As prices of DDR5 memory kits set records in the U.S., in Europe they begin to show signs of descend. At least, this is what a DDR5 pricing graph published to a renowned PC enthusiasts community is meant to show. We also analyzed the pricing of several DDR5 kits from prominent suppliers in Germany, and we can certainly say that these kits cost less than they used to cost just weeks ago.

The chart allegedly depicts aggregated pricing of an 'average' 32 GB DDR5 kit across the European Union from late July 2025 to February 2026. Prices hovered around €95 (minimum, green) – €100 (average, blue) through early autumn, then began climbing sharply in October, accelerating through November and peaking in early February at roughly €430 – €470 on average, with minimum prices slightly lower. Toward the end of the period, both lines trend downward, which may either indicate a modest correction after the spike or an actual drop in prices due to certain factors.

While the graph from deserves attention, it lacks clarity and details (which kits, which countries, retailers, is VAT included, etc.), so we decided to do our own price trend checks of five popular 32 GB DDR5-6000/6400 dual-channel kits* from renowned brands like Crucial, Corsair, G.Skill, Kingston, and Patriot in Amazon Germany using the CamelCamelCamel service.

Among the 32 GB DDR5-6000/6400 kits that we checked, only two models — from Corsair and Kingston — demonstrated steep declines: from around €480 in early February to around €425 now for Corsair and from around €550 in early January to €463 at press time for Kingston. Nonetheless, all memory kits that we checked are now priced below their peaks several weeks ago. Note that all retail prices in Europe include VAT, unlike retail prices in the U.S.

We also checked price trends for the same 32 GB DDR5 kits in the U.S., and while the prices are far from where they were in September, some of them (G.Skill, Patriot) are also showing a modest correction, though we certainly cannot say that they are heading downwards.

While $400 is certainly way too high for a 32 GB DDR5-6000 memory kit in 2026, we are not going to see prices decline to normal levels due to shortages of memory chips, which is going to happen either when excessive demand for all kinds of memory drops, when new DRAM production capacities come online in late 2026 – 2027, or when DRAM makers transit to more efficient process technologies. Yet, the signs of correction clearly show that the retail DDR5 kits' prices are way too high, which affects demand significantly enough for retailers to slash their price tags.

*We used the following kits for our checks, as memory prices currently depend on supply, we did not specify based on whether the kits feature AMD Expo or Intel XMP profiles:

• Crucial Pro 32 GB DDR5-6400 (CP2K16G64C38U5B)
• Corsair Vengeance RGB 32 GB DDR5-6000 (CMH32GX5M2E6000C36)
• G.Skill Flare X5 Series 32 GB DDR5-6000 (F5-6000J3636F16GX2-FX5)
• Kingston Fury Beast 32 GB DDR5-6000 (KF560C30BBEK2-32)
• Patriot Viper Venom 32 GB DDR5-6000 (PVV532G600C36K)

For the first time since announcing its seismic decision to kill its consumer SSD and memory brand Crucial, Micron has addressed the notion that it is leaving consumers behind in a new interview. The company also warned that despite breaking ground on new memory fabs, we shouldn't expect to see meaningful output impacting memory supply until at least 2028.

Micron's push back against the criticism of its decision to shutter Crucial comes by way of a WCCFTech interview with Christopher Moore, Micron's VP of Marketing, Mobile and Client Business Unit. The outlet wasted no time pressing Moore in Micron's controversial, but not entirely unexpected, decision to shutter the Crucial brand late last year. In early December, the company said that it plans to wind down its consumer business by the end of next month (January), reallocating its output and time to enterprise-grade DRAM and SSDs for AI buildouts.

Moore was asked if memory suppliers were inclined towards catering to the AI sector, "leaving consumers behind" as a result. "Well, first I would want to try to help everybody understand that the perception may not be exactly correct, at least from our point of view," Moore said. He stated that while he would "never want to tell someone what to think or that they're wrong... our viewpoint is that we are trying to help consumers around the world." Moore then cited Micron's sizeable businesses in the client and mobile market. Moore hinted that Micron is still technically serving consumers by supplying LPDDR5 to OEMs like Dell and Asus for inclusion in laptops, amongst other things. While this is technically correct, the news will be of little comfort to the DIY community and enthusiasts facing colossal price increases.

While the report claims Micron is in contact with "every single PC brand out there", the company simply cannot afford to ignore AI demand.

Some hope on the horizon for PC builders and the consumer sector is increased DRAM capacity in the supply chain. Micron recently announced it would begin work on a $100 billion New York 'megafab', where it plans to produce 40% of the company's overall DRAM output by the 2040s. Moore also noted its upcoming ID1 facility in Idaho, which is scheduled to come online in mid-2027. However, he warned that it will be 2028 before we see "real output, meaningful output," in its DRAM supply chain. Don't forget that Micron can't even keep up with current demand. Its CEO said in December that it can only meet half to two-thirds of demand, meaning that even the upcoming new capacity will initially go towards making up shortfalls for existing demand. As such, while 2028 might mark the first meaningful dent Micron makes in DRAM supply, it could be months more before consumers start to see any shift in pricing for PC builds.

'Twas the festive season but Santa didn't know how
to deal with the new RAM apocalypse, now!
Who could've predicted this horrid event
in which memory costs rose 500 percent?

He didn't expect this holiday twist —
It's memory-intensive to create his list!
Even Crucial got killed in this big AI boom.
(This boom smells as funky as Zotac's perfume).

But he still had some systems lying around.
Surely one of them would have to be sound!
He had an XPS, Optiplex — oh, what the hell?
Which one is which? They're all just a Dell.

So instead, he switched to an Apple Silicon Mac
and played Cyberpunk to forget the setback.
Then he jumped to a Windows 10 system he'd moved to Linux
and entered his password: 123456.

❄️❄️❄️

But none of these systems made the list the right way.
They all fell like Steam on Silksong's launch day.
So he figured he'd do these holidays with a twist:
He'd simply put everyone onto his list.

Compiling this data would take quite a while.
Santa had what Torvalds called "random turd files."
He found several old lists made in Microsoft Word
and even more data he had stored on a bird.

To collate the data, he had the elves convene,
using prototypes of the new Steam Machine,
and old rigs he updated with a bit of skill,
and a few lucky builds that he found at Goodwill.

Now that the rigs were doing the heavy lifting,
It was time for Santa to put his mind to gifting.
Like 3D-printed loafers that are beyond description
And some great Tom's Hardware Premium subscriptions!

❄️❄️❄️

Gold RTX 5090's would be all the rage.
There were VPNs for people verifying their age.
Elves made hair clips out of old M1 Pros,
And for Intel, a new permanent CEO.

There were non-binary RAM kits on sale at some shops,
and 5070 Ti's with all of their ROPs.
Nintendo Switch 2s without stapled screens,
and TSMC sneakers that go great with your jeans.

"Hey Santa!" cried one elf, looking over his shoulder.
"Why are all these anime girls in your temp folder?"
Santa explained that his JPGs weren't a crime,
But blamed them on a bug in XPG Prime.

With that unfortunate question answered,
He called Rudolph, and Comet, and Dancer, and Prancer.
"Can you stop by the store and try to obtain
some new Xboxes before the price rises again?"

❄️❄️❄️

Then Mrs. Claus came downstairs in shock.
"Santa, look what I found under stockings and socks!"
He felt his jaw drop like it was full of rocks —
It was a whole bunch of RAM he had never unboxed!

Santa couldn't help it — he let out a snort.
Like RDNA 1 and 2, he could have more support!
But he and his crew were saving the day,
with existing computers, the old-fashioned way.

And Santa and friends were all having a ball.
It was the best holiday lead-up that he could Recall.
He might even have extra time to clean up his house
to smell fresher than Asus's aromatic mouse.

It took all of Santa's brains and all the elves' might,
But they had a list, and had gifts, and were ready for flight.
To ensure they had the fuel to get this year's trip right,
They stopped by a Denny's for some Breakfast Bytes.

No matter what you’re celebrating this year,
Hanukkah, Kwanzaa, or Christmas with cheer;
We at Tom’s Hardware wish you the best.
We’ll be at our benches, running our tests.

🎅🎄🎁☃️❄️ 🎅🎄🎁☃️❄️ 🎅🎄🎁☃️❄️

The Tom’s Hardware staff first published a holiday poem in 2014 on Christmas Eve. It was updated a little bit and published again each year at the same time. The poem was given a complete overhaul in 2018 and has been rewritten with new stories and references every year since. For more holiday fun, see the versions from 2019, 2020, 2021, 2022, 2023, and 2024.

In keeping with the spirit of times, Micron on Wednesday announced plans to wind down its Crucial consumer business worldwide by the end of February 2026. The company is reallocating its output and investments to enterprise-grade DRAM and SSD products amid growing demand from the AI sector.

Micron will continue shipping Crucial-badged consumer products through retailers, online stores, and distributors until the end of its fiscal second quarter, which concludes in late February 2026. After that point, Micron will no longer supply the consumer channel with products under the Crucial name, but will continue to ship its Micron-branded enterprise portfolio, which will remain available through commercial and server-focused partners.

After Micron ceases to ship its Crucial devices, it will continue honoring warranty obligations and technical support for existing Crucial products. Customers who already own Crucial-branded memory modules, SSDs, and other products will continue to receive service coverage well after shipments stop, so the decision will not leave installed hardware unsupported.

As expected, Micron is abandoning its consumer business due to reallocation of its 3D NAND and DRAM output and production capacity to enterprise-grade SSDs, high-bandwidth memory (HBM) for AI accelerators, and server-grade memory modules.

"The AI-driven growth in the data center has led to a surge in demand for memory and storage," said Sumit Sadana, EVP and Chief Business Officer at Micron Technology. "Micron has made the difficult decision to exit the Crucial consumer business in order to improve supply and support for our larger, strategic customers in faster-growing segments."

Micron established its Crucial brand some 29 years ago, in 1996, when the enthusiast-grade hardware market began its rapid growth. Over time, Crucial got particularly successful in retail, where its name became synonymous "with technical leadership, quality and reliability," as Micron puts it. However, the current market situation is by far not favorable for these types of products, so instead of shrinking its Crucial portfolio, the company decided to wind it down entirely, albeit without selling off the Crucial brand (at least for now).

There are several reasons behind Micron's decision.

Firstly, client memory modules and SSDs sit at the lowest-margin end of Micron's portfolio as they compete in highly volatile, price-competitive, and promotion-driven market. Even though the Crucial and Ballistix brands still matter, they are squeezed between high-end enthusiast brands and low-end consumer brands, which makes their evolution difficult. By contrast, data center and enterprise products lock in long-term contracts, higher ASPs, and more predictable demand.

Secondly, the supply environment has changed permanently. AI infrastructure requires every single wafer with memory it can consume, something that has never happened with any industry megatrend previously. This means that every wafer Micron assigns to consumer parts is a wafer not going to a hyperscaler or enterprise contract. As a result, keeping a consumer line would directly limit Micron's ability to fulfill orders from its largest customers, which is a risk for profits and strategic relationships.

Thirdly, even a small consumer business would still require a minimum viable supply chain, including product development, firmware validation, compliance testing, sales teams, retail relationships, and global warranty operations. Such fixed costs barely shrink when volume shrinks, so a reduced consumer business would still burn resources while losing the economies of scale that make the segment viable.

As a result, it strategically makes far more sense to wind down consumer operations entirely and free production capacity, R&D, and product engineering resources for premium products like HBM4/HBM4E/C-HBM4E, enterprise drives, and high-density server memory modules.

Micron indicated it will try to reduce the impact of the decision on employees through internal reassignments into existing vacancies elsewhere in the company.

Crucial has recently launched the Pro Overclocking DDR5-6400 C32, its latest memory kit designed to compete with the best RAM available in the retail market. Although higher data rates are accessible, many memory vendors focus on DDR5-6400 due to its optimal balance between performance and affordability. Crucial already has several DDR5-6400 memory kits in its arsenal, including a CUDIMM variant; however, these featured conservative memory timings. Recently, Crucial has done some extra binning and optimization, now bringing a DDR5-6400 C32 memory kit to the market. The new addition to the brand's Pro Overclocking series caters to enthusiasts seeking improved performance.

Although the new memory kit retains what Crucial refers to as an origami-inspired heat spreader, the company has made subtle modifications to the design. Instead of a uniform solid white or black color, Crucial has introduced a camouflage pattern with a highly gamer-oriented aesthetic. The aluminum alloy heat spreader remains available in either black or white finishes, although Crucial employs fancier names such as "Stealth Matte Black" and "Snow Fox White," respectively.

Like other memory kits from the Pro Overclocking lineup, there is minimal marketing on the memory modules. The Micron logo remains visible; however, this time, the Crucial logo is presented as a product of CNC machining with a diamond-like finish to enhance light reflection. The Pro Overclocking memory modules have a height of 1.38 inches (35 mm), fitting into the low-profile category.

The memory kit consists of two 16GB memory modules. Each module features a single-rank design with eight integrated circuits (ICs), arranged on a single side. This represents the first occasion that we're seeing these Micron 5QH75 D8HKZ ICs, which, according to the SPD dump, appear to be Micron's new H-die ICs. Given that the memory module is 16GB, each IC has a capacity of 2GB. Meanwhile, the 12=BL 1G1 power management IC (PMIC) indicates to us that it's from Richtek.

As with any overclocked memory kit, the memory modules default to JEDEC's specifications. In this case, it's DDR5-5600 with timings at 46-45-45-90. Both Intel XMP 3.0 and AMD EXPO support are present in Crucial's Pro Overclocking memory modules. The primary profile is configured for DDR5-6400 with timings specified as 32-40-40-103 at 1.35V. Conversely, the secondary profile is designated for DDR5-6000 with timings of 36-38-38-80 and a DRAM voltage of 1.35V. See our PC Memory 101 feature and How to Shop for RAM story for more timings and frequency considerations.

Comparison Hardware

The Intel system features the Core Ultra 9 285K processor and the MSI MEG Z890 Unify-X motherboard with firmware version 7E20v1A60. Conversely, the AMD system utilizes the Ryzen 9 9900X processor and the MSI MPG X870E Carbon WiFi motherboard with firmware version 7E49v1A64. The Corsair iCUE Link Titan 360 RX LCD CPU liquid cooling solution efficiently sustains optimal temperature levels for both the Arrow Lake and Zen 5 processors.

The MSI GeForce RTX 4080 16GB Gaming X Trio addresses the more graphics-intensive workloads, ensuring that our gaming RAM benchmarks do not have a graphics bottleneck. TeamGroup's A440 Lite PCIe 4.0 SSD strikes a balance between performance and storage capacity, offering 2TB of high-speed storage that reaches speeds of up to 7,400 MB/s, ideal for Windows 11 24H2 installations, benchmarking software, and gaming.

Meanwhile, the Corsair RM1000x Shift ATX 3.0 power supply provides a dependable and sufficient power source for our testing systems, directly supplying the GeForce RTX 4080 with a native 16-pin (12VHPWR) power cable. Ultimately, the Streacom BC1 open-air test bench is capable of housing all of our hardware.

Intel Performance

Crucial's latest memory kit demonstrated commendable speed on the Intel platform. It faced no significant issues securing a top position, only trailing behind the Dominator Platinum RGB DDR5-6400 C38 in application performance. Conversely, the Pro Overclocking's gaming performance was comparatively modest and ranked at the lower end of the competitive spectrum.

AMD Performance

Memory kits do not consistently demonstrate identical performance across various platforms; therefore, testing is conducted on both Intel and AMD motherboards. The Pro Overclocking DDR5-6400 C32 memory kit did not perform notably well on the AMD platform and was generally slightly slower than the existing Pro Overclocking DDR5-6400 C38 memory kit.

Overclocking and Latency Tuning

Our best overclocking result on the Pro Overclocking DDR5-6400 C32 was DDR5-6600, maintaining identical memory timings while increasing the DRAM voltage from 1.35V to 1.45V. The result was marginally lower than the DDR5-6400 C38, which was anticipated, given that we were able to preserve the CAS Latency (CL) of 32 clock cycles as opposed to 38 clock cycles on the latter.

Lowest Stable Timings

The factory timings for the DDR5 6400 memory kit were already approaching the stability threshold. Efforts to reduce various memory timings were unsuccessful. Additionally, increasing the voltage to 1.45V, which is 7% above the default setting, did not allow us to lower any of the memory timings.

Bottom Line

The Pro Overclocking DDR5-6400 C32 exhibited a varied array of results. In comparison to the DDR5-6400 C38 variant, Crucial refined the CL timings from 38 to 32 clock cycles; however, this adjustment was accompanied by an increase in tRAS from 84 to 102 clock cycles. The trade-off balances out the performance, which explains why the C32 version is not necessarily consistently faster than the C38 version on every platform. With DDR5 memory, all timings are important, and you should not be fixated solely on the CL timing.

Crucial markets the Pro Overclocking DDR5-6400 C32 memory kit with two distinct pricing options based on the selected color trim. The Stealth Matte Black variant is priced at $174.99, whereas the Snow Fox White variant incurs a $12 premium, resulting in a retail price of $186.99. For comparison, the Pro Overclocking DDR5-6400 C38 memory kit currently sells for $129.98. Regardless of the choice, Crucial provides a limited lifetime warranty on all Pro Overlocking memory kits.

If you're looking to upgrade the storage in your desktop PC or laptop, this budget-oriented SSD from Crucial is now at its lowest-ever price. The smaller P310 M.2 2230 is one of the best SSDs for the Steam Deck and other Windows gaming handhelds, and this larger 2280 version of the P310 is just an updated version of the same drive with a larger form factor. So, if you want an inexpensive solution for the shrinking SSD space of your game library, the Crucial P310 could be the answer. This version also includes a heatsink for additional cooling.

If you want to get your hands on one of these drives, pick up the Crucial P310 (1TB) with heatsink for just $74 - its lowest-ever price on Amazon. If you'd like more detailed information on this Gen 4 SSD, look at our review of Crucial's P310 M.2 2280 SSD.

This heatsink version of the P310 from Crucial is the 2280 form-factor variant. Although the standard P310 is perfect for popping into a laptop, thanks to its excellent power efficiency, this version is not advised due to its height and space constraints in most laptop drive bays. The Crucial P310 on sale has a 1TB capacity and uses Micron's 232-layer QLC NAND flash memory. The controller used is the Phison E27T, which helps to get the P310 to run with sequential read/write speeds of 7,100/6,000 MB/s.

Even though the P310 has excellent performance, rivaling the speeds of many of the top Gen 4 SSDs, the sustained performance of the P310 isn't the best because of its use of QLC NAND instead of TLC. The less available space on the drive, the quicker the performance deteriorates. However, as long as you can find this drive at the right price, the performance of the P310 outweighs its use of QLC. It's not worth it if you can find a TLC drive with a heatsink for less money and better performance.

Don't forget to look at our Amazon coupon codes for April 2025 and see if you can save on today's deal or other products at Amazon.

Toto is mainly known for its advanced toilets and sanitary ceramics. However, the company's expertise in ceramic production can also be applied to semiconductor manufacturing. Since the 1980s, Toto has made electrostatic chucks (e-chucks) indispensable in modern semiconductor manufacturing. According to Nikkei, operating profits from these products are expected to exceed $100 million this year.

In contemporary semiconductor manufacturing, electrostatic chucks (ESCs) securely hold a silicon wafer (or other substrate) in place using electrostatic forces rather than mechanical clamping or vacuum-based methods. The ESC is a key component in many steps of chip production, including EUV lithography steps, plasma etching, chemical vapor deposition (CVD), physical vapor deposition (PVD), and other steps requiring precise wafer positioning and minimal contamination.

While e-chucks have traditionally been used for CVD, PVD, and plasma etching, but not for DUV lithography steps as they are carried out in an ambient environment or immersion fluid, and therefore, a vacuum system under the wafer is good enough to maintain wafer flatness and position. However, with EUV, things are different. EUV lithography operates at very short 13.5nm wavelengths and requires a high-vacuum environment to prevent absorption of EUV light. Therefore, chipmakers use e-chucks instead of vacuum chucks as they are easier to use in such environments. Also, they can provide more uniform clamping force, reduce stress, minimize distortion, and improve overlay and critical dimension (CD) control.

It takes over 4,000 steps to process a wafer and make a chip. Usage of EUV steps has been increasing in recent years, just like steps requiring precise wafer positioning, so usage of ESCs has been on the rise, driving Toto's revenues and profits. Ceramics used for e-chucks must be both strong and resistant to cracking. Toto has developed materials with uniform properties, applying its expertise in molding and firing from its long history in toilet manufacturing. However, competition in this field is growing. In e-chucks, Toto faces Shinko Electric Industries, which has strong ties to chip equipment manufacturers and Applied Materials.

To strengthen its position, Toto has invested heavily in manufacturing. In 2020, the company spent ¥11.8 billion constructing a ceramics production facility in Oita, Japan. Between April 2020 and April 2024, it increased its ceramics production workforce by around 20%. According to company president Noriaki Kiyota, plans for a new factory are also being considered.

For fiscal 2024, Toto expects its ceramics division to generate ¥20 billion ($130 million) in operating profit, with a nearly 40% margin. This is far higher than the company’s projected overall margin of 7%. By fiscal 2026, it aims to increase that figure to ¥25 billion and expand its product range.

Toto now looks beyond wafer processing to downstream steps like cutting and packaging. According to the report, three-dimensional chip stacking requires new base substrate technologies, and ceramics are expected to play a more prominent role in these processes.

"Even if there are ups and downs, semiconductors will undoubtedly keep growing exponentially," Noriaki Kiyota, president of Toto, told Nikkei.

If you find yourself needing some extra storage for your games, or for moving data between devices outside of a network connection, then having a portable storage solution is a nice thing to have. Smaller is better for portability, but it doesn't also have to equal less capacity. Today's deal is on a 2TB external SSD that you can connect via USB C to all manner of devices, from your games console to your mobile phone, to the usual laptops and desktop PC, or even your camera's.

Head to Amazon for today's deal and pick up the portable Crucial X9 Pro (2TB) external SSD for just $119. It's not the lowest price ever -- that was $87 back in 2023 -- but, you aren't going to see prices like that again for storage unfortunately, especially outside of the larger sales events like Prime Day or Black Friday. The MSRP list price of this drive is $179 with the Camelizer Amazon price checker showing prices steadily fluctuating between $119 and $155.

The X9 Pro measures around 2.5" and comes in a solid and rugged plastic chassis that's rated for IP55 dust and water resistance, It also advertises a resistance to drops from heights no greater than 7.5 feet, which should more than cover the average use-case scenario. Inside the chassis is a small SSD with speedy read and write speed transfer rates up to 1050MB/s over USB 3.2. The X9Pro uses a Type-C USB connection and cable.

I have one of these handy drives and use it to record video from my phone. Large 4K video files quickly soak up my phone's valuable storage so offloading it to the X9 Pro keeps my mobile speedy and makes it easier and faster to connect to my PC for video editing. The small size means I can just keep it in my pocket while I'm out and about, and always have it to hand when I need it.

Don't forget to look at our Amazon coupon codes for January 2025 and see if you can save on today's deal or other products at Amazon.

The man who exploded a Tesla Cybertruck outside of a Trump hotel on January 1st in Las Vegas used ChatGPT to plan his blast, according to new findings from the Las Vegas Metropolitan Police Department. In a recent press conference by the police department and partners in the ATF and FBI, specific prompts submitted to ChatGPT were revealed, along with information that some specific prompts returned information that were crucial in planning the explosion.

Matthew Livelsberger, the man who blew up the Cybertruck shortly after killing himself, asked ChatGPT a long list of questions about the plan over one hour in the days leading up to the event. These include questions about sourcing the explosives used in the blast, the effectiveness of the explosives, whether fireworks were legal in Arizona, where to buy guns in Denver, and what kind of gun would be needed to set off the chosen explosives.

Most importantly, Assistant Sheriff Dori Koren confirmed that ChatGPT was instrumental in making the blast plan work. ChatGPT returned prompts to Livelsberger which revealed the specific firing speed a firearm would need in order to ignite his chosen explosive. Without ChatGPT, the incident may not have been as explosive as it proved to be, though the ATF also confirmed in the conference that not all explosives detonated as were likely intended to in the initial blast.

"We knew that AI was going to change the game at some point or another, in really all of our lives," shared LVMPD Sheriff Kevin McMahill. "This is the first incident that I am aware of on U.S. soil where ChatGPT is utilized to help an individual build a particular device, to learn information all across the country as they're moving forward. Absolutely, it's a concerning moment for us."

McMahill was also not aware of any governmental oversight or tracking which would have been able to flag the 17+ prompts asked of ChatGPT, all relating to sourcing and detonating explosives/firearms, submitted within a one hour period.

While full info on the ChatGPT prompts has not yet been released by the Las Vegas police, the prompts shown in the press conference were straight-forward and written in simple English, without traditional backdoor terms used to "jailbreak" ChatGPT's content detection system. While this usage of ChatGPT violates OpenAI's Usage Policies and Terms of Use, it is not clear at this time whether safeguards or content warning violations were raised in Livelsberger's use of the LLM.

OpenAI and the Las Vegas Metropolitan Police Department have not yet responded to requests from the press for further information on the usage of ChatGPT in the event; we will update our coverage as more becomes available.

The Crucial T500 4TB is here, for those who find that 2TB of storage just isn’t enough these days. Using Phison's E25 controller, the T500 has a power-efficient architecture with DRAM, now available as a 4TB pseudo-budget SSD. Some of our favorite PCIe 4.0 drives simply haven't bothered with higher capacities, like the SK hynix Platinum P41, so attention has turned to the T500. Crucial promised a 4TB model a year ago, and it has now delivered: the same great taste but in an extra large tin.

For the most part, nothing has changed but the capacity. The T500 was heralded as a top notch PCIe 4.0 SSD that could match or beat any other drive in its class, all while being more power-efficient. Half the channels with the same amount of fun, and unlike some of its competitors, it kept the DRAM intact. It’s a great successor to drives like the SK hynix Gold P31 for laptops, and it also works great for desktops and consoles. What’s not to love?

Perhaps the most obvious potential misstep is the price. In some respects this is supposed to be a “budget” SSD, even though its performance is in line with heavy hitters. Often, it manages to look like a WD Black SN850X or Samsung 990 Pro substitute, but really it just brings DRAM to a potentially less expensive drive, taking on the likes of the Teamgroup MP44. This becomes more obvious when looking at its sustained write performance irregularities, although for the most part you’re not going to notice that in daily use. Still, its price is higher than it should be at 4TB, which keeps it from being the best option in many cases, even if it delivers exemplary all-around performance.

Crucial T500 Specifications

The 4TB Crucial T500 is currently priced at $309 for the non-heatsinked version and $301 for the heatsink model. Those prices can and will change — a week or two back, the non-heatsink model was going for $289 and the heatsink variant cost $319. Whether you want the heatsink or not, check the current pricing, as you can easily remove the heatsink if you really don't want it. If the bare drive ends up costing less (as it logically should), a variety of M.2 heatsinks will get the job done, or you can use your motherboard's M.2 heatsink if it has one. Do note that the 4TB model isn't a single-sided design, however.

The Crucial T500 4TB can hit up to 7,000/6,900 MB/s for sequential reads and writes, and up to 1,050K/1,350K random read and write 4K IOPS. These numbers are slightly lower than the 2TB SKU, and the 1TB SKU as well for that matter. The reason is the hardware. The controller in use has four chip enable (CE) signals for each of the four channels, giving sixteen CE total. As the drive is using 1Tb dies there are 32 dies in total present, more than the CE count, which add some overhead. We’ll point out physical layout changes below that relate to this configuration. If you’re concerned about performance loss, we would state that these dips are not significant in everyday use and that this is a capable 4TB SSD.

The 4TB T500 has a warranty for five years and up to 2,400TB of writes, which is right in line with the smaller SKUs. The drive also supports TCG Opal 2.01 for hardware encryption.

Crucial T500 Software and Accessories

The 4TB T500 has the same software available as discussed in the original Crucial T500 2TB review — Crucial’s Storage Executive SSD toolbox and Acronis True Image. The toolbox has typical features including health management, secure erase, firmware updates, and encryption support. Acronis True Image is for imaging, cloning, and backing up your data, which is useful especially when upgrading your primary drive.

Crucial T500: A Closer Look

One of the biggest questions we had was whether the 4TB T500 would be single-sided. The answer, unfortunately, is no. It’s theoretically possible to stack 16 dies per package (16DP) and 1Tb flash is in abundance, but there are reasons to prefer four 8-die packages (8DP/ODP). Due to the T500’s layout, which includes a diagonal and sizable controller as well as a package of DRAM, the drive must be double-sided at 4TB when using 8DP.

Plenty of other companies have put 4TB on a single-sided drive, though often that's with DRAM-less drives. But the Samsung 990 Pro 4TB as an example puts four 1TB NAND packages on one side of the stick, and that can be better for cooling in a laptop or console. The 45-degree rotation on the Phison E25 controller would already make it difficult to do four NAND packages on a side.

But the flash density and performance also requires additional ICs on the PCB that we'll discuss below, this is unfortunately the easier way for Crucial to go, and results in a double-sided form factor. This might preclude the drive’s use in some systems at this capacity, although most laptops today can handle double-sided drives.

The hardware is effectively the same as before with the Phison E25 controller, LPDDR4, and Micron’s 232-Layer TLC flash. One thing that stands out is the presence of additional ICs listed as DDR4000 by Texas Instruments. This is not the first time we’ve seen these. They were also present on the Corsair EX100U and Sabrent Rocket nano V2 portable drives.

On this drive there are a total of four of these ICs, one for each of the E25’s four channels. These could be used for many things, but on SSDs they are commonly used to keep signaling within acceptable parameters for the 2,400 MT/s flash at hand. This becomes more of an issue as more flash is added, and for 4TB — which requires 32 dies, or 8 dies per channel — this is a sure way of maintaining performance with the extra loading. The previous drives we tested had these at 2TB with slower flash, but those drives were only using two-channel controllers. The use of these ICs on the 4TB T500 further encourages a double-sided design due to space constraints.

MORE: Best SSDs

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

We’re starting out by looking to the future with the PCIe 5.0 Crucial T700 and at the past with the early adopter Sabrent Rocket Q4. Moving forward, 4TB drives will become more commonplace where once you looked to QLC flash to make things work. The T500 will inevitably be compared with top-tier 4.0 drives like the Samsung 990 Pro — which updated its hardware with the 4TB launch, and the drive remained single-sided — but is more fairly compared to Phison E18-based legends. These drives used the first high-end 4.0 hardware available and remain good picks, particularly for sustained workloads, albeit with higher power use and worse efficiency. These include the Sabrent Rocket 4 Plus-G, the Seagate FireCuda 530, the Teamgroup Carda A440 Pro, and the Netac NV7000.

The T500 also has to face off against other budget 4TB drives, though these are DRAM-less designs. The Addlink A93 and Lexar NM790 remain popular choices with often lower price tags. Or, you can go with QLC flash to save some money, as with the Patriot VP4300 Lite or more entry-level WD Blue SN5000. QLC flash still has a poor reputation, but for extra inexpensive storage it remains a reasonable compromise. 

Trace Testing — 3DMark Storage Benchmark

 Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams. Future gaming benchmarks will be DirectStorage-inclusive and we also include notes about which drives may be future-proofed.

The T500 is the fastest PCIe 4.0 SSD on the market when it comes to 3DMark. This does not translate to significantly faster loading times or higher FPS, but it does mean it makes a great game drive. Its controller also has Phison’s DirectStorage-optimized I/O+ firmware, so it’s future-proofed for upcoming games. Drives built on Phison’s newer E31T controller will have higher bandwidth and power efficiency, but the T500 still stands out by having DRAM. This could make for more consistent performance with DirectStorage titles, but that remains to be seen.

Trace Testing — PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices. The results are particularly useful when analyzing drives for their use as primary/boot storage devices and in work environments. 

The T500 is also the fastest 4.0 drive in PCMark 10. It’s an excellent all-around drive that can work well as the sole drive in your PC — or any system for that matter. For laptops, PCIe 5.0 SSDs remain out of reach, although a few built on the E31T are coming very soon. The T500 remains competitive, though, thanks again to its DRAM, the presence of which some people prefer. This makes the T500 the natural SK hynix Gold P31 successor and more or less irreplaceable for the time being if DRAM’s your thing. Application performance should be excellent across-the-board. 

Console Testing — PlayStation 5 Transfers

The PlayStation 5 is capable of taking one additional PCIe 4.0 or faster SSD for extra game storage. While any 4.0 drive will technically work, Sony recommends drives that can deliver at least 5,500 MB/s of sequential read bandwidth for optimal performance. In our testing, PCIe 5.0 SSDs don’t bring much to the table and generally shouldn’t be used in the PS5, especially as they may require additional cooling. Check our Best PS5 SSDs article for more information.

Our testing utilizes the PS5’s internal storage test and manual read/write tests with over 192GB of data both from and to the internal storage. Throttling is prevented where possible to see how each drive operates under ideal conditions. While game load times should not deviate much from drive to drive, our results can indicate which drives may be more responsive in long-term use. 

Our performance results for the 4TB T500 on the PS5 are a little lower than we expected. In our original T500 review we did notice some discrepancies in our PS5 testing, but Crucial was aware of this and later put out a firmware update. The 4TB model comes with new firmware pre-applied, but the PS5’s firmware has also since been updated in a way that does reduce our scores a little bit on all drives. Even with this considered, the 4TB T500 underperforms to a small degree, but not enough that you shouldn’t consider it very usable for the console. Any relatively recent PCIe 4.0 SSD should generally work fine in the PS5, with price per GB often being the determining factor.

Transfer Rates — DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We write 31,227 files of various types, such as pictures, PDFs, and videos to the test drive, then make a copy of that data to a new folder, and follow up with a reading test of a newly-written 6.5GB zip file. This is a real world type workload that fits into the cache of most drives. 

Our DiskBench data fits in just about any pSLC cache these days, so the results here are an attempt to show real world transfer performance before the cache runs out. For larger, sustained transfers, see our write saturation testing below. We mention this because the T500 performs excellently in DiskBench but still has some issues when its cache gets depleted. We do think that, in the real world and especially with the size of the cache for the 4TB T500, this drive should perform excellently in almost all cases.

Synthetic Testing — ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes and at different queue depths for both sequential and random workloads. . 

The T500’s ATTO performance is good, with only two dips worthy of notice. The first is with sequential reads at the 2MiB block size. This is possibly because of a boundary that exists with flash interleaving, where essentially you can have four six-plane dies for each of four channels for peak performance. Given the typical plane size of 16KiB, this may create a boundary between 1MiB and 2MiB. In the real world, this is not a significant issue.

There are also dips at 2KiB for both reads and writes, although not nearly as bad as with the T700. While consumer flash today has 16KiB physical pages, it’s not uncommon to have these broken down into 4KiB logical pages. This matches typical cluster and physical sector sizes, although a logical sector size of 512B is normal. This would leave 2KiB I/O as a subpage, which has interesting ramifications for the SSD controller and its DRAM mapping memory. This performance dip is likely deceptive as Phison’s work with enterprise storage, where 2KiB I/O may need optimized handling, informs its consumer controller design. As the T500 has plenty of DRAM and ideally you have some queue depth for that type of I/O, this result may be an anomaly.

CDM gives us a fuller picture by revealing random 4KB QD1 performance — the T500 is very good here — as well as sequential throughput. The T500 is as fast as any other high-end PCIe 4.0 SSD, so no surprises there. Given the relatively minute differences, you’re not missing anything by going with the T500. You can safely buy it expecting more or less the best performance possible for a drive of its class. 

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of pseudo-SLC (single-bit) programmed flash that absorbs incoming data. Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC (three-bit) or QLC (four-bit) flash. Performance can suffer even more if the drive is forced to fold, which is the process of migrating data out of the cache in order to free up space for further incoming data.

We use Iometer to hammer the SSD with sequential writes for over 15 minutes (we did two hours in this case) to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds. This process shows the performance of the drive in various states as well as the steady state write performance. 

In our original review of the T500 we noticed some inconsistencies in write performance, particularly when the pSLC cache was exhausted. Crucial later put out a firmware update intended to improve in this area, but in our testing things were still relatively irregular. The 4TB, with its relatively large cache, still seems to suffer in this area, although technically speaking the base write performance of the TLC flash is excellent. This is revealed in the steady state write speeds which are supposed to yield the native flash performance, more or less. 3.5 GB/s for a four-channel controller like the T500’s would be exceptionally fast.

The 4TB T500 begins by writing at around 6.5 GB/s in the single-bit pSLC mode for up to almost five minutes. This is a huge cache and is one of the reasons post-cache performance looks so slow. The drive is eventually forced to empty the cache to free up the full amount of space and this naturally bottlenecks things. In this slower folding mode, where the drive is forced to copy data from pSLC to the native TLC flash while also juggling incoming writes and migrating blocks from pSLC to TLC mode, the drive writes at under 600 MB/s. This is rather sluggish, but far from the worst we’ve seen. QLC-based drives even at 4TB can only achieve half this speed or lower.

All of this makes for some deceptive results. We know the Phison E25 controller is designed for up to about sixteen dies of interleaving, given that it’s a four-channel and not eight-channel design. We also know how fast this flash is, so realistically it wouldn’t plod away at a 3.5 GB/s average, either. A more reasonable target would be what we see on the A93 and NM790, which uses YMTC’s comparable 232-Layer TLC flash. That result would still be very good. We’re instead seeing an average of the pSLC and folding modes, which added together and divided by two is 3.5 GB/s or so.

So, what does this mean? Well, the T500 is simply not very consistent when dealing with larger writes. If the drive is fuller the cache will be smaller so the size of the transfer required to make this happen will change. For the most part, you should never see poor performance from this drive as the cache does a good job for 99% of workloads. You shouldn’t be buying this drive for anything super intense, like massive sustained writes. That said, these results can make people nervous, in part because many drives in the past have had pSLC issues in the long-run. Getting “stuck” in TLC mode isn’t as bad as it looks with our results, though, but we can understand why you might want something that performs more consistently here. 

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre stock storage. Desktops may be more performance-oriented with less support for power-saving features, so we show the worst-case.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption but performance-per-watt, or efficiency, is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

For temperature recording we currently poll the drive’s primary composite sensor during testing with a ~22°C ambient. Our testing is rigorous enough to heat the drive to a realistic ceiling temperature.

The 4TB T500 is slightly worse in all metrics when compared to our original 2TB sample: higher idle, peak, and average power consumption, with somewhat lower overall efficiency. This isn’t an exact comparison as there have been firmware updates since then, so we did not include the original results. While packing in more flash can reduce power efficiency at some point, modern drives are very good at switching dies to use as little power as is possible. The 4TB does have the DDR switch/multiplexers — NAND flash runs on some similar principles as DRAM memory with, say, NV-DDR3 — which could eat a small amount of power. Regardless, the drive is still fairly efficient, particularly as it has DRAM, and power consumption is not a huge concern.

In our testing the drive hit a maximum temperature of 57°C. This is well below the throttling threshold, so any well-cooled system should have no problems running this drive. It will also be fine in a laptop or PS5. You can definitely add a heatsink, though — or buy the model that comes with one — for peace of mind. This is not a drive that should get too hot as you'll want to avoid massive sustained workloads anyway.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Crucial T500 Bottom Line 

If you’ve been looking for a 4TB T500, here it is. Crucial always promised to offer this capacity — a capacity that’s often clamored for, since it feels like many drives top out at 2TB these days — and it has delivered. It’s a power-efficient drive with DRAM, which is part of what drives the desire to get it at 4TB. Great performance for a laptop drive? Sounds like an amazing product. Although, this drive is double-sided and has some other caveats that keep it from being the perfect dream come true.

All-around performance remains excellent, and in fact it’s the fastest PCIe 4.0 SSD you can get when all things are considered. Throughput and latency are both top notch. This comes with higher power consumption than equivalent DRAM-less drives, but it has DRAM. That’s a fair trade-off, but what isn’t is the drive’s sustained performance. Despite firmware updates, the T500 is still inconsistent when it comes to longer sustained-write workloads. Its large cache should hide this well for all but the most hardcore users, but it remains a sticking point that keeps this drive from being flawless. If you need DRAM and sustained performance, you’re better off getting a 4TB drive with an eight-channel controller: the WD Black SN850X, the Samsung 990 Pro, or an E18-based drive like the Sabrent Rocket 4 Plus-G.

Crucial sells the T500 both with and without a heatsink, which is a great option to have for certain builds. This is something else that makes it a great all-around drive, but ultimately it falters with pricing at 4TB. At the time of review it’s still being sold at $50+ over decent DRAM-less drives like the Lexar NM790 or Teamgroup MP44 which makes it a hard sell unless you really want DRAM. That begs the question: is DRAM worthwhile?

This is an awkward question to answer because areas where DRAM is especially useful — with a fuller drive, running heavier workloads, that sort of thing — the T500 can be less consistent due to its cache handling. For everyday use, on the other hand, you probably wouldn’t be able to blind guess which drive has DRAM. It’s more of a nice-to-have.

One thing that challenges the T500’s relevance is the fact that its primary competitors, such as drives that use the DRAM-less Phison E27T — which is a possible sidegrade for the MP44 and others, and the default for the Corsair MP600 Elite and others — are excellent choices for laptops even over something like the NM790. These drives run cool with high performance and none of the sustained write shenanigans. The problem is, none of these are available at 4TB. What is available are 4TB drives with QLC instead of TLC flash, if that’s a compromise you’re willing to make. And it's a reasonable option if you just need extra storage in a secondary drive at the lowest possible price.

If you have your heart set on that much storage with TLC flash, though, the T500 is a top runner for now, even with its high relative cost. But it does have some real direct competition from existing drives, like the Samsung 990 Pro 4TB and WD Black SN850X 4TB, both of which also cost $309 at the time of writing. Depending on the workload, the T500 can come out ahead of or behind those drives (though we never tested the SN850X 4TB, we did just recently check out the SN850X 8TB).

Overall, the T500 is a good achievement. It’s basically a modern replacement for the SK hynix Gold P31 — or as a higher capacity option than the SK hynix Platinum P41. It’s faster and more efficient than the Gold P31 and is now available at up to 4TB. If you’re looking to upgrade your laptop from an older or OEM drive, or even a Gold P31, the T500 looks pretty good. With pricing the way it is, though, it’s probably not the best PS5 option. For desktop use, it’s somewhat of a push depending on what you’re trying to do. If money is no object and you just want an excellent all-around drive, it’s arguably the best drive on the market, and we can highly recommend it. 

MORE: Best SSDs

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Sibelco, one of the two companies mining ultra-high-purity quartz at Spruce Pine, North Carolina, has restarted production a little over ten days after Hurricane Helene devastated the area. According to its press release, Sibelco only suffered minor damage, and all its employees are safe and accounted for. With the company restarting shipments to its customers and ramping production to full total capacity, the chipmaking industry is assured that it won’t have issues with the pure quartz supply needed to make the silicon base of semiconductors.

Hurricane Helene was a Category 4 hurricane that severely affected the Spruce Pine area, thus raising some fears that it would disrupt the accessible and affordable supply of quartz needed to create silicon ingots. These ingots are sliced into thin wafers and polished, then etched to form the chips we find on our computers. However, you can’t just melt ultra-pure silicon in any container to create the silicon ingots you need to make chips. Impurities in the container could react with the molten silicon, so you need an equally pure quartz crucible to hold it.

Most chip makers and their suppliers have enough silicon wafers or silicon ingots to weather a disruption in the supply chain, so many did not expect any significant industry repercussions from the tragedy. Besides, there are other quartz sources globally, although they’re likely not as readily available and affordable as what the North Carolina mines supply.

Nevertheless, Sibelco’s production restart is welcome news for the entire industry. After all, the supply chain horror stories that started during the 2020 COVID pandemic and extended until 2022 are fresh in our memories, and we don’t want a repeat of that.

It reveals how vulnerable our global semiconductor supply chain is to major regional events. Although the U.S., the E.U., and several other nations are working to establish a more robust chipmaking industry, investments being made today, like the U.S. CHIPS Act, will take years, if not decades, to bear fruit. So, in the meantime, our only hope is that nothing more untoward will happen that will disrupt the delicate global network needed to make the chips that drive our society.

Crucial has finally launched the 4TB variant of its successful T500 PCIe Gen4 SSD. Available with or without a heatsink, the T500 is a brilliant choice for desktops, laptops, or PS5, and is already Tom's Hardware's top pick for a laptop SSD in its 1TB/2TB variants.

First released in 2023, the Crucial T500 is among the best PCIe Gen 4 drives. While some among us are firmly in the PCIe Gen 5 era, many people are still staying in the Gen 4 camp, whether it's for the DRAM cache, the reduced cooling costs (Gen 5 SSDs still run much hotter than Gen 4 on average), or simply because of the PS5 and many laptops still top out at Gen 4 speeds. PCIe Gen 4 drives are still a very active market, and the T500 beats the whole pack in our testing. 

The Crucial T500 4TB runs at an advertised 7000 MB/s read and 6900 MB/s write. This is a slight downgrade in speeds from the 2TB model's 7400 / 7000 MB/s, but this is still very close and fast enough to saturate a Gen3 connection. The 4TB model may, as a result, fall behind the Samsung 990 Pro, our current best 4TB SSD, which advertises 7,450 / 6,900 MB/s at 4TB. The T500 2TB beats the 990 Pro 2TB in all of our tests, so Samsung may win an important victory in the Gen 4 4TB space. 

Crucial's T500 SSD gets its special sauce from a customized Phison controller. Aiming for efficiency, it crams four channels and a DRAM cache under its hood. As a single-sided drive, the drive is tailor-made for laptops or PS5s. The T500 fits in the Crucial stable just below the Crucial T700—a pricier, faster Gen5 drive—and its predecessor, the P5 Plus. 

The Crucial T500 4TB is available now from Crucial's website and other online retailers. The drive carries an MSRP of $399 without a heatsink and $449 with heatsink, though Crucial's website currently has the drive on sale for $309/$319, barely undercutting the 990 Pro's current price tag. Watch this space to see if the T500 4TB has what it takes to earn another slot in our best SSDs rankings.

A computer engineer and tech enthusiast has converted his Crucial BX500 512 GB SSD from quad-level cell (QLC) to single-level cell (SLC) NAND, reducing capacity to 120GB in the process. TechPowerUp database maintainer Gabriel Ferraz shared details of his feat, trading capacity for performance and durability, in the video below. Before the end of the video, he also charts the impacts on durability and performance.

The Crucial BX500 512GB SSD is a SATA connector drive that uses QLC NAND to provide a keenly priced half gig of storage for the consumer market. Many buyers will be happy with this product as it is, especially if they are replacing a slow HDD. However, enthusiasts don’t usually warmly embrace QLC storage as it has a reputation for lower performance and questionable durability. Sometimes drives feature DRAM or SLC caches (or use part of your PC RAM) to address some of these concerns.

Crucial’s BX500 512GB SSD contains the key components of a Silicon Motion SM2259XT2 controller and NAND flash dies from Micron (the parent company).

Ferraz noted that four NAND dies were onboard his drive and they had NY240 part numbers. This data is important for the QLC to SLC conversion process. For this delicate process, an app called MPtools for the Silicon Motion SM2259XT2 controller was downloaded from the USBdev.ru site, and the correct die reference numbers were carefully inputted. In the video, we get a walkthrough of the process which results in the system seeing a new 120GB SLC drive.

With the process complete Ferraz talks us through the benefits. Probably the primary benefit in the case of the BX500 is the boost in endurance. According to his tests and calculations, post-modification the SSD’s endurance jumps to 4,000 TBW. This is a sizable increase in TBW, in the order of about 3,000%.

In performance testing, using standard tools like ATTO and Crystal Disk Mark, many of the potential benefits of the modification didn’t show themselves due to the drive’s SATA III interface. Thus aspects of performance like data transfer speeds weren’t as impressively scaled as they could have been, but we see more promising uplifts in latencies and access times. However, Ferraz seemed most impressed by the efficiency gains he observed.

Please be very careful if you feel the impulse to try this SSD mod as Ferraz says you could “kill the drive” if something goes wrong during the process.

Silicon Motion has been in touch with Ferraz, and apparently, there may be some collaboration to deliver similar SLC mode conversion abilities to M.2 NVMe drives.

August 25, 2023 Update: We've added new testing for the 2TB PNY XLR8 CS3140 SSD on page 2.

Original Review published June 22, 2022:

PNY has been around for a long time, known in particular for its flash-based products, but also for its memory and GPUs. The company's new XLR8 CS3140 SSD tops out at 7,500/6,850 MBps of sequential read/write throughput for the 2TB and larger models, and it's also flexible in that it has both a version with a heatsink and one without, and you can also use a separate PlayStation 5 heatsink for the latter. 

Last year, PNY reduced the endurance rating on its CS3030 SSD by almost 80% due to the impact of Chia cryptomining, which was known for killing SSDs with heavy write workloads. But PNY has now returned to high endurance ratings with the XLR8 CS3140.

We tested the new 1TB XLR8 CS3140 model, which is a good starting capacity, but the 2TB model offers better peak performance because it is packed with more flash dies. Naturally, the 4TB is even more compelling if you're after more raw capacity. However, a basic primary drive is quite fine at 1TB, especially when coupled with strong hardware and a good SLC caching scheme like we see with the previous-gen CS3140. Let's see how the PNY CS3140 fares in our test suite. 

Specifications

The PNY CS3140 is available in the 1TB, 2TB, and 4TB capacities. These all have a five-year warranty backed up by 700TBW, 1400TBW, and 3000TBW of write endurance, respectively. Performance tops out at 7,500/6,850 MBps of sequential read/write throughput for the 2TB and larger models.

PNY doesn't officially list the IOPS measurements, but we can guess that this drive would be similar to other SSDs that have the same hardware combination. We had to dig up the endurance numbers on PNY's Taiwanese site, and the company seems to operate on the principle of “the less the information, the better.” That does make some sense in an era where SSD manufacturers often change the SSD's hardware after launch, but you'll need to do your homework to find concrete specs for these drives.  

The drive supports 256-bit AES encryption. PNY sells this drive with and without a heatsink, and we are testing the version without. This makes it a good choice for the PlayStation 5, possibly as-is, but PNY sells a separate heatsink for SSDs that is designed especially for the PS5. 

Software and Accessories

The CS3140 doesn't come with anything special. PNY does offer a download for its PCIe SSD toolbox software for firmware updating on their site. This software also gives some basic information about the drive, such as SMART readings, and allows for secure erasing.

A Closer Look

The drive is quite plain, sporting a rear label with basic details about the drive. Under the top label, we see the controller, DRAM, and four NAND packages in a standard configuration.

The flash is labeled IA7BG94AYA, which is 176-layer Micron TLC, a.k.a. B47R. This flash has proven excellent when coupled with the Phison E18 controller. As there are four packages, each should have four 64GB flash dies in a QDP configuration.

Based on data pulled from the drive, this flash is standard B47R running at 1200 MT/s. We’ve seen faster, but this is sufficient with two dies per channel. The endurance and quality of the flash look good, so there are no surprises here.

The Phison E18 controller needs no introduction - it’s been used on a wide variety of drives to good effect. It tends to be optimized for peak performance, specifically in sequential work, and it’s particularly good when paired with newer flash.

The DRAM is SK hynix’s H5AN8G6NCJR-VKC, which is 1GB of DDR4 in the 512M x 16b configuration. This would be C-die DRAM with the normal DRAM-to-NAND ratio, more than enough for any user. We can also spot the Phison power management package (PMIC) near the controller.

MORE: Best SSDs

MORE: Best SSD for the Steam Deck

MORE: Best Hard Drives

MORE: Best External SSDs

Comparison Products

There are some new drives out on the market that weren’t here when we reviewed the 1TB CS3140, including PCIe 5.0 SSDs like the Crucial T700 and Corsair MP700. We also have reviewed some great 4.0 SSDs like the Samsung 990 Pro, the Solidigm P44 Pro/SK hynix Platinum P41, and the WD Black SN850X. There’s also the Sabrent Rocket 4 Plus-G and Adata Legend 960 Max. Lastly, we have the budget-oriented Solidigm P41 Plus. Keep in mind that we got the PS5 heatsink with the 2TB CS3140 which makes it great for the PS5 even if the drive is dated in standard benchmarks.

A Closer Look

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The CS3140 is below average in 3DMark. It scored slightly worse than the 1TB version.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The CS3140 is near the bottom in PCMark 10, too, faring worse than the 1TB version.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

Things aren’t quite as bad with DiskBench. The CS3140 is close enough to the other high-end PCIe 4.0 SSDs. The 5.0 SSDs - the T700 and MP700 - pull away due to their faster interface, while the slower P41 Plus is at the bottom.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The CS3140’s sequential performance in ATTO and CDM is adequate but not on the level of the better PCIe 4.0 SSDs. This is fine as sequential performance at this level isn’t super important. Random write latency is good — surprisingly good even, but random read latency is fairly bad and significantly worse than the 1TB version of this drive.

The reason for that is obvious; PNY switched to 112-Layer Kioxia TLC (BiCS5) which does not perform as well with this controller as the original 176-Layer Micron TLC (B47R). However, if you’re buying a drive for capacity and especially for the PS5, this is not a big deal.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data. Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash.

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 2TB CS3140 writes in pSLC mode for over 30 seconds at up to 6.2 GB/s. This makes for a cache over 200GB, which is fairly large. Luckily, the drive does not easily hit a folding state, and instead drops to around 1.6 GB/s in TLC mode. This is not a bad result in any way. The 1TB version with Micron TLC is faster in TLC mode and it would probably be even faster with more dies at 2TB. Still, the BiCS5 2TB CS3140 is fairly consistent. Its recovery response has it lingering in TLC mode, which is luckily not too bad.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

For temperature recording we currently poll the drive’s primary composite sensor during testing with a 24C ambient.

The 2TB CS3140 with BiCS5 flash is not very efficient, especially compared to similar drives using Micron’s TLC instead. This is to be expected as this flash is less efficient in general and does not seem to pair quite as well with Phison’s E18 controller. This result isn’t crtical if this drive is used in a high-end desktop or PS5, but it does demonstrate the downsides of swapping the flash. It’s worth pointing out that other manufacturers can and have changed hardware over the course of an SSD’s lifespan.

We only reached a brief maximum of 67C during testing, which is far away from significant throttling. This drive would be quite happy in a PS5 with the optional heatsink. We would recommend it for that or perhaps some extra fast storage if the drive is priced right. Right now, at the time of the review update, the 2TB SN850X is probably the better option.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

MORE: Best SSDs

MORE: Best SSD for the Steam Deck

MORE: Best Hard Drives

MORE: Best External SSDs

PNY XLR8 CS3140 1TB SSD

RATING: ★★★★ (4 Stars)

PROS

+ Consistent performance
+ Strong peak and all-around performance
+ Power-efficient
+ Good PlayStation 5 option from PNY

CONS

- Pricing

OUR VERDICT
The 1TB PNY CS3140 is a solid drive that can be used in a PC or in the PlayStation 5, but the latter is best with PNY’s separate PS5 heatsink. Performance is strong, but there’s not much special about this drive versus the competition.

Comparison Products (1TB)

We compared the PNY CS3140 to other PCIe 4.0 drives, including the more budget-oriented WD Black SN770. Contenders using proprietary controllers include the WD Black SN850, the Samsung 980 Pro, and the Crucial P5 Plus. The other three drives use the same Phison E18 controller, but the Gigabyte Aorus Gen4 7000s and Corsair MP600 Pro use the older 96-layer Micron TLC while the last one, the Kingston Fury Renegade, is equipped with the same 176-layer flash as the CS3140.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

3DMark shows off WD’s excellent optimization, but thanks to their upgraded flash, we also had respectable results for the CS3140 and Fury Renegade.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

Again we see strong optimization from WD and Crucial, with the newer flash helping out the CS3140 and Fury Renegade. PCMark is often considered a “real world” benchmark, and we can look particularly at the latency results to see that the older 96-layer flash on the 7000s and MP600 Pro is appreciably slower.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

This bandwidth-heavy test favors the high-end PCIe 4.0 drives, particularly those based on the Phison E18 controller. The newer 176-layer flash also brings improvements, particularly as we are testing at 1TB rather than 2TB of capacity. The Phison E18 can peak with 32 dies of flash, and considering these use 64GB dies, that means 2TB. The 176-layer flash is also faster per die than the 96-layer flash for copying tasks.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The queue depth 1 (QD1) sequential read and write results from ATTO paint a simple picture: drives using the Phison E18 come out on top, with a significant edge given to those with the newer 176-layer flash. To some extent, there is a “wall” due to the limits of the x4 PCIe 4.0 interface, as seen with queue depth 8 sequential reads in CrystalDiskMark. The write result shows a bit more delineation between the tested drives. At this point, we see a theme, with the CS3140 underperforming the Fury Renegade, possibly due to inferior firmware optimization.

Random QD1 performance is good to excellent on the CS3140, where we again see the newer flash pulling ahead. Again, no real surprises, which is not a bad thing — this hardware configuration performs very well on multiple fronts.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The CS3140 writes at above 5.8 GBps for almost 20 seconds, revealing an SLC cache of around 116GB. This is not a particularly large cache but should be more than sufficient for a 1TB SSD. The drive then hits its direct-to-TLC phase at up to 2 GBps, which is quite fast for a 1TB drive. This reflects the choice of having a smaller SLC cache while using the faster 176-layer TLC. Performance would be higher at 2TB due to additional interleaving. As the cache is dynamic, it will shrink as the drive is filled. It’s still possible to hit a slower state with sufficient writing, but this is difficult to achieve.

Beyond flash choice, the size of the cache is one thing that separates drives based on the E18 controller. As we can see here with the Fury Renegade’s results, it has a larger, faster cache than the CS3140, but no middle-ground TLC state — this allows some flexibility for the manufacturer and the user. While cache recovery remained slow on both drives, the CS3140's ability to maintain TLC speeds ensures it has better steady state performance for edge cases.

Which is better is arguable, particularly as high-end PCIe 4.0 drives tend to focus on raw bandwidth. However, we believe in the future that sustained performance may become more important. For example, if you’re able to stream assets steadily from a drive via DirectStorage. That would be reads and not writes, of course, but there are nevertheless cases where consistency trumps peak performance alone.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

We also monitor the drive’s temperature via the S.M.A.R.T. data and an IR thermometer to see when (or if) thermal throttling kicks in and how it impacts performance. Remember that results will vary based on the workload and ambient air temperature

Efficiency remains strong on the CS3140, thanks to the flash and the ability to quickly finish workloads with high peak performance. It still trails the Fury Renegade by a small bit.

According to both SMART and the IR gun, the CS3140 idles in the mid-40Cs and peaks in the upper-70Cs. This is hot but not unexpected as the drive lacks a heatsink. The smaller SLC cache helps when the drive hits the TLC flash, but sustained writes still get it pretty close to the throttling mark. In fact, we did see light throttling after about 215GB of writes when the drive hit 78C. Therefore, we recommend a heatsink if you anticipate having heavier workloads and/or a hotter environment.

Test Bench and Testing Notes

We use a Rocket Lake platform with most background applications such as indexing, windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Conclusion

We didn't expect any surprises, and we didn't find any — the PNY XLR8 CS3140 gets close to our performance expectations without any weirdness. As with similar drives, the peak and general performance ranges from good to excellent. The SLC caching scheme is conservative, making for consistent performance in sustained workloads, and the new flash ensures good power efficiency. The PNY CS3140 is also a strong choice for a PlayStation 5 drive due to its option for a tailored heatsink.

On the other hand, there’s not much to set this drive apart from the competition. It is nice to see a functioning SSD toolbox — this is something lacking with, say, Inland drives — but the pricing isn’t particularly competitive. That being said, this is an adequate alternative and would be a good choice if you can find it on sale. PNY isn't pulling any endurance shenanigans here as far as we could tell, even if the company is a bit shy about some of the drive's specifications.

We scored the Kingston Fury Renegade a bit higher. The Fury Renegade edges out the CS3140 in almost every test while coming with cloning software, although currently, it is more expensive. The Kingston SSD is also designed for the PlayStation 5 and is perhaps a more trusted brand, perhaps except for those users who remember the ancient NAND switch-a-roo.

We also previously reviewed the Silicon Power XS70, a similar drive that's also designed for the PlayStation 5. It costs less than the CS3140 and comes with a PS5-capable heatsink out of the box. All of this makes it hard to give the PNY XLR8 CS3140 a higher score, but it is still a strong drive and contender. The biggest challenge for the XLR8 CS3140 is that it's in an increasingly crowded market, but it is a competent alternative.

MORE: Best SSDs

MORE: Best SSD for the Steam Deck

MORE: Best Hard Drives

MORE: Best External SSDs

Even with proper bed leveling and one of the best 3D printers, sometimes you can experience issues with your 3D prints not sticking well to the 3D printer bed either because of the nature of the design you are printing or the type of material you are using. One of the ways to address this issue is by using a brim. 

A brim is an additional thin material you add to your design in the slicer to improve adhesion between the printed object and the 3D printer bed. It achieves this by increasing the contact area between the object and the build plate, providing better stability and preventing warping or detachment during printing. A brim is not intended to be a permanent part of the final 3D printed object. Once you finish 3D printing, you can remove it.

Benefits of 3D Printing Brim

Some of the benefits of a brim in your designs include:

• Improved adhesion: As we have mentioned above, the primary purpose of a brim is to enhance adhesion between the print and the build platform, reducing the likelihood of issues resulting from poor adhesion. Also, you will not need adhesion aids like glue sticks or hairspray to help the print attach to the bed.
• Stability: For objects with complex geometries and overhangs, adding the brim will act as an anchor, preventing those parts from detaching or becoming misaligned.
• Minimized warping: A brim helps counteract the effects of differential cooling, which can cause warping in materials like ABS. The extended surface area improves bonding, reducing the risk of warping.
• Adaptability to print speeds: A brim is particularly valuable when printing at higher speeds, as it helps counteract the increased forces that can make the print more prone to lifting or shifting.
• Cleaning up the design is easier: Post-processing your 3D prints, especially on the bottom section, is easier as you can peel away or trim the brim easily without affecting the design itself. Also, you can easily clean any remaining traces on the object without affecting the part itself, unlike when it was 3D printed without it.

Not all 3D prints require a brim, but there are specific situations where its use is essential. You need it when:

1. 3D Printing With Materials That Are Prone to Warping

Some materials like ABS are prone to warping due to its thermal properties, specifically, the rate at which it cools and contracts. The temperature differences cause the corners of the print to lift, leading to a failed or distorted object, and sometimes the design can completely detach from the bed. If you are using such a material, you should consider using a 3D printing brim, as it provides an extra adhesion around the base of the print.

2. When 3D Printing Objects With Overhangs

If your 3D print has overhangs or other delicate parts in the initial layers, you should use a brim, as those parts tend to detach from the build plate. A brim will extend the contact area of these sections and make them stable.

3. 3D Printing Parts With Small Base Areas

Parts with small base areas might have issues adhering properly to the build plate as the 3D printer operates and vibrates. So adding a brim can make a big difference as it will increase the contact surface between the print and the platform and distribute adhesion forces more evenly.

So even if forces are applied to the design from the nozzle, especially when it reaches the top areas, it will be difficult for it to detach. You can also adjust the thickness of the brim so that it can be thicker and hold your design firmly.

4. 3D Printing at High Speeds

When printing at higher speeds, the increased forces acting on the design can make it detach from the bed, particularly during the initial layers. Using a 3D printing brim in such scenarios is a valuable strategy as it will provide an additional surface area for adhesion, distributing the forces more evenly and reducing the chances of lifting.

5. Multiple Small Objects

When 3D printing multiple small objects, there's a higher risk of them detaching because of the forces that act on them during the print. A brim provides additional adhesion points for each object, improving stability.

Also, a brim can help ensure uniformity in the first layer of each small object, which is important for achieving consistent quality across all objects in the print. This is particularly beneficial when printing objects that require precise dimensions or alignment.

How to Add a Brim to Your 3D Prints

You can use any 3D printer slicer to add the brim to your 3D prints. If you are using Cura slicer, you can follow the steps below.

1. Import the 3D model you want to print.

2. Navigate to build plate adhesion from the print settings section and choose Brim.

3. Adjust brim settings to customize it further to fit your needs. The following settings are available:

• Brim minimum width: This determines the minimum length of the filament that will be used when 3D printing the brim. The default value is always 250mm, and it is always enough. 
• Brim width: Determine how wide the brim should be in millimeters. A wider brim adheres more to the printer bed.
• Brim line count: You choose how many lines of the brim will be printed around the base of the model. This setting also affects the width of the brim.
• Brim distance: It determines the horizontal distance between the first layer of the print outline and the first brim line. You should always retain it at 0mm.
• Brim only on outside: You can enable this option so that the brim can be printed on the outside of the model to make it easier to remove it after 3D printing, as it reduces the amount to be removed.
• Brim inside avoid margin: This setting determines the distance of the brim that will be removed from the outer brim that touches the internal of the other part. You should retain the default value, 2.5mm, and see if it works for your design. If not, you can adjust it.
• Smart brim: When you enable it, it improves brim removal by swapping the print order of the innermost and second innermost brim lines.

4. Review and slice: Review all your settings, including the brim settings, and ensure everything is okay before exporting your G-code file after slicing.

How to Achieve a Good 3D Printing Brim

Sometimes your brim might have issues like not correctly sticking to the bed, or even it can be challenging to remove from the print because it is either too thick or too thin. You can consider doing the following to solve those issues.

1. Use Appropriate Brim Settings

You can adjust the settings highlighted above until you get the most appropriate one for your 3D prints. For example, you can choose a brim width that provides sufficient adhesion while not being excessive.

If there are adhesion issues, you can consider increasing the brim line count to provide more support by printing more lines around the object. However, you should not add too many as it can make the removal process challenging. Additionally, ensure that the smart brim settings are selected, too, as it makes it easier to remove the brim.

2. Properly Level Your 3D Printer Bed

Bed leveling is often the first culprit most of us consider when facing any 3D printer issue, as it plays a crucial role in the quality and success of the first layer. If it isn't level, some areas might be too close to the nozzle, while other areas might be too far away, leading to poor adhesion or even detachment of the brim. So before you start 3D printing, you should properly level your 3D printer bed.

3. 3D Printer Material You Use

Different materials may require adjustments to brim settings. Materials like ABS are prone to warping and need a wider brim or increased line count for it to be printed well. A wider brim involves adding more lines to the edges of the part to help create a stronger, more stable part. For other materials like PLA, you can lower the brim settings.

4. Use the Correct Temperature

If you are using too high or too low temperatures, it can affect your brim. Too high bed temperatures can cause the brim to adhere too firmly to the build surface, making the removal process more challenging and increasing the risk of damaging the main print when trying to detach it.

Also, high temperatures can cause the filament to be heated much, and become fluid, which can result in the deformation of the brim, causing it to lose its intended shape and compromise the accuracy of the first layers.

If the temperature is too low, it can cause insufficient bed temperatures resulting in weak adhesion between the brim and the build surface. So it’s important to choose the optimal 3D printing temperature for your specific filament to achieve the perfect print.

MORE: Best 3D Printers

MORE: Best Budget 3D Printers

MORE: Best Resin 3D Printers

Your CPU is the brains of the operation when it comes to your computer, but unlike with your own brain, we can get an exact measure of how much processing capacity is in use at any given time.

Whether you’re a heavy multitasker, or trying to get the most performance out of a video game, monitoring CPU usage is an important way to know if you’re getting all the performance your processor is capable of, or whether it’s behind any performance issues you may be experiencing. 

In this how-to, we’ll be showing you various ways to monitor your CPU usage, from quick and simple, to slightly more complicated, but much more detailed. If you’re overclocking, or worried that your CPU might be running too hot, you’ll also want to learn how to check your CPU temperature.

Some Important Notes About Interpreting CPU Usage

Generally CPU usage is reported as a single percentage, showing how much of your CPU’s resources are currently being used. So if that number is at 100%, your entire CPU is busy. If it’s at 25%, then only a quarter of the CPU is occupied. However, there’s a small wrinkle to this, because modern CPUs consist of multiple cores.

Suppose you have a quad-core CPU (it’s probably time to upgrade!) and one of the cores is completely maxed-out, while the rest are idle. You’ll see only 25% CPU usage, but your performance is limited to the maximum performance of that core. We see this often in video games that don’t make good use of multiple CPU cores, so the frame rate of the entire game is limited to the fastest performance of a single CPU core or thread.

In this case, a single CPU percentage for usage isn’t all that useful, which is why some of the methods we’re about to explore will let you see every individual CPU core doing its thing, to help you see where any bottlenecks may be.

How to Check CPU Usage With Windows Task Manager

Windows has robust and detailed performance monitoring tools built into it, and with every version of the operating system the information on offer is more detailed. Here’s how to access these tools in Windows 11 (this also works in Windows 10):

1. Launch Task Manager. The easiest way is by hitting Ctrl + Shift + Escape. 

2. Open the Performance tab and click on “CPU.” Here you can see a graph showing what percentage of your total CPU power is in use. If that’s all you wanted to know, then you can stop here.

3. (Optionally) right-click on the graph and select Change Graph to > Logical Processors. This will show you the utilization of each individual “thread” your CPU offers. This number may not reflect the number of physical CPU cores, since some types of CPU cores can handle two threads at once. This is what “logical” processors refers to rather than physical processors.

With the logical processor view open, you can tell whether your CPU’s load is evenly spread across all logical processors, or if one or two threads are being hammered at 100% while the rest of the CPU is under lower stress.

How to Check CPU Usage With the Xbox Game Bar

Using the Windows Task Manager to check CPU usage works great if you’re using a windowed application, but unless you have a multi-monitor setup, you’re out of luck if you’re using a full-screen application like a video game. 

Luckily, Windows includes a feature known as the Xbox Game Bar, which lets you pin a performance monitoring widget to your screen on top of a full-screen app. Here’s how to use it:

1. Launch Xbox Game Bar by Pressing Win + G.

2. (Optionally) Move the performance widget. You can use your mouse to drag the Performance widget to a spot on the screen where it won’t obscure something important in your full-screen app.

3. Pin the Performance Widget with the Pin button. By clicking on the pin button you’ll lock the performance widget in place even after closing the Game Bar.

4. Close the Game Bar with Win + G. Once you’ve pinned the widget, you can close the Game Bar and the performance info will remain overlaid on your full-screen app.

Do get rid of the overlay, just open the Game Bar again, and unpin the widget, the close the bar, and continue with business as usual.

How to Use MSI Afterburner to Monitor CPU Usage

While the Xbox Game Bar gives you a basic CPU usage meter you can use in games and other full-screen applications, it doesn’t show you details about the per-thread CPU usage. If you want to see what your entire CPU is doing while running a full-screen 3D application, the MSI Afterburner combined with Riva Tuner Statistics Server (RTSS) is the way to go. Here’s how to set it up:

1. Download and install MSI Afterburner. Head to the official website to download, unzip, and install the app. During installation, make sure not to uncheck the box for installing RTSS.

2. Open MSI Afterburner. You’ll be asked to provide special privileges for the app, you must say “Yes” to continue.

3. Click on the setting “Cog” button. This opens the MSI Afterburner properties window where you can modify its settings.

4. Navigate to the “Monitoring” tab. This is where you can select which aspects of your system will be monitored, and which will be displayed in the on-screen overlay.

5. Tick every entry titled “CPU Usage” under “Active hardware monitoring graphs”. Each one of these entries represent a logical CPU in your computer. In this case there are 20, but this will differ based on your hardware.

6. Hold Shift and click on every CPU usage entry to highlight it. Select all of the numbered CPU usage entries.

7. Click on “Show in on-screen display.” This ensures that these monitored CPUs appear in the on-screen display.

8. Navigate to the On-screen Display tab and set a shortcut. By default the “Toggle on-screen display” shortcut is set to use the “Home” key, but that’s not always convenient, so set whichever key you would prefer. We chose Shift + P in this case.

9. Click OK and run your game. With your choices locked in, run your game or 3D app and all of the CPU usage measurements should appear. If you don’t see anything, use the toggle you set. If that doesn’t work, make sure MSI Afterburner is running in the background.

How to Use HWInfo64 to Check Your CPU Usage History

All of the methods we’ve covered so far let you see what your CPU is doing live as it happens, but there some instances where you may want to see a record of CPU usage, so that you can do some detective work about usage patterns and see if there are particular times when your CPU usage spikes or otherwise behaves oddly.

The free utility HWInfo 64 offers exact this feature, and here’s how you can use it:

1. Download and install HWInfo64. You can find it on its official website. If you don’t want to install it to your system, you can opt to download the portable version instead.

2. Run the app and click “start”. It may take a while for the app to finish initializing.

3. Click on Monitoring > Sensor Status. This will open the sensor status window where you can see all the different data sources being monitored.

4. Note the Total CPU Usage sensor. You can hover the mouse pointer over this or any sensor to get an explanation of what it measures and how it’s calculated.

5. Click the Start Logging button.

6. Choose a location to save your CSV file. HWInfo64 saves your log as a CSV file. Pick a location where you’ll easily find it.

7. When you’re done logging, click the “stop logging” button.

8. Open your CSV file. Open the CSV file with the viewer of your choice. Here in Excel, we used New Workbook > Data > Import from Text/CSV > Load.

The first two columns log the date and time, you can compare this to the column for Total CPU Usage, or any of the columns that show individual core or thread usage.

Now that you know these different methods to check your CPU usage, you’ll never have to wonder about what your CPU is actually doing, no matter what sort of task you’re using it for.

The Seagate Lightsaber Collection Special Edition FireCuda is the SSD to get for Star Wars fans. A sleek heatsink and switchable heroic lightsaber faceplates are combined with addressable RGB LEDs for an unmistakable look. It’s a potential showpiece for your build, and you can control the lighting through software to match your specific aesthetic. The drive itself is nothing special, but it’s fast enough for any sort of use and has Seagate’s comforting three-year data recovery service to boot, making it a contender for our list of Best SSDs.

Underneath the Seagate Lightsaber FireCuda, as we will henceforth call it, is essentially Seagate’s older FireCuda 530. This drive is known for good performance in sustained workload and is a high-end PCIe 4.0 SSD with a high endurance (TBW) rating. Seagate’s warranty comes with one attempt at data recovery, and they also support the drive with its SeaTools software. That means this drive is a safe bet, even if it’s not the fastest. The challenge would be in acquiring one!

The Lightsaber FireCuda is meant to be shown off, which means it’s intended for desktop use. It arrives with three changeable faceplates and a long ARGB connector cable. The latter can be a pain to work with, depending on your predilections for cable running. The heatsink is functional and designed by EKWB, which matches well with the higher sustained performance. Otherwise, this is a standard FireCuda 530, which is among the more consistent SSDs based on Phison’s E18 SSD controller.

Specifications

The Seagate Lightsaber FireCuda is available at 1TB and 2TB, priced currently around $159.99 and $239.99, respectively. It can be difficult to find these drives, and the price is fairly high for what you get, but if you want the specific aesthetic, it’s not too bad. The drive is fastest at 2TB, hitting 7,300 MBps / 6,900 MBps for sequential reads and writes and 1M / 1M IOPS for random reads and writes.

Seagate warranties this drive for five years and 1,275TB of writes per TB of capacity, which is pretty high. It is not recommended to use the TBW rating when selecting a drive, as even the standard 600TB per TB is an immense amount of writes. Seagate also backs the SSD with its three-year Rescue Data Recovery Services which is nice, although probably more useful for hard drives.

Software and Accessories

The Lightsaber FireCuda comes with the standard manuals and packaged content, including a notice of its Rescue Data Recovery Services. It’s also possible to download SeaTools on Seagate’s site for Windows, Linux, and in a bootable format, which is nice to see. The SeaTools application offers standard monitoring and tools for Seagate’s HDDs and SSDs.

A Closer Look

The Seagate Lightsaber SSD arrives in a comfortable box with three faceplates to use, including those of Obi-Wan Kenobi, Luke Skywalker, and Darth Vader, all officially licensed. Also included is a 40cm, or almost 16”, LED RGB connector cable. This cable lets you control the addressable RGB (ARGB) lightning with a wide array of RGB software applications.

The drive is clad in an effective heatsink crafted by EKWB, the same company that helped with the Seagate FireCuda 530 that forms the basis of this drive, although the design is different.

The LED bar is thick underneath, but otherwise no surprises. This drive still uses the Phison E18 SSD controller, 176-Layer Micron TLC (B47R), and DDR4 memory. This drive is double-sided with two DRAM and eight NAND packages. It has the standard 2GB of DRAM, and at 2TB, there is an excellent amount of flash for interleaving, so that’s where the drive reaches its peak performance.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

Some things have changed since our Seagate FireCuda 530 review, so the Star Wars variant of the drive faces some new challengers, including the PCIe 5.0 Crucial T700. It’s also up against the original FireCuda 530, the Samsung 990 Pro, the Solidigm P44 Pro, the WD Black SN850X, the Sabrent Rocket 4 Plus-G, the Crucial P5 Plus, the Adata Legend 960 Max, amd the Corsair MP600 Core XT. All of these drives are high-end PCIe 4.0 except for the last, which is a mid-range, QLC-based part.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The Lightsaber FireCuda performs the same as the original FireCuda 530, as expected. Results are mediocre, including against the DirectStorage-optimized Rocket 4 Plus-G.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The Lightsaber FireCuda shows its age, losing out to fresher PCIe 4.0 SSDs, aside from the Legend 960 Max and MP600 Core XT.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

The Lightsaber FireCuda copies fairly fast, within striking distance of the P44 Pro. All drives fall behind the might of the PCIe 5.0 Crucial T700.

Synthetic Testing - ATTO / CrystalDiskMark 

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The Lightsaber FireCuda shines brighter in ATTO, which is unsurprising as SSDs powered by Phison controllers perform well in this benchmark. It’s also clear to see that the PCIe 5.0 Crucial T700 pulls away from the pack. Sequential performance in CDM is fairly good on the Lightsaber FireCuda, aside from QD1 sequential reads, where the FireCuda 530 is even worse. The T700 dominates here, too.

Turning to random performance, the Lightsaber FireCuda manages to hang with the crowd but can’t quite reach the 990 Pro and T700 when it comes to 4K random reads at QD1. This is a key metric used to estimate real-world performance and feel, and the 990 Pro’s faster flash is still unbeatable.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. 

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The original FireCuda 530 was one of the most consistent SSDs we’ve tested, thanks to its conservative pSLC cache. The Lightsaber FireCuda is virtually identical. It writes in pSLC mode at up to 6.9 GBps for over 30 seconds with a cache size of around 225GB. The write speed then declines to about 3.8 GBps in TLC mode, fast enough that it eventually reduces to folding at less than half that speed.

This is a good result with the Lightsaber FireCuda getting close to the Rocket 4 Plus-G at steady state, the latter drive among the fastest we’ve tested aside from the Legend 960 Max and T700. This makes the drive a good option for certain heavier workloads. pSLC recovery time to TLC performance levels is also good, but recovery to maximum speed takes longer. Avoiding additional writes and letting sequential writes hit TLC can be beneficial for reducing unnecessary wear, which can help support the high TBW rating.

Power Consumption

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

The Lightsaber FireCuda is surprisingly efficient, although far from the best. Using the LEDs should not be a concern, and this drive is adequately cooled. It’s precluded from being used in a laptop due to the heatsink, and the same is true of the PlayStation 5, although you wouldn’t be able to see the lightsaber in those cases, anyway.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Bottom Line

The Seagate FireCuda Lightsaber SSD is merely an average performer within the high-end PCIe 4.0 array of SSDs, yet it easily stands out by its appearance. It arrives in thoughtful and tasteful packaging with an unambiguous Star Wars aesthetic, and who doesn’t love lightsabers? However, that’s mostly what this is - an SSD for Star Wars fans to show off in their desktop builds. The pricing reflects that and it’s essentially a collector’s item, which is perfectly fine. You can go with a cheaper drive and even design your own solution if you can’t find one of these.

Of perhaps greater contention is the focus on RGB LEDs. Historically, LEDs have had a rough road with SSDs. The Kingston HyperX Fury RGB SSD was prone to overheating due to using so many LEDs. Other drives, like the Patriot Viper VPR100, saw potential performance drops depending on the RGB mode. We had no issues with the Adata XPG Spectrix S40G, which like the Lightsaber FireCuda, has addressable RGB. However, RGB, in general, is a meme for flashy builds, and it’s not uncommon to have issues getting RGB software to behave.

It’s not really fair to compare this drive to other standard SSDs, as it’s designed for a specific purpose unrelated to its performance. However, it is generally fast enough to be “fast,” and the heatsink plus high sustained performance make it good for a wide range of applications. It’s possible to get an equivalent-performing SSDs and embellish it yourself, if you desire to get the right aesthetic. The Lightsaber FireCuda’s design won’t match all builds, even with the RGB LEDs. On the other hand, the inclusion of three faceplates means that you don’t have to settle for one specific Star Wars hero’s lightsaber.

On the whole, this is a cool drive, even if it comes from Seagate, a storage company that hasn’t really gotten deep into the SSD game. It would be interesting to see how well its data recovery service works with SSDs. It is nice to know that your expensive lightsaber SSD is backed by something a little extra. Aside from that, the biggest downside is dealing with the cable. If that’s not an issue for you and you can find this drive, it could be a worthwhile grab.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Crucial T700 sets a new bar in performance for consumer SSDs, pushing speed higher than any other PCIe 5.0 NVMe drive on the market. It reigns as the king of the hill for sequential bandwidth and random IOPS, at least for now. Its native DirectStorage firmware optimization is an added benefit over almost all PCIe 4.0 SSDs for future gaming, as well. Crucial departs from the reference heatsink design with its own effective solution while also offering a cheaper, bare-drive variant for easier installation with motherboard and custom M.2 heatsinks. Together, this makes it a nice option for enthusiasts and early adopters.

The T700 is part of a wave of PCIe 5.0 SSDs about to deluge the consumer SSD market, all so far based on the same Phison E26 controller and Micron flash. This puts pressure on the manufacturers to differentiate their products in other ways, particularly with cooling. This could bring some innovation to the market, but current industry trends have created lingering concerns about the future of NAND that could eventually reduce the buyer’s pricing advantage from such competition.

Competing controllers are also on the way with InnoGrit’s IG5666 and SMI’s SM2508, and flash of the 232-Layer generation should be in wider production from Micron and other manufacturers. Faster drives based on the Phison E26 SSD Controller have also been announced offering up to 14 GBps, so early drives like the T700 have a limited window to shine. We’ll know more after Computex, which is currently underway, but many of these drives won’t be available until later in the year. At the time of review, this drive is the fastest of its crop.

Specifications

The Crucial T700 is available at 1TB, 2TB, and 4TB capacities in both bare and heatsinked variants with MSRPs of $179.99/$209.99, $339.99/$369.99, and $599.99/$629.99, respectively. This is a significant premium over slower PCIe 5.0 SSDs like the heatsinked Inland TD510, where the 2TB model can be had for $249.99 at the time of review. It’s likely that the T700’s price will come down some time after launch, but these prices are what Crucial currently lists on its store.

Let's also be clear that paying $30 extra for the small metal heatsink is asking far too much. If you have a motherboard with a decent M.2 heatsink, you should just get that model. The included heatsink does work fine in our testing, but a $10 premium would be more in line with what you're getting.

The T700 is capable of reaching 12,400 MBps / 11,800 MBps for sequential reads and writes and 1,500 million IOPS for both random reads and writes. The primary benefit over slower PCIe 5.0 SSD variants is the sequential performance increase, but write IOPS are also higher. Crucial warranties this drive for the typical 600TB written per TB of capacity at up to five years. The retail drive has TCG Opal 2.01 support for encryption, which was not enabled in our previous Crucial T700 Preview.

Software and Accessories

Crucial offers its own SSD toolbox called the Crucial Storage Executive, which has the typical features you would expect. This includes drive information, SMART attributes, firmware updates, secure erase/sanitize, overprovisioning, and other features related to encryption. Third-party software like CloneZilla is recommended for cloning and imaging.

A Closer Look

The Crucial T700 can be purchased with or without its custom, passive heatsink. The heatsink is carefully designed to avoid the need for a fan, using a variety of materials and an airflow-optimized shape. The backplate makes it usable in motherboards that have flush thermal padding, which can be removed for this drive.

The non-heatsinked variant of the drive comes with a label that should not be removed, and the drive should not be operated without sufficient cooling applied. Such cooling would include a motherboard’s M.2 heatsink or your own aftermarket solution. The drive should not be used bare in the PlayStation 5 or in a laptop.

The drive itself has an SSD controller, one DRAM package, and two NAND packages on the top side. The rear side has two more NAND packages for a total of four. There's also a power management integrated circuit, or PMIC, labeled Phison.

The T700 utilizes Phison’s E26 SSD controller, an 8-channel design capable of running a bus speed of up to 2400 MT/s to help saturate four lanes of PCIe 5.0 connectivity. Initial E26 drives top out around 10 GBps at 1600 MT/s, but Crucial has opted for 2000 MT/s here. Future drives, some of which have been already announced, will push 2400 MT/s for 14 GBps or more. Higher speeds put more stress on the hardware, so cooling becomes more critical, especially with sustained workloads.

The 2TB T700 comes with 32Gb, or 4GB, of LPDDR4. This is more than enough memory even for the 4TB SKU. LPDDR4 is more power-efficient than DDR4 but the impact is not enough to help keep the E26 solution from being very power-hungry.

The flash is Micron’s 232-Layer TLC, or B58R, capable of running at 1600-2400 MT/s. This hexa-plane flash comes in 1Tb dies, so each 512GB package has four dies for a total of sixteen. Given the E26’s 8-channel design, it’s therefore possible that peak performance will be realized at the 4TB capacity.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The 2TB Crucial T700 SSD is tested both with Crucial’s heatsink (labeled "HS" in the charts) and with out Asus motherboard’s heatsink (labeled "(Mobo)" in the charts). It’s up against other current PCIe 5.0 SSDs including the Corsair MP700, the Gigabyte Aorus Gen5 10000, the Inland TD510, and we also have results from the engineering sample (ES) Phison E26. The high-end PCIe 4.0 drives in our charts include the Adata Legend 960 Max, Crucial P5 Plus, Sabrent Rocket 4 Plus-G, Samsung 990 Pro, and Solidigm P44 Pro.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The T700 is the fastest drive we’ve tested to date in 3DMark. The performance improvement over other PCIe 5.0 SSDs is a modest 4~7 percent, with a larger gap between the T700 and leading 4.0 options.

Speaking of gaming, drives using the E26 SSD controller come with Phison’s I/O+ firmware for DirectStorage optimization, which is only available on one PCIe 4.0 SSD, the Sabrent Rocket 4 Plus-G. Faster drives should be out by the time that makes a real impact, however.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

PCMark 10 shows a similar pattern with the T700 again on top. The real-world differences between the top drives are very small, especially among the PCIe 5.0 models, but the slightly higher transfer speeds do push Crucial to the top.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

Copy speeds in DiskBench are faster, with significant gains over previously-tested drives. The main reason to buy PCIe 5.0 drives right now is for this type of performance if you are capable of making use of it — meaning, you have a Z690/Z790 or X670 class motherboard.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The T700 rules the day in ATTO, particularly as the block size increases past 64KB. There’s a dip at 2MB which, as mentioned in our Crucial T700 Preview, could be a controller quirk or a facet of the flash organization. It's present on the other Phison E26 drives as well for read performance. The T700 also dominates in the CDM sequential results, as expected.

One area the T700 falls short is with random 4KB performance at queue depth 1, still an important result for everyday workloads. In particular, random read latency is critical, and here the T700 comes in second to the Samsung 990 Pro. This is due to Samsung’s 176-Layer TLC having superior read latency. The T700 does do slightly better than previous PCIe 5.0 SSDs, possibly as a small boost from the higher MT/s rate.

Micron’s 232-Layer flash was designed more with density in mind and this may be the case with competing flash to come, which dents the appeal of current 5.0 drives if you’re seeking performance over capacity.

One additional note is that our drives under test are not operating as the primary, or OS, drive. The preview T700 sample had reduced random write performance when used as the primary drive, but this has been fixed for release.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data. Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash.

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 2TB T700 writes at up to 12 GBps for over 20 seconds. This matches what Crucial states, which is that about 11% of user capacity is allocated for the dynamic cache. When the drive is empty this is over 220GB from 660GB of SLC, or approximately one-third of the drive’s raw NAND capacity. This is a decently-sized cache to help balance burst and sustained performance even as the drive fills.

The precise pattern of writing will depend on your cooling and the drive’s placement, as demonstrated by the difference between our T700 as native and with a motherboard heatsink. We're testing drives in a PCIe expansion card with the Asus motherboard, and this includes a very large metal cover that serves as a heatsink — far larger than a typical M.2 cooler.

After the cache is exhausted, the T700 writes at around 3.6 GBps in TLC mode. It can maintain this for quite a while before hitting a folding state at around half this speed, and when the cache recovers the drive goes back to TLC. Recovery to TLC performance should be relatively rapid and is adequately fast. This flash should be able to sustain higher levels of TLC performance at 4TB with additional interleaving, but this remains to be seen.

Thermals and Temperatures

We've also begun to collect data on drive temperatures during our DiskBench and Iometer write saturation testing. The latter, as you can imagine, is far more demanding and represents a worst-case scenario. Using the version of the T700 that includes a heatsink during the write saturation testing, it reached a peak temperature of 87C on the controller. At that point, it begins throttling, which is why you see periodic dips in the above write saturation tests around the 850 second mark.

What ends up happening is a short period (less than a second) of reduced performance, then the drive returns to full (saturated) speeds for around 10 seconds. Rinse and repeat until the workload is done.

Please note that this is not a real-world scenario that would normally be encountered by a consumer drive. During our 15 minute stress test, Iometer wrote 2.95 TB of data to the T700. In other words, it more than filled the drive. During the DiskBench testing, where we copy 100GB of files to the drive three times in short succession, drive temperatures peaked at 51C.

Certainly, a larger/heatsink and/or direct active cooling can dramatically drop temperatures. With the large Asus heatsink on the expansion card, peak temperatures during our Iometer testing only reached 54C. This time, due to the lower temperatures and lack of throttling, the T700 drive managed to write just over 3TB of data in nine and a half minutes. The DiskBench tests in this case didn't even break 40C.

Power Consumption 

We use a Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs, including all of the PCIe 5.0 SSDs we've tested, can consume a few watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects we'll look at, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

The T700, like the other drives based on the Phison E26 SSD controller, pulls a lot of power. It’s not nearly as efficient as even the fastest PCIe 4.0 drives. This is something that will be ameliorated in time with more efficient controllers and flash. Considering that these 5.0 drives are designed for high-end desktop use only, the issue of power draw is less concerning than it would normally be.

The T700 preview sample had not been optimized for the lowest power states, PS3 and PS4, that is the idle states. This is not a huge factor as the drive should not be making its way into laptops and is hardly a drive you’d choose for a low-power build. Still, the drive at release can use these power states at a typical 85 milliwatts.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Bottom Line

The Crucial T700 is the fastest SSD we’ve tested to date. Some minor wrinkles have been ironed out since our preview, and the drive as a whole produced good results. Drives that are just as fast or even faster have been announced, but this one is available here and now. Crucial has gone to great lengths to offer a decent, passive heatsink, while leaving the option open for you to use your own M.2 cooling solution. This makes it attractive in comparison to existing, slower PCIe 5.0 SSDs that use the reference cooler design with a fan, like the Inland TD510.

The Gigabyte Aorus 10000 also has a heatsink option, albeit with less clearance, but that drive is also slower. If you’re shooting for a drive that comes with a heatsink, however, it is currently $30 less at 2TB than Crucial’s price. If you're looking for a sans-heatsink alternative, the 2TB Corsair MP700 is $40 less than the bare T700, although it may be difficult to find in stock. There should be more options after Computex, especially as we get into August with both pricing and availability expected to improve.

The current outlook divides the PCIe 5.0 SSDs into three rough categories: 10 GBps, ~12 GBps, and ~14 GBps (in the future), all for maximum sequential reads but with boosts to sequential writes possible as well. Directly competing in the middle level with the T700 will be the Team Cardea Z540 and the KLEVV CRAS C950, among others. Faster known drives include the MSI Spatium M570 Pro and ADATA’s Project NeonStorm, as well as the IG5666-based Team Group Z54A. More will certainly be revealed during Computex. Pricing and cooling will differ between these options, but ultimately the lower tiers are early adopter products by design.

As for the PCIe 5.0 selling point a drive like this may have, the performance results speak for themselves: For anything but sequential bandwidth, you should go for a much cheaper PCIe 4.0 SSD. Even on a PCIe 5.0 capable board, getting a drive that performs similarly aside from sequential speeds while being much more efficient makes sense — and many 4.0 drives don't require a heatsink. Furthermore, these early 5.0 drives are as slow as 10 GBps with the T700’s 12 GBps currently the fastest available, neither of which make full use of 5.0 bandwidth.

These SSDs need more time to mature. They can be fun for early adopters and enthusiasts that want to set some records. There's still the possibility of higher capacities, not yet realized, which is also nice. If you are already feeding into that mindset with your build then the premium isn’t that crazy, but if you are not hell-bent on cutting-edge storage you should stick with drives like the Solidigm P44 Pro or WD Black SN850X.

This could influence overall system building decisions, and DirectStorage does not currently have sufficient presence to be a factor if you’ll be upgrading again in a year or two, or even three.

That said, the Crucial T700 is nice for what it is, and it’s important that manufacturers keep pushing the limits. Better products at better prices are a net benefit and encourage market penetration of newer storage hardware, which will hopefully hasten the arrival of DirectStorage-enabled games and applications. The T700 is very fast and we applaud it for that today, but we can’t wait to see better things in the future. 

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Today at Amazon, the Crucial P3 2TB SSD is marked down to one of its best prices yet. This SSD usually goes for around $100 but right now is marked down to $88. It’s a PCIe Gen 3 SSD that uses 3D NAND. It comes in a range of capacities starting at 500 GB and getting as big as 4TB, however, this offer is only for the 2TB edition.

All of the drives in this line have an M.2 2280 form factor. We reviewed the Crucial P3 SSD when it debuted last fall and overall appreciated it. It’s supported with a good warranty and already had a low price point before today’s discount. This drive would be well-suited as a secondary drive for many users.

The Crucial P3 2 TB SSD uses an NVMe PCIe Gen 3 x4 interface to connect with. It can reach moderately high read/write speeds with Crucial indicating its cap to be around 3500/3000 MB/s. This speed is consistent also for the 1TB and 4TB models.

Users get a limited 5-year manufacturer’s warranty with the purchase that voids should the drive surpass 440 TBW. You can also take advantage of Amazon’s 30-day return policy.

Visit the Crucial P3 2TB 3D NAND M.2 SSD product page at Amazon for more information and purchase options. It’s not clear for how long the discount will be made available.

Early samples of the Crucial T700 have the poise to be the fastest consumer SSD when the retail units hit the retail market on May 30. The T700 or one of its competitors will undeniably disrupt the list of the best SSDs very soon.

Corsair's competing MP700 recently got pushed into the spotlight due to several scenarios where reviewers tested the SSD without a cooler, contrary to Corsair’s recommendation. The publications didn’t conduct the tests in bad faith, but rather to see what would happen if a PCIe 5.0 was to operate without a heatsink or cooler. And the experiments exposed a flaw in the MP700’s firmware related to the thermal throttle curve. Phison’s firmware engineering team has quickly jumped on the issue and has prepared a firmware update fix that's currently going through internal validation.

The T700 and MP700 share similar components. Both SSDs use the Phison PS5026-E26 PCIe 5.0 controller and Micron 232-layer 3D TLC NAND. In Crucial’s case, the T700 utilizes slightly faster NAND (clocked at 2,000 MT/s) than the (1,600 MT/s) variant in other E26-powered drives. The higher-binned NAND propels the T700 to sequential read and write speeds up to 12.4 GB/s and 11.8 GB/s. However, even though the T700 and MP700 use identical E26 controllers and similar Micron NAND, they don’t use the same firmware.

Larger and more prominent vendors have their own firmware teams working on SSD firmware. But smaller brands, or those not as heavily involved in the SSD realm, rely on reference firmware from the manufacturer. The rules vary by the SSD controller manufacturer. Some, like Silicon Motion, allow complete control over the firmware, while others don’t.

In Phison's case, the company does firmware customization for its customers; in this case, that would be Corsair and Crucial. As a result, the T700 and MP700 don't have the same firmware. The vendors do not make their firmware, and given that these two drives exhibit different failure mechanisms, it is safe to assume that the firmware is different. Regardless, clearly there was a flaw in the MP700 firmware. And the solution seems to be as easy as tweaking the thermal throttle curve. That's why Phison was able to cook up a fix so quickly.

Crucial's PCIe 5.0 drive hasn't suffered as drastic an issue, but that doesn't mean things are great with that drive if you run it without a cooler, either. German publication ComputerBase pointed out that the T700 suffered from thermal throttling in CrystalDiskMark when a heatsink wasn't cooling the drive. The SSD reportedly hit temperatures of 86 degrees Celsius. The SSD throttled hard, as evidenced by the low write speeds that stuck around the 100 MB/s mark. The publication's feedback aligns with our experience. Without a heatsink, our T700 sample performed fine in tests like DiskBench, but exhibited thermal throttling in longer benchmarks, such as CrystalDiskMark and IOmeter. The recorded average write performance after 15 minutes on the bare drive was 1,027 MB/s, down from the 3,681 MB/s on the one with the heatsink. Thermal throttling was working as expected. And honestly, when a drive is capable of  write speeds of nearly 12 GB/s, in most mainstream use cases, you aren't going to be writing to it for anything close to 15 minutes straight.

Crucial sells the T700 in two variants: as a bare drive and with a heatsink. The pricing difference is $30. However, the bare drive doesn't mean you should run the PCIe 5.0 SSD without a heatsink or cooler. In fact, Crucial, like Corsair, advises customers to use a heatsink with its PCIe 5.0 SSD.

The bare drive version targets consumers who plan to use your motherboard's integrated M.2 heatsink with the T700. It's similar to how Intel stopped bundling stock coolers with the chipmaker's K-series processors, since enthusiasts frequently ditched the stock cooler for an aftermarket cooler. Depending on the motherboard, CPU cooler, and graphics card choice, sometimes it's impossible to fit a beefy M.2 cooler into the mix. So while Crucial's T700 bare drive offers consumers an option to save $30, you should be sure your motherboard has a cooler that you can bolt onto the drive, especially if you're going to be hitting it with lots of sustained writes.

The Gigabyte Aorus 10000 was one of the first PCie 5.0 NVMe SSDs to be announced, and today this speedy SSD has finally landed on our test bench. It stands apart with an included optional heatsink, which has aesthetic appeal. With performance that peaks at 10 GB/s of sequential throughput, this drive has plenty of horsepower for demanding workloads. 

Gigabyte traditionally markets its SSDs to OEMs, and the Aorus 10000 is difficult to find at the time of review. Since the drive comes bare with the optional heatsink included, this could be a good choice if you want to use your own cooling solution or run multiple drives. However, it is not recommended to operate the drive bare, so laptops are generally out of the target applications. Gigabyte also offers additional software support for this drive which helps gives it a leg up over products like the Inland TD510. The real question is pricing: The drive has been listed at $399.99 for the 2TB model, but we aren’t sure if that is official pricing —Gigabyte has yet to share details.

Specifications

The Gigabyte Aorus 10000 is intended for sale at 1TB and 2TB. Performance can reach up to 10,000 / 9,500 MBps for sequential reads and writes with no official IOPS rating listed. The warranty is the standard 5-year with 700TB of writes per TB of drive capacity. The 10000 stands apart from other early PCIe 5.0 SSDs with its inclusion of a custom heatsink.

Software and Accessories

Besides including an optional heatsink, you can use your motherboard’s instead, Gigabyte also has a downloadable SSD toolbox. This application has basic functionality such as drive information, S.M.A.R.T., TRIM optimization, and secure erase. The drive is also supported in the Gigabyte Control Center. You can use free software like Clonezilla for cloning or imaging purposes.

A Closer Look

The included heatsink is quite robust and matches the aesthetic of the drive, although the height might cause issues in some builds. The overall design is effective and not too ridiculous. Gigabyte includes the necessary screw. The bare drive can be used with your own heatsink or, more likely, the M.2 heatsink available on your motherboard. It should not be used in a laptop or PlayStation 5 in its bare state.

The Aorus 10000 has the standard E26 layout with the SSD controller, a single DRAM package, and two NAND packages per side for a total of four. This is a sufficient number of packages to mount 8TB of the expected flash. There is also space for a second DRAM package on the rear, although 4GB of volatile memory should be plenty even for 8TB of NAND.

This uses the standard 4GB of SK hynix LPDDR4 and four packages of 512GB Micron 232-Layer TLC. Each package has four 1Tb, or 128GB, dies.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The Gigabyte Aorus 10000 competes against two other PCIe 5.0 SSDs, the Phison E26 ES and the Inland TD510. The Crucial T700 is still currently only for preview due to unfinished firmware, but we’ll have a full review soon. We also have several high-end PCIe 4.0 drives for comparison, including the Samsung 990 Pro, the Solidigm P44 Pro, the Adata Legend 960, the Corsair MP600 Pro LPX, the Crucial P5 Plus, the Gigabyte Aorus 7000s, and the Sabrent Rocket 4 Plus-G.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The Aorus 10000 is very fast in 3DMark, beating previous drives, with only the Inland TD510 able to keep pace. The Phison ES is an electronic sample, so this drive isn't available at retail. 

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The Aorus 10000 also does well in PCMark 10, although not quite as well as the E26 ES. The latter did have some quirks that needed ironing out for launch, but consistency and reliability come before outright performance, so peak performance can vary with shipping products. 

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The Aorus 10000 is outstanding in both ATTO and CrystalDiskMark with sequential reads and writes. It isn’t quite as fast in random IOPS, but the 4KB performance, on the whole, is good to great. This flash will need more time to mature. As it is, it’s designed for density and bandwidth first.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. 

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The Aorus 10000 writes at up to 10.2GBps in pSLC mode, 3.5GBps in TLC mode, and nearly 1.4GBps in its folding state. The real star of the show is the peak sequential write speed, but sustained write performance is not poor. We will likely see better things here in the future. pSLC recovery is slow for the up-to 220GB cache, but the drive is able to get back to its relatively good TLC performance mode without too much of a problem. If you’re interested in consistent, high sustained performance, some 4.0 drives remain equally or more viable, and at lower cost.

Power Consumption

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

The Aorus 10000 is simply not efficient. This is not as relevant for a drive of this type as it’s meant for fast burst performance and is an early adopter product. We will see more efficient drives later this year and next for laptops and other non-desktop devices. The Aorus 1000’s massive heatsink positions it primarily for high-end desktop use where efficiency is less of a concern.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Bottom Line

The Gigabyte Aorus 10000 is part of the welcoming committee for PCIe 5.0 SSDs. It’s not particularly impressive when considering what the 5.0 interface can produce in terms of total bandwidth, but it’s a good start. This is an early adopter product, but it still performs quite well on the whole with very high sequential performance. Performance is good, and it’s ready for DirectStorage when it actually arrives.

Gigabyte separates itself from the pack with an included optional heatsink that does not rely on the reference design. This drive is usable with your own heatsink or your motherboard’s M.2 heatsink. This is nice, but the question mark here is pricing. The Inland TD510 already exists with aggressive pricing, and there are faster drives in the pipeline. It is also challenging to find the Aorus 10000 in stock. Still, it’s a perfectly suitable product for what it is, being an excellent choice for enthusiasts who want the latest and greatest, particularly if their system matches the included heatsink.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Crucial is preparing to launch its ultra-fast T700 SSD, which is now available to pre-order starting at $209.99 for the 1TB model. The drive is currently the world's fastest Gen 5 SSD, surpassing all of the SSDs on our list of best SSDs with read speeds of nearly 12 GB/s. The drive launches later this month, on May 30.

The T700 is a PCIe Gen 5.0 x4 NVMe M.2 SSD, rated for up to 12,400 MB/s in sequential read speeds, and up to 11,800 MB/s in sequential write speeds, as well as random read speeds and write speeds of up to 1,500K. It's a double-sided 2280 M.2 drive featuring Phison's bleeding-edge E26 controller, 4GB of LPDDR4 memory, and Microsoft DirectStorage API.

Performance will vary depending on the drive capacity you choose: the 1TB model will feature slightly lower read and write speeds (11,700 MB/s sequential read speeds, 9,500MB/s sequential write speeds, 1,350K random read speeds, 1,400K random write speeds, as well as an endurance rating of 600TB). The 2TB and 4 TB both get the faster drive speeds mentioned earlier, with the only difference being the drive endurance — which is rated at 1,200TB for the 2TB model and 2,400TB for the 4TB model.

We previewed the drive two weeks ago, and found it indeed lives up to its superlative as the world's fastest SSD. The prototype T700 we received outperformed every single SSD we've tested so far, including the Phison PS5026 engineering sample, Aorus Gen 5 10000, Samsung 990 Pro, and more in a plethora of tests including DiskBench, PCMark 10, and 3DMark. We expect the production version to perform just as well — if not better — than the prototype version we tested.

Unsurprisingly, Crucial's new SSD won't come cheap and cost roughly twice what PCIe Gen 4 drives go for these days. The drive will start at $179.99 for the 1TB model, while the 2TB model will cost $339.99, and the 4TB model will cost a whopping $599.99. Heatsink versions of the T700 will cost an additional $30 on top of the initial price.

They may be a little late, but the first crop of PCIe 5.0 SSDs have finally arrived at retail. The Inland TD510 is one of the first available and affordable models, and its leading-edge performance of 10GB/s of sequential read throughput and 1.5 million random read IOPS is a good indicator of what we can expect from the new wave of drives.

The Inland TD510 offers unprecedented bandwidth with a compatible system, and performance as a whole is quite good. The heatsink is sufficient but comes with a largely ineffective fan, and this basic reference design matches Inland’s no-frills approach to SSDs. There’s no extra software support, but you do get an extra year of warranty. The TD510 is a good choice for early adopters, but faster PCIe 5.0 drives are coming later this year. Let's see how the Inland TD510 fares in our test suite. 

Specifications

The Inland TD510 is currently only available at the 2TB capacity, but this is a perfectly reasonable amount of storage. A 4TB option would be nice, but 2TB is quite cost-effective, selling for $279.99 regularly and at the time of review. This carries a premium over high-end PCIe 4.0 SSDs, but it’s competitive for the new crop of 5.0 drives.

The drive can eke out up to 10,000 / 9,500 MBps for sequential read/write throughput and 1,500K / 1,250K for random read/write IOPS. This is on the lower end for 5.0 SSDs but still offers a bump over PCIe 4.0 SSDs. The TD510 is warrantied for 1400TB of written data, which is pretty standard, and it gets an extra year of warranty as Inland covers it for six years versus the standard five.

Software and Accessories 

The Inland TD510 is a no-frills offering, so it comes with no software. The included reference heatsink is listed as removable, so you can use a motherboard’s M.2 heatsink instead. For software, we suggest the free CrystalDiskInfo for S.M.A.R.T. monitoring and Clonezilla for cloning.

A Closer Look

Removing the heatsink wasn’t quite as easy as the TD510’s packaging would have you believe. We surely don't recommend you attempt the process if you don't have to. The small fan is optional, and we recommend not plugging it in as it does not provide much in the way of improved cooling performance but does generate noise. This is because the reference heatsink doesn't have an air channel that allows the air to exit the small chamber underneath the fan.  

Under the heatsink, we find an SSD controller, a DRAM package, and two NAND packages. There are two more NAND packages on the underside, for a total of four.

The Phison E26 controller will become a familiar sight on PCIe 5.0 SSDs. Initial drives reach up to 10 GB/s, but future models can go higher, as with the Crucial T700. This controller is often paired with 4GB of LPDDR4. This is twice the typical amount of DRAM, and this type of memory is more efficient than DDR4.

NV066 is the FBGA code for 4Tb of Micron’s 232-Layer TLC (B58R). Each package is in a quad die configuration (QDP) for an overall total of 2TB. It should be possible to achieve up to 8TB with this layout.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The Inland TD510 faces two other PCIe 5.0 SSDs, the Phison E26 ES and the Gigabyte Aorus 10000. However, the Crucial T700 is not a final product yet, so we didn't include those results here. There are also a slew of high-end PCIe 4.0 drives for comparison, including the Samsung 990 Pro, the Solidigm P44 Pro, the Adata Legend 960, the Corsair MP600 Pro LPX, the Crucial P5 Plus, the Inland Gaming Performance Plus, and the Sabrent Rocket 4 Plus-G.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

No surprises here, as the TD510 is one of the fastest drives we've ever tested in 3DMark. The Phison E26 listed in the charts is an early engineering sample, not a product you can buy on the shelves. However, this gives us a taste of the theoretical heights of performance with PCIe 5.0 SSDs powered by the Phison E26 SSD controller. Drives with similar levels of performance will arrive later this year.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The TD510 also does quite well in PCMark 10, beating all of the PCIe 4.0 SSDs and effectively tying the Gigabyte Aorus Gen5 10000. 

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

Nothing copies files like a PCIe 5.0 SSD. The TD510 comes out on top thanks to the additional bandwidth afforded by the PCIe 5.0 interface. 

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The TD510 is a solid performer in ATTO, although it drops at the 2MB block size for sequential reads. This matches the E26 ES. This may be a quirk of the E26’s optimization or a result of how the flash interleaves, given that 16 dies at six planes each with 16KiB physical pages would suggest a super-page up to 1.5MiB in size. The TD510 also dominates all PCIe 4.0 drives during the Crystal Disk Mark sequential read and write tests.

Random performance is unremarkable, with the 990 Pro besting the TD510 in read latency and several drives beating the TD510 in write latency. Again, this is a result of flash architecture, as the need for higher density requires compromise in other areas.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. 

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

There are no surprises here as the TD510 matches the E26 ES quite closely. The ~220GB pSLC cache can write at over 10.2GBps before the drive moderates to 3.5GBps in TLC. There’s also a slower folding state at around 1.4GBps on average. It’s possible we could see higher results with a 4TB drive. Recovery to TLC is relatively quick, but pSLC recovery takes more time.

Power Consumption

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

We had hoped that our power efficiency results for the E26 ES would be improved with further controller optimization. This is not the case, as the TD510 is quite inefficient. That being said, this is not a drive you buy for efficiency, especially as a large heatsink is all but required. There’s a bright future for more efficient drives, but this one isn’t quite there yet. Luckily, this drive is not intended for use in a laptop.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Bottom Line

The Inland TD510 surprised people by being available early. It surprised people again by dropping in price relatively rapidly. Anyone with access to a Micro Center can and has been able to enjoy PCIe 5.0 storage bandwidth for some time now by purchasing the TD510. It’s still an early product, but it is faster than 4.0 SSDs and is really meant for early adopters due to its requirements and the fact that faster drives are on the way.

The TD510 comes with the reference heatsink, which has an essentially worthless fan, which holds back our final score a bit. After getting rid of mechanical storage, the last thing you want is another failure point with added noise. However, even if the fan failed, it is so ineffective that it wouldn't make much of a difference.

The Inland TD510's DirectStorage optimization is nice for future-proofing, meaning you'll be able to blast through loading scenes in optimized games released in the future. The drive isn’t very power efficient, but that isn't too discouraging — the drive is clearly aimed at performance addicts with desktop PCs. The TD510 is fast for what it is; just be aware that the early PCIe 5.0 drives aren't as fast as newer models that will come later in the year.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Crucial T700 sample we have in the labs today is unquestionably the fastest consumer SSD in the world, at least for now, delivering up to a blistering 12.4 GB/s of sequential throughput and 1.5 million random IOPS over the PCIe 5.0 interface. That's 70% faster than today's highest-end PCIe 4.0 SSDs, and 20% faster than the current crop of PCIe 5.0 drives.

The T700 is slated to be the first of a new set of faster PCIe 5.0 SSDs that offer even better performance than what we saw with the initial PCIe 5.0 SSD we covered in our Phison E26 SSD controller preview. The T700 uses the Phison E26 SSD controller, a robust design used in several other 5.0 SSDs, paired with speedy 232-Layer TLC flash, thus creating the fastest SSD we’ve ever tested in our labs. Impressively, it delivers this level of performance with passive cooling thanks to its well-designed heatsink, but if you remove the heatsink the SSD will also work well in motherboards with proper M.2 heatsink coverage. 

The drive also comes with DirectStorage enabled, which will allow games that support the feature to deliver the utmost in performance. Crucial provided an early T700 sample to us for testing, and the drives will launch in a few months (by the end of the second quarter). Let's see what they'll look like when they arrive. 

Specifications

The Crucial T700 delivers up to 12.4 / 11.8 GBps of sequential read/write throughput and 1.5 / 1.5 million IOPS in random read/write workloads with the 2TB and larger models (the 1TB model is slightly slower). The drive will be available in 1TB, 2TB, and 4TB capacities, with pricing yet to be determined. Crucial backs the drive with a five-year warranty and endurance of 600TB of writes per 1 TB of capacity.

The drive supports AES-256 encryption and TCG Opal 2.01, including for crypto erase (TCG OPAL was not implemented at the time of testing). Crucial also has other optimizations in the works, but the drives we're testing are considered Engineering Samples (ES), so the firmware is not final. Crucial expects that performance in random write workloads will improve in the future. 

Software and Accessories

Crucial offers its own Crucial Storage Executive SSD toolbox with the typical features you would expect. The utility includes drive information, SMART attributes, firmware updates, secure erase/sanitize, overprovisioning, and other features related to encryption. The T700 is not explicitly supported per Crucial’s website, but this will likely change when the drive arrives at retail. Third-party software like CloneZilla is recommended for cloning and imaging.

A Closer Look

The T700 comes with or without a heatsink, depending on your needs. If your motherboard already has a substantial M.2 heatsink, or if you want to install your own solution, the bare drive is an option. In that case, the drive has a copper thermal label that is not intended to be removed. This SSD is not intended to be used without a heatsink, such as in a laptop or PlayStation 5.

The default model comes with a pyramidal, central T-pose heatsink designed to keep the drive cool without any (direct) active cooling. The heatsink is constructed of multiple materials and has air channels to maximize heat dissipation via airflow in the case. The SSD components — SSD controller, NAND flash, power management integrated circuit (PMIC), and voltage regulators — have good contact with the heatsink via thermal material. The back side of the heatsink is nickel-plated copper that provides support and heat spreading, while the top is aluminum with a significant amount of surface area for heat dissipation.

All T700 models are double-sided with two NAND packages per side, for a total of four. The top of the SSD has the requisite SSD controller package and a single DRAM package.

The Phison E26 is the first consumer PCIe 5.0 SSD controller to hit the market. We previously conducted preview testing of an engineering sample, and now a whole host of drives are arriving with this controller on board. This controller has DirectStorage-optimized firmware, which we also previewed before its official release. That firmware has already found its way into one PCIe 4.0 SSD, the Sabrent Rocket 4 Plus-G.

The T700 is not the first PCIe 5.0 SSD on the market, as the Inland TD510 has been out for a while. The T700 is also not the first to be announced by a big brand name, as we can see with the Corsair MP700. However, the T700 is the fastest implementation of the E26, as it can reach up to 12.4 GBps in sequential read workloads thanks to its speedier flash with an I/O speed of 2000 MT/s. This is in contrast to the 1600 MT/s flash used in the earlier E26 SSDs, limiting them to around 10 GBps. The E26 controller supports up to 2400 MT/s flash, which would put a cap of around 15 GBps. But for now, the T700 is the fastest around.

The T700 has DRAM identified as D7CJG, meaning this is LPDDR4 in a 1G x 32b configuration. LPDDR4 is more power efficient than DDR4 memory (we saw the benefit of LPDDR4X in our Lexar Professional NM800 Pro review). In this configuration, we have 4GB of DRAM for 2TB of NAND, the same as in our E26 preview. There may be reasons for such a high memory-to-flash ratio, but we'll have to review the amount used with other capacities.

The 232-Layer Micron TLC (B58R) flash takes up the mantle from Micron’s very successful 176-Layer TLC (B47R). Micron has gone from four planes to six and has made other improvements that make multi-planar operations faster for superior internal parallelization. The move to 1Tb (128GB) dies over 512Gb (64GB) is also an important consideration for capacity: bigger dies, more storage.

Denser dies also mean less relative parallelization for the same capacity because fewer dies are required. This is mitigated to some extent by the increase in internal parallelization, or planes. The T700 can still reach peak performance at 2TB like older PCIe 4.0 SSDs, but has the opportunity to be better at 4TB as well, depending on multi-planar optimizations. The performance of TLC flash during sustained write workloads could also improve without a corresponding improvement to read latency, although read performance as a whole should improve due to independent plane advancements.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

Besides the prototype Phison E26 ES, the Crucial T700 is also up against the Gigabyte Aorus Gen5 10000, the Inland TD510, the Sabrent Rocket 4 Plus-G, the Adata Legend 960, the Corsair MP600 Pro LPX, the Crucial P5 Plus, the Samsung 990 Pro, and the Solidigm P44 Pro. We've tested all of those, though we haven't yet reviewed the Aorus and Inland models.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The T700 is the fastest drive we’ve tested in 3DMark, just barely edging out the Phison PCIe 5.0 E26 sample that we tested before the official launch. The T700's performance leap over Crucial’s last flagship, the P5 Plus, is significant. The other two currently shipping Gen5 drives (Inland TD510 and Aorus Gen5 10000) come in slightly behind the E26 engineering sample, with nearly identical performance — no surprise, since those two feature identical components.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The T700 is also the fastest drive we’ve tested to date in PCMark 10. It provides a significant advantage over PCIe 4.0 SSDs, although it doesn't take a massive leap over the Phison engineering sample, which again outpaces the Aorus and Inland models. 

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom 50GB dataset. We copy 31,227 files of various types including pictures, PDFs, and videos to a new folder. A secondary test copies those same files from the drive to itself (so it's both reading and writing data simultaneously). Finally, we follow-up with a reading test of a newly-written 6.5GB zip file.

The T700 copies at very high speeds, even significantly faster than the Phison E26 sample. This time the Phison sample basically ties the Aorus and Inland drives.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The T700 does require its heatsink to perform properly, especially during tougher or prolonged tests. Nevertheless, the T700 smooths out the bumps we saw with the Phison E26 sample in ATTO and sets all-time highs. Likewise, the drive also provides improved sequential throughput in the CrystalDiskMark benchmark.

We expect fewer improvements in 4K random workloads as this drive mainly benefits from a faster bus, and this bears out: Random read and write latencies see little to no improvement over PCIe 4.0 drives.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash.

We use Iometer to hammer the SSD with sequential writes for (over) 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The T700 obviously needs a heatsink for heavy sustained write workloads. Sustained write performance starts high, hitting nearly 12 GBps in pSLC caching mode for over 20 seconds. This cache size is ample while still providing good TLC performance after depletion of around 3.6 GBps. A slower 1.4 GBps folding state follows, and on the whole, the post-pSLC pattern is reminiscent of what we saw with the Phison E26 sample.

In steady state, after 15 minutes of sustained writes, the T700 still averages nearly 3.8 GBps, the fastest of any SSD we've tested to date. The main improvement comes from the sequential burst speed, with the T700 ending up about 6% faster than other E26 drives we've tested.

Sustained performance may improve with a 4TB drive, and the peak will be higher with 2400 MT/s flash. The T700's recovery to TLC speeds is relatively rapid, but pSLC is unsurprisingly a bit slower to recover.

Power Consumption

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

We want to immediately mention that Crucial is still optimizing the T700's power consumption in idle states. This did not impact our testing due to how we currently test, and it shouldn't be a huge factor anyway as this drive is not intended for laptop use.

The Crucial T700 is more efficient than the E26 ES, indicating that the controller is more efficient when able to run at higher speeds. This makes sense in our testing as the tests finish faster. This does mean higher peak consumption, making a heatsink all the more important for sustained workloads. Idle consumption with our desktop setup also shows improvement, and this controller is being optimized over time.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use appropriate cooling for all SSDs.

Bottom Line

The Crucial T700 takes a good thing and makes it even better. It’s hard to find fault with the product, especially as the non-Phison-reference heatsink is attractive and effective, and performance is equal to or better than other PCIe 5.0 drives. Faster, cheaper, and more capacious drives are on the way, and that’s a good thing.

There are a lot of DirectStorage skeptics, but Crucial seems pretty excited about the technology. The company is also especially proud to produce the fastest 5.0 SSD currently announced, even if it’s not using an in-house controller. The problem is that an announced drive isn't the same as a publicly available drive, and more critically, we don't have pricing details for the T700 yet.

When you look at the plunging SSD prices, you can now find good PCIe 4.0 2TB drives starting at just $120, like the Crucial P5 Plus 2TB. The T700 is unquestionably faster, but current Gen5 2TB SSDs start at $280 (Micro Center's inhouse-brand Inland TD510), and that's already $120 lower than the launch price from just a couple of months back. Unless you absolutely need maximum throughput, PCIe 4.0 drives are a far better value and are likely to remain so.

This is far from the only PCIe 5.0 SSD that Crucial has planned, but it's a nice surprise. The P5 and P5 Plus were also innovative in their own ways, although with proprietary controllers they have not held up well against the competition, except in pricing. And pricing is one area that Crucial has done well by introducing products like the P3 and P3 Plus. Unfortunately, Crucial hasn't shared pricing information for the T700 yet. However, we do know they'll arrive in a few months (by the end of the second quarter of this year).

Enthusiasts have been waiting for Crucial to bring out a leading-edge product, and the T700 looks to be just that. The key innovation is the use of higher speed Micron NAND, however, which is something other manufacturers can easily acquire. Whether they can match the T700 in other respects remains to be seen.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Addlink S90 Lite is a mid-range PCIe 4.0 NVMe SSD that offers nothing new but provides a satisfactory experience at 2TB without breaking the bank. It’s efficient and cool-running enough for laptop and PlayStation 5 use, and also would be a good way to add extra storage to your desktop. It’s very similar to the Corsair MP600 GS but is easier to find at 2TB and its price is competitive with other options. In some cases you could go cheaper with QLC or potentially pay more for the better-performing, TLC-based WD Black SN770. The S90 Lite is somewhere in-between, but nowhere near the performance or value of the best SSDs on the market.

Although unexceptional, the S90 Lite does have some quirks. At 2TB it has an overabundance of NAND flash dies which negatively impacts its performance, giving the Black SN770 an edge. It also has a different pSLC cache response than the MP600 GS which makes it a lot less useful for sustained workloads, although on the bright side that prevents the drive from overheating. Some drives in this class, like the Silicon Power UD90 and Team Group MP44L, have swapped to QLC at high capacity, and that’s also possible for the S90 Lite. Such a configuration would be more comparable to the Solidigm P41 Plus or Crucial P3 Plus, so beware.

Specifications

The Addlink S90 Lite comes in the 512GB, 1TB, 2TB, and 4TB capacities, although the 4TB model seems difficult to find. At the time of review these are on sale for $38.88, $59.88, and $109.99, respectively. This is competitive with similar drives like the Silicon Power UD90, the Team Group MP44L, and the WD Black SN770, to name a few. Some drives in this class have moved to QLC at 2TB while our S90 Lite sample has TLC, which makes this drive most attractive at that capacity.

The S90 Lite is capable of up to 5,000/4,200 MBps for sequential reads and writes, and 780K/800K IOPS for random reads and writes. There is no official TBW but Addlink indicates the warranty is based on “percentage used” by SMART. The warranty is good for five years.

Software and Accessories

Addlink provides an SSD toolbox which displays various information including drive specifications, SMART status, and firmware revision. The toolbox is capable of performing a secure erase.

A Closer Look

The 2TB Addlink S90 Lite uses a simple, single-sided design. Under the basic label there is a controller flanked on each side by two NAND packages. Placing the controller in the middle can provide advantages with heat spreading.

The Phison E21T controller has shown up on a lot of great budget drives. It’s proven to be a winner in that regard, offering excellent performance for a DRAM-less design. Its primary competition is WD’s proprietary controller in the Black SN770 and, as an alternative with QLC, the Solidigm P41 Plus's SM2269XT. It’s fine with either TLC or QLC and many models come with both depending on capacity.

The IA8HG94AYA NAND packages are Micron’s 176-Layer TLC, or B47R. The use of four packages for 2TB of flash implies each package is 512GB. As this flash uses 64GB dies, each package has eight dies in an 8DP configuration. Single-sided drives are ideal as they can fit in a wide range of devices and may be easier to rig for cooling.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The 2TB Addlink S90 Lite is facing the WD Black SN770, the Solidigm P41 Plus, the SK hynix Gold P31, the Crucial P3 Plus, the Kingston NV2, the Inland Performance, and the Samsung 990 Pro. The 990 Pro is one of the fastest drives we’ve tested even though it has had its issues. The S90 Lite is more likely to compete with the SN770 and NV2, but the P41 Plus and P3 Plus are popular budget QLC drives within its price range. The Gold P31 remains a popular PCIe 3.0 option. The Performance, for its part, represents older PCIe 4.0 SSD technology with DRAM.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The S90 Lite is average to below average in 3DMark. This result is lower than expected, but it’s still better than the NV2.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The S90 Lite is again middling and barely edges the NV2 in PCMark 10. Even the P3 Plus, which has QLC with the same controller, does better. Performance on the whole is still satisfactory, however.

The initial reflex is to suspect poor or restrained flash, similar to the situation with the Team Group A440 Pro and Corsair MP600 Pro LPX. However, analysis of other results, particularly random latency as explored below, suggests instead that it’s simply a matter of overhead from an excessive amount of dies to meet 2TB. We saw similar results with the 2TB Corsair MP600 GS. This is one limitation of 4-channel controllers when paired with 512Gb dies, but competing flash has the option of 1Tb dies, which makes the Black SN770 a better option for 2TB right now.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

The S90 Lite does better in DiskBench, which fits our expectations for this benchmark. It can’t match the Black SN770 which has been an amazing value SSD in recent months.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

For sequential performance we first look at ATTO where the S90 Lite performs as expected. Phison controllers excel at sequential reads especially. The drive disappoints with writes but performance is acceptable. The sequential results from CDM show the S90 Lite hanging in there. The Black SN770 and P3 Plus, two prominent rivals in this price range, are within the same performance ballpark. The P3 Plus is the undisputed choice for 4TB. For 2TB, the S90 Lite is positioned right between these two drives, which feels appropriate. Random 4KB performance is, thankfully, good to great; this is a surprisingly responsive SSD.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. 

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 2TB S90 Lite writes at almost 4.65 GBps in pSLC mode. It does this for approximately 31 seconds within its 145GB cache. It then hits a slow folding mode at just 200 MBps. This is a poor result and much slower than what we saw with the 2TB Corsair MP600 GS, but the 2TB S90 Lite also has a significantly larger cache. This is a trade-off that makes the drive far less useful for heavier or sustained workloads and it may suffer more when fuller. It compensates for this with rapid recovery of its cache.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

Temperatures are gauged at both idle and load states via sensor and an infrared thermometer. The typical ambient temperature is at 24C. The load state involves sustained writes at maximum speed with measurement ensuing if and until throttling is demonstrated to discover the equilibrium temperature.

The 2TB S90 Lite is fairly efficient although it does not match the prowess of the Gold P31. This drive would be more efficient at 1TB for reasons suggested above. This would be a good choice for a laptop if you need this much space.

This recommendation is also valid with regard to thermals. Our S90 Lite idled at 41C and reached a maximum of 64C after massive sustained writes. The drive did not throttle, although this is a consequence of the drive being very slow once its cache is exhausted.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Conclusion

The Addlink S90 Lite is nothing special, but it’s always nice to have another cheap option to add more fast storage. We tested it at 2TB and that is where it’s priced best. There are potentially some cheaper options at this capacity that utilize QLC, and the Black SN770 is a perennial favorite in recent sales for a TLC option. The S90 Lite has to thread that needle. It’s very similar to the MP600 GS but at this late stage the S90 Lite has to be scored a bit lower, particularly as it has a glaring issue with sustained performance. On the whole, the drive provides a “good enough” experience regardless of where it’s used and it has good availability.

Current DRAM-less, PCIe 4.0 SSDs are an excellent bargain and have been for some time. You get very good performance with an experience that subjectively matches much more expensive, high-end options. These drives are also usually efficient and cool-running. One weakness is capacity as there are fewer controller channels, which can also impact maximum sustained performance. Still, this is the space to watch for value, and controllers with faster bus speeds will max out the interface. Faster, denser TLC will also mean better all-around performance and higher capacities. The S90 Lite is living on borrowed time.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Adata Legend 960 is a high-end PCIe 4.0 NVMe SSD that has arrived late to the market. Performance is merely average aside from sustained writes, which means it doesn’t excel at anything except for very specific workloads. The drive runs cool enough but is not particularly efficient so is not eligible for our list of best SSDs, but would work well in a PlayStation 5. Warranty and software support are good but not exceptional. This leaves the Legend 960 as something worthwhile to pick up on sale but it does not stand out from other high-end PCIe 4.0 drives in any meaningful way. Meanwhile, faster, PCIe 5.0 SSDs are due out soon.

Specifications

The Adata Legend 960 is available in 1TB, 2TB, and 4TB capacities, at $109.99 and $167.11 for the first two capacities. The 1TB price is high given recent sales and general SSD market pricing trends. The 2TB is currently on sale but faces a lot of competition within the same price range, such as the Corsair MP600 Pro XT, the Corsair MP600 Pro NH, and Corsair MP600 Pro LPX.

The 4TB model was not available at the time of review. It has sometimes been challenging to find drives of that size, with the WD Black SN850X and the Inland Gaming Performance Plus being two popular and available options. The Inland Performance Plus that we previously reviewed is also popular and now has updated hardware that matches the Gaming Performance Plus, sans the heatsink.

The Legend 960 is rated for up to 7,400/6,800 MBps for sequential read and write and 750K/630K IOPS for random read and write. The former is essentially the maximum for PCIe 4.0 while the latter falls behind a bit compared to other products, although this is more than enough performance for most. The hardware is certainly capable of more. Adata backs the drive with 780TB of warrantied writes per TB capacity, slightly higher than typical, with all options having a 5-year warranty.

Software and Accessories

Adata provides a download for its SSD Toolbox software package. This application has drive information, diagnostics, cloning, TRIM optimization, a firmware updater, and the ability to perform a secure erase.

A Closer Look

The Adata Legend 960 comes with an installable heatsink/heatspreader that is easy to apply. This is a nice option for a high-end drive like this, especially as it lets you use your own solution that may dissipate heat more effectively if you so choose. The SSD has one DRAM and two NAND packages per side for a total of two and four, respectively. There’s also the prerequisite SSD controller located in the center, a layout that should improve heat dissipation with the spreader.

The controller is an SMI SM2264, a newer design, and this is the first drive we’ve tested that utilizes it. The SM2264 is an eight-channel, 12nm, quad-core ARM Cortex-R8 SSD controller with 1600 MT/s channels. It has 4KB codeword LDPC error correction, double that of the previous generation SM2262EN’s 2K, and can also handle double the dies with ideal support up to 64.

The R8 is a distinct processor revision from the R5, the latter commonly used for SSD controllers, including most or all of the Legend 960’s competitors, and having it manufactured may be why SMI is late to the market. The R8 is also used for the SM2269XT, found on the Solidigm P41 Plus, and it should also form the basis of the upcoming SM2268XT, a controller we will hear more about soon. An R8-based design is potentially very powerful and will be used in SMI’s PCIe 5.0 solutions as well.

The DRAM package is labeled K4A8G165WC-BCTD, which is Samsung DDR4 in a 512Mx16b configuration for 1GB of DRAM. The two modules together are 2GB in total which is the ideal amount for 2TB of flash.

The Legend 960 is using Micron’s 176-Layer TLC (B47R) with thirty-two 64Gb dies in four packages. Each package contains eight dies for 512GiB of flash. This is an ideal amount of dies for interleaving and this particular flash has proven to provide good to excellent performance in our testing.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The 2TB Adata Legend 960 is up against some of the best drives we’ve tested. Our comparison list includes the Samsung 990 Pro, the WD Black SN850X, the Kingston Fury Renegade, the Crucial P5 Plus, the Solidigm P44 Pro, the Sabrent Rocket 4 Plus-G, and the HP FX900 Pro.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The Legend 960 is mediocre to below average in 3DMark. This is a bit disappointing, given that it took SMI a long time to get this controller to market in comparison to Phison’s E18 and InnoGrit’s IG5236.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The Legend 960 does somewhat better in PCMark 10, but still lags the best SSDs, namely the 990 Pro, the P44 Pro, and the Black SN850X. The P44 Pro is also sold as the Platinum P41.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

When it comes to file copying, the Legend 960 performs adequately. There’s not a huge amount of improvement possible given the limitations of the PCIe 4.0 interface and the flash. Controllers with faster channels are on the way, such as the previously mentioned SM2268XT at 3200 MT/s, but improvements here will largely come with PCIe 5.0 drives and 232-Layer TLC.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

No surprises in ATTO. The sequential performance results are satisfactory, minus a dip with smaller reads at QD1. This may be peculiar to SMI’s current R8-based SSD controller architecture as we saw a similar pattern with the SM2269XT on the Solidigm P41 Plus.

Sequential performance in CrystalDiskMark is also average. This is not the end of the world as the numbers are still pretty good. Random performance is often more important and here the Legend 960 unfortunately trails the competition. Good things are possible with the flash being used, so perhaps this is just conservative controller optimization or design to offer third parties more options for their controllers. Such a trade-off is visible with sustained write performance. We are, in any case, more excited for the SM2268XT in the PCIe 4.0 space.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. 

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 2TB Legend 960 writes in pSLC mode for just over 30 seconds at just under 6.5 GBps, demonstrating a cache of almost 200GB. The TLC mode is an impressive 3.59 GBps with a not-too-slow 1.28 GBps folding mode following it. The drive is able to recover after some time back to TLC, allowing it to write an incredible amount in total, second highest of all time for 2TB aside from the Rocket 4 Plus-G.

This is a very good result and the drive is mostly consistent which helps make up for its otherwise lackluster performance results. pSLC recovery is also present and not overly slow. The cache’s size is adequately sized, making this an effective design for any type of use.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

Temperatures are gauged at both idle and load states via sensor and an infrared thermometer. The typical ambient temperature is at 24C. The load state involves sustained writes at maximum speed with measurement ensuing if and until throttling is demonstrated to discover the equilibrium temperature.

The Adata Legend 960 exhibits mediocre power efficiency under load, but its idle power result is excellent. Our idle power testing currently measures what many users would be running on a desktop for maximum performance, which means that the difference is probably not huge relative to the entire system power draw. Additional power-saving features will further reduce power draw, particularly on laptops. The efficiency results paint a better picture of power consumption under load and the Legend 960 is nothing special there.

The 2TB Legend 960 idled at 34C. After 1TB of consistent writes, the drive hit up to 80C but did not quite throttle. This is likely a borderline case given the performance with pSLC as the drive is rated to increasingly throttle at 78C and 81C. This drive could benefit from a heatsink in some systems and for some usage scenarios, such as sustained writes. It seems that Adata has anticipated this as the Legend 960 Max comes with an ample heatsink. The heat spreader is otherwise sufficient.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Bottom Line

The Adata Legend 960 is a mediocre PCIe 4.0 NVMe SSD with average performance and below average power efficiency. The included heatspreader is sufficient to keep the drive from throttling which matches well with its excellent sustained performance, although a full heatsink would be better. Adata’s reputation has taken some hits in recent years, but the drive is backed by a decent warranty with sufficient software support. This makes it a capable product, particularly for desktop and PS5 use, but it is not one that particularly stands out.

One reason it doesn’t stand out is that it’s late to the market, which makes sense since it’s using SMI’s delayed SM2264. It is cool to see this technology and competition in the marketplace, but the Legend 960 faces two important hurdles: drastic NAND price drops and the looming take over of PCIe 5.0 SSDs. It’s a perfectly fine drive but it’s put in a tough spot; it has to be priced right to be worthwhile.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Acer Predator GM7 is a fast and very efficient PCIe 4.0 SSD. In fact, it’s the most efficient SSD we have tested to date by a significant amount. The GM7's all-around performance is good to great, and its performance in sustained workloads is also quite good. The drive is also single-sided and runs cool, making it one of the best SSDs for laptops and one of the best SSDs for the PS5. The GM7 is also noteworthy for using NAND flash from the China-based YMTC fab, but it doesn't use the company's newest 232-Layer TLC flash, meaning the best might be yet to come. 

Acer’s drives are built by Biwin, a company that also supplies HP. The Predator GM7 uses a new Maxiotek controller that can push the limits of the PCIe 4.0 interface with just four channels. More interestingly, this drive also uses 3D TLC flash from YMTC, a China-based foundry. Recent political events have taken YMTC from being disruptive, with the first 232-Layer flash on the market, to a pariah, as it lost a contract with Apple and may be unable to import its flash to the U.S. This has been quite challenging for the NAND manufacturer. Our current Predator GM7 sample uses a revised version of YMTC’s 128-Layer flash, but the more performant 232-Layer variant is promised and would make for an even better drive. Competing drives will likely arrive with Micron’s 232-Layer TLC and the TenaFe TC2201 controller, but the GM7 is the efficiency king for now.

Specifications

According to the datasheet we received, the Acer Predator GM7 will be available in 512GB, 1TB, and 2TB flavors. The product’s box also lists a 4TB SKU, which would be a tall order for this drive’s budget controller. It’s technically capable of 4TB but would need 1Tb dies. Performance peaks at the 2TB model.

The 1TB Predator GM7 is rated for up to 7,400/6,300 MBps of sequential read/write performance, but the datasheet implies a future 2TB model could hit 7,400/6,700 MBps. The performance in random workloads is currently not defined, but the similar Zhitai TiPlus7100 SSD can reach up to 900K/700K in random read/write workloads, and the controller is capable of up to 1M/1M.

The Predator GM7 has a five-year warranty, and the 2TB model has a 1,300 TBW (total bytes written) endurance rating. Thus, the drives would likely have a 650TB-per-TB-of-capacity rating, but be sure to look for an official spec for the other capacity points.

The datasheet lists NVMe 2.0 support, and the controller also supports this spec revision, but the box and CrystalDiskInfo indicate NVMe 1.4 for this drive. Our model is listed as BL.9BWWR.118 on the box, and given the product naming conventions of the Acer Predator GM7000, we made some guesses for 512GB and 2TB models. 

At the time of review, the Predator GM7 was not widely available, but pricing is expected to be $49.99, $89.99, and $159.99 for the three launch capacities. This is not competitive with mid-range PCIe 4.0 SSDs like the Silicon Power UD90, which have been $74.99 or less for 1TB, but the pricing gives it a good starting point for sales against high-end PCIe 4.0 drives like the WD Black SN850X. 

Software and Accessories 

The Acer Predator GM7 does not list free Acronis cloning software on its datasheet as the Predator GM7000 does. We recommend Macrium Reflect Free for your imaging and cloning needs. For information about the drive, use CrystalDiskMark or similar software to check drive health and other data.

A Closer Look

The 1TB Acer Predator GM7 is a plain, single-sided SSD. There is a small controller with no DRAM package and two NAND packages. It is possible to populate this PCB with up to four NAND packages. It would be possible to get up to 4TB with 1Tb dies, but this drive would be best at 2TB with that flash, and possibly best at 1TB with the 512Gb dies our sample uses. The single-sided design makes it a good fit for some slim devices, like laptops.

The controller is the Maxio MAP1602A, also known as the MAP1602 or MAP1602-I. Maxio, or Maxiotek, is a spin-off from JMicron, which once made SSD controllers. This particular controller is made in the 12nm TSMC process node with an ARM Cortex-R5 architecture, distinct from the 22nm TSMC node used on the MAP1202. It is four-channel and DRAM-less, suggesting it should be very efficient. Unlike other DRAM-less PCIe 4.0 controllers we’ve tested, this one has a 2400 MT/s bus that allows it to saturate the interface. This will also make it more power efficient.

The flash is Biwin-binned and packaged YMTC TLC. Like the TiPlus7100, this drive was expected to have YMTC’s newest 232-Layer TLC but actually arrived with a stopgap variation of its 128-Layer design. The 232-Layer TLC will have six planes like Micron’s 232-Layer TLC, as seen in our Phison E26 Preview, along with independent plane operation similar to Micron’s. This flash also has a center X-decoder, like SK hynix’s 176-Layer TLC in the Platinum P41, improving sub-page I/O performance. Other changes will help bring the complexity and cost of the flash down — an area that has been an issue — and allow high die density.

The four-plane 128-Layer variation found on our review sample does not have the new Xtacking 3.0 technology and has a different 2x2 plane layout than the original 128-Layer design. YMTC’s Xtacking uses wafer-on-wafer technology to bond separate CMOS circuitry — page buffers, charge pumps, etc. — on a flipped flash array. Architectures from other flash manufacturers instead have the CMOS under the memory array. YMTC’s design reduces development time and the time to market, plus it has benefits for die density and efficiency, but it is more costly. Challenges include bonding wafers with distinct characteristics, such as different coefficients of thermal expansion (CTE), which require novel vias/plugs to handle the stress.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The Acer Predator GM7 is up against the WD Black SN850X, the Crucial P5 Plus, the SK hynix Gold P31, the Acer Predator GM3500, the Samsung 980 Pro, the Solidigm P44 Pro, and the Kingston Fury Renegade.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The Predator GM7 scores average to above-average scores in 3DMark. This is about where we would expect it to fall given its hardware — above PCIe 3.0 and older PCIe 4.0 drives, but below newer PCIe 4.0 options.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The Predator GM7 does a bit better in PCMark 10, managing to beat the Fury Renegade. The GM7’s fast bus helps it keep up and deliver a solid user experience.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

The fast bus also helps here, with file transfers, but the Predator GM7 can’t quite keep up with newer drives like the SN850X or P44 Pro or with the Fury Renegade’s newer flash. It’s possible that with 232-Layer TLC the Predator GM7 would perform a bit better.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The sequential test results for the Predator GM7 are quite good, but not record-breaking. The drive’s bus and interface speeds continue to help. This is reflected in the sequential CrystalDiskMark results, where the drive does quite well. This isn’t quite bleeding-edge performance, but you can get close to that with some cost and power savings. Performance in random workloads isn’t quite as good, although the flash may be a factor here. An upgrade to YMTC’s 232-Layer TLC would likely help a significant amount.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. 

We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 1TB Predator GM7 exhibits three primary performance states during sustained write workloads: pSLC, TLC, and folding. It writes at about 6GBps for 28 seconds in pSLC for a cache of almost 170GB. This is not as fast as other drives, and the cache is on the small side, but it does well enough. TLC mode writes at nearly 2.07GBps, which is quite fast and in the vicinity of flash on other four-channel PCIe 4.0 drives like the Team Group MP44L. Folding writes at 770MBps, which, while not amazing, is also not terrible. Cache recovery appears rapid but falls back to TLC speeds in short order.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

Temperatures are gauged at both idle and load states via sensor and an infrared thermometer. The typical ambient temperature is at 24C. The load state involves sustained writes at maximum speed with measurement ensuing if and until throttling is demonstrated to discover the equilibrium temperature.

Power efficiency is where the Predator GM7 shines: It is by far the most efficient SSD we have tested. This is expected as the controller is four-channel and DRAM-less and is paired with YMTC’s Xtacking flash technology. It’s more efficient than other four-channel, DRAM-less PCIe 4.0 drives we’ve tested because the faster bus allows full use of the interface to transfer files quickly. We anticipated this in previous reviews and pointed out that the TenaFe TC2201 may have similar results. However, better things are to come from newer flash and, eventually, “low-end” PCIe 5.0 controllers.

Acer markets this drive as having thermal throttling and power management for efficient heat dissipation and better laptop battery life. It’s also great for the PlayStation 5. The drive is set to throttle at a relatively high 90C initially, and with its small cache, it won’t get there with a normal ambient temperature. Our GM7 idled at 43C for the controller via infrared but never throttled even with a full drive’s worth of writes, reaching only 73C. It may benefit from a heatsink in certain environments for very specific applications.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Conclusion

The Acer Predator GM7 is the most efficient SSD we have ever tested, and it's no slouch with regard to performance. It's good to great in most benchmarks, and performance in sustained workloads is also reasonably good. The great power efficiency helps keep the drive cool, and the single-sided design makes this a great choice for laptops and the PS5. We already liked mid-range PCIe 4.0 drives with four-channel, DRAM-less, 12nm controllers as some of the best values on the market, but pushing the bus further makes the GM7 an even more impressive specimen.

Not everything is perfect, however, as pricing will require threading a needle. The GM7 and similar drives have to cost less than the highest-end drives and cannot cost much more than mid-range PCIe 4.0 alternatives. Sequential performance is not everything, so the choice of flash is also a consideration. The revised 128-Layer YMTC TLC on this drive performed well, but it could be better in some areas — such as random workloads — if Biwin upgrades the drive to 232-Layer flash. Competition is also coming in the form of SSDs outfitted with the TenaFe TC2201 controller and Micron's 232-Layer flash.

This drive is otherwise fairly easy to recommend, if you can find it. Availability is always a concern, which is complicated by the fact that the flashmaker has come under restrictions from the US government. However, we're told those restrictions won't impact this 128-Layer version of the drive, and the impact on possible future 232-Layer variants isn't quite clear yet.

Acer's warranty is solid, but you may want more substantial software support as you would have with, for example, WD's drives. However, if you're looking for a budget-leaning drive that gets you high-end-like performance with great efficiency, it's hard not to suggest the Predator GM7. It demonstrates good things to come for the SSD market and may also be a YMTC harbinger.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Kingston NV2 is one of those SSDs that is just too good to be true. The price is exceptional, especially at 2TB, and it is sold as a PCIe 4.0 drive. Yet it has no definitive hardware inside and its performance for both of our samples is distinctly bottom of the barrel. It also runs a little hot and inefficient in our testing, too. 

Miles behind the best SSDS, the NV2 may still make a good secondary or backup drive if your budget is absolutely strapped but it is not recommended for primary or laptop use. It is on the whole a good example of caveat emptor - buyer beware.

The NV2 follows Kingston’s previous NV1, a drive very much in line with Kingston’s philosophy of providing cheap drives at scale. The SATA-based A400 is the perfect example of this as it was a prevalent drive in cheaper builds. The NV1 followed suit and became more popular than it perhaps deserved. The NV2 is similar in that it uses a hodge-podge of hardware - different controllers and NAND flash from drive to drive - offered at an insanely low price. It should become commonplace especially in regions with limited SSD options. If you actually have solid alternatives, however, please look elsewhere.

Specifications

The Kingston NV2 is available in 250GB, 500GB, 1TB, and 2TB. At the time of review, pricing for these capacities was $22.99, $34.99, $54.00, and $124.99, respectively. This drive is often on sale with the 1TB and 2TB SKUs becoming even better values. This drive is the A400 of NVMe drives and a fitting replacement for the NV1, which could be good or bad depending on how you look at it. In essence it is a dirt cheap NVMe SSD that meets the minimum requirements to get the job done.

The drive can manage up to 3,500 / 2,800 MBps for sequential read and write, respectively, but has no random performance specifications. This makes sense because it can come with more than one controller and more than one type of flash. The sequential write specifications are such that it can only have QLC at 1TB and 2TB, however. The sequential values are low for a PCIe 4.0 drive for a good reason: Kingston set them for the weakest possible controller and flash.

The NV2 has a 3-year warranty and can manage 320TB of writes per TB capacity. This is exactly as expected for a budget drive.

Software and Accessories

The Kingston NV2 is a barebones drive but Kingston does offer an SSD Manager on its site. This has typical SSD toolbox functionality and is able to display disk information and health, update drive firmware, and securely erase drives. It only works on Microsoft Windows.

A Closer Look

Although we are looking at the 2TB in the photos here, mention will also be made of the 1TB because the flash and controller on the Kingston NV2 will vary from drive to drive. The 2TB drive is single-sided with a DRAM-less controller and four NAND packages. There’s not much to this drive but it should always be single-sided for where that matters.

The 2TB model we sampled is using the SMI SM2267XT controller. This is one of those entry-level PCIe 4.0 SSD controllers that barely qualifies for the 4.0 moniker. It’s similar to the Phison E19T controller used on drives like the WD Black SN750 SE or the Inland TN436. This technology has a 1200 MT/s bus which, with four channels, means it can saturate a x4 PCIe 3.0 link, much like the controllers on the SK hynix Gold P31 and WD Blue SN570.

For all purposes this makes the controller a jacked-up SM2263XT with higher bandwidth and IOPS but not really new technology like the SM2269XT. This performance requires a much higher core clock rate. This means the power efficiency should be fairly terrible for a budget drive when compared to other, newer PCIe 4.0 DRAM-less options like the HP FX900 and Silicon Power UD90.

Our 1TB sample also uses the SM2267XT, but this drive has been spotted with the newer SM2269XT as well. Given the limited performance specifications, it’s possible this drive could also come with the comparable E19T as well. Other, faster controllers closer to the SM2267XT, such as the InnoGrit IG5220 and Phison E21T, are also technically possible.

Four 512GB NAND modules with four 128GB dies deliver a total of 2TB. This is Intel’s 144-layer QLC. This flash is used on the Intel 670p and Solidigm P41 Plus. It remains arguably the best QLC on the market but is still QLC.

Our 1TB sample arrived with Kioxia 112-layer BiCS5 TLC, instead, but QLC is a possibility at that capacity.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The 2TB Kingston is up against other 2TB drives including the Crucial P3 Plus, the Adata XPG Gammix S50 Lite, the Corsair MP600 GS, the SK hynix Gold P31, the Solidigm P41 Plus, the WD Black SN770, and the Samsung 990 Pro.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The NV2 comes in dead last and by a significant margin. It’s the worst for score, bandwidth, and latency. There’s really no excuse here as the P41 Plus, which has the same flash, is far superior and the S50 Lite, with the same class of controller but with DRAM, is also faster. The NV2’s older, DRAM-less controller just can’t hang. It should still be close enough with load times so is fine for a dedicated games drive.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The NV2 is below average in PCMark, although it manages to still beat the popular Gold P31. The S50 Lite is using the DRAM-equipped version of the NV2’s controller, the SMI SM2267, but with older flash.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

The NV2 copies at a slow speed, no better than the PCIe 3.0 Gold P31 or the entry-level PCIe 4.0 drives, entry-level meaning glorified PCIe 3.0 drives like the S50 Lite and NV2. There’s little reason to waste a 4.0 slot on these drives.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

The NV2 is essentially dead last in ATTO. CDM sequential performance is below average to poor. 4KB performance is also relatively poor, although it beats out the P3 Plus for reads at a queue depth of 1. This is adequate for a gaming drive but basically only provides a budget experience.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 2TB NV2 maintains 2.95GBps in pSLC mode for over 190 seconds for a cache of 561GB. This implies that the entire drive is capable of pSLC mode so the dynamic cache should usually be about one-fourth the size of the free space. The cache is relatively quick to recover, as expected.

QLC mode flatlines around 240MBps which is not terrible for QLC but still pretty poor on the whole. This is about what we would expect from this flash when using such a wide cache. Comparisons to the P3 Plus and P41 Plus are enlightening to see how cache size and pSLC speed impact QLC performance. This is a drive that will suffer when fuller, particularly with prolonged writes.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

Temperatures are gauged at both idle and load states via sensor and an infrared thermometer. The typical ambient temperature is at 24C. The load state involves sustained writes at maximum speed with measurement ensuing if and until throttling is demonstrated to discover the equilibrium temperature.

NV2 power efficiency is fairly terrible, as expected. This will still be an unfortunately popular drive for laptops as it’s single-sided and dirt cheap. If this drive isn’t regularly pushed, and it shouldn’t be, it may be passable with its power usage, but there are many better options.

Normally with QLC-based drives we won’t see throttling because the QLC mode is too slow. The 2TB NV2 runs hot, though, another characteristic of its controller since it’s 28nm technology. The drive idled at 56C by IR and 37C in SMART, a wide deviation between measurements, then hit 86C and 70C, respectively, after roughly 500GB of sustained writes. At this point the drive began to throttle.

These temperatures make it difficult to recommend for laptop use. While it takes a lot of writes to get the drive to throttle, the high idle temperature is concerning. A heatsink is recommended for other use.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Conclusion

The Kingston NV2 is simply not a drive we can recommend in most cases because the range of hardware it can have makes it too risky for the buyer. You could get the great SM2269XT or the mediocre SM2267XT controller. You could get BiCS5 TLC or much slower QLC. None of these will have DRAM and it’s probably unfair to call this a PCIe 4.0 drive on top of all that. 

In testing, the drive also ran hot even when idle and it draws a decent amount of power, making it less than ideal for many machines, including laptops. Its saving grace is its amazing price, which does make it perhaps good for something other than a primary drive.

Despite this, you may not want to roll the dice and hope you get lucky. It is probably better to approach it from a worst-case perspective: this is a budget drive designed to be a cheap way to add more NVMe-class storage. If you have worked with the A400 or NV1, then you know what to expect with the NV2. Many motherboards come with their own M.2 heatsinks or shields, which can also help alleviate any heat problems this drive might have, mitigating one of its downsides.

Our samples both use the SMI SM2267XT controller which did not perform well with either TLC or QLC. The flash, in either case, is actually not bad. It’s just that there are so many better options on the market. Pricing really is the ultimate factor here that you must consider when looking at the NV2. It may be okay for a gaming drive if it’s priced competitively against SATA alternatives, but this implies you want higher capacity. At lower capacities it’s simply a way to upgrade an old machine cheaply - but we did not review those SKUs.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Corsair MP600 GS is another decent mid-range PCIe 4.0 SSD, although it feels like it’s arriving too late and at too high a cost. It’s nice to have this level of performance in this price range and get it at up to 2TB, but there’s a lot of competition in this challenging market. 

The MP600 GS has no real issue stepping over older PCIe 3.0 SSDs, including those with DRAM, in both performance and efficiency. Between drives in its own class however, the main thing that sets it apart is its unique pSLC setup. There’s also the Corsair name to consider, which may carry some weight, but the actual hardware is nothing new.

This drive could use a price cut, or Corsair could have made it more attractive with a heatsink option. The uniquely-sized pSLC cache is such that the MP600 GS has very good sustained write performance and consistency, which would also be a good fit with a heatsink. It seems likely Corsair simply wanted to jump on the mid-range bandwagon – and there’s nothing wrong with that. But without a price cut at least, the MBP600 is too little, too late.

Specifications

The Corsair MP600 GS is available in the 500GB, 1TB, and 2TB capacities. It’s currently priced at $57.99, $92.99, and $177.99 at those capacities, respectively, which are all a bit on the high side compared to similar SSDs. That said, it’s nice to see a 2TB option, as budget drives are often limited to 1TB, which we saw on the Silicon Power UD90 in North America.

The MP600 GS is a mid-range PCIe 4.0 SSD offering up to 4,800/4,500 MBps for sequential reads and writes and up to 530K/1000K read and write IOPS at 2TB. This drive has a full 5-year warranty, with up to 600TB of rated writes per TB of capacity. Corsair also lists encryption support which, although supported on the Phison E21T controller, is only optionally activated by the manufacturer. Similar drives, like the UD90, lack this feature.

Software and Accessories

Corsair does provide an SSD toolbox for download. This piece of software allows you to check the drive’s health data via S.M.A.R.T. and discover other pertinent information about the drive. Features include secure erase, disk clone, and TRIM functionality. It feels a bit rudimentary, but it gets the job done without having to go searching for a third-party option.

A Closer Look

Corsair has gone with a minimalist design on the MP600 GS. Under the top label, we spot a controller wedged between four NAND packages – two to each side. This is single-sided at 2TB, which means it's slim enough for any M.2 device.

The controller is a Phison E21T. We’ve seen this on the UD90 and Team Group's MP44L, which have the same basic hardware as the MP600 GS. The E21T has also shown up with QLC on the Crucial P3 Plus and Crucial P3. We have nothing bad to say about this controller. Its peers include the InnoGrit IG5220 and the SMI SM2269XT, the former particularly well-known and used in the HP FX900 and Patriot P400. However, a new manufacturer by the name of TenaFe looks to shake up the field, so drives with its TC2200 controller may soon invade the E21T’s space.

The flash is labeled IA8HG94AYA, which we know to be Micron’s 176-layer TLC. This flash is ubiquitous and has proven to offer great performance with decent power efficiency.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The 2TB Corsair MP600 GS faces the WD Black SN770, the Crucial P3 Plus, the Solidigm P41 Plus, and the Corsair MP600 Pro XT, all at 2TB. 1TB competitors include the Samsung 980, the HP FX900, and the Patriot P400. All but the MP600 Pro XT and 980 are within the same class of drive as the MP600 GS with the former being high-end and the latter PCIe 3.0. The P41 Plus and P3 Plus also stand out as the only QLC-based drives in the list.

Trace Testing - 3DMark Storage Benchmark

Built for gamers, 3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The MP600 GS delivers mediocre marks in 3DMark. On the whole, there are only two in-class drives that beat it:  the Black SN770 and the P41 Plus. The Black SN770 remains an amazing drive. The P41 Plus only barely edges out the MP600 GS, but it’s a QLC option that’s cheaper than both the MP600 GS and the SN770. The 980 is merely average here, but this is a good showing for a PCIe 3.0 SS

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The MP600 GS does poorly in PCMark, although it’s not really far off from most of the drives tested. Once again the Black SN770 and P41 Plus stand out. At least some of this is from firmware optimization. Even if the results do not demonstrate a direct performance relationship, they do reflect an intention of designing a drive geared toward a better real world experience.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

The MP600GS is above average in DiskBench, but it’s essentially on the same page as competing drives like the FX900 and P400. The Black SN770 has traditionally done well on this test, too. The MP600 Pro XT easily pulls away.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

Read results from ATTO are stellar on the MP600 GS, as we would expect from Phison controllers. This also applies to the P3 Plus and the MP600 Pro XT. ATTO writes are more subdued, limited to some extent by the controller interface, with the MP600 XT Pro pulling easily ahead again. We have actually seen anomalies in ATTO testing with some Phison drives that have been confirmed by other review sites. Things are more clear with CDM.

CDM sequential performance for the MP600 GS is as expected - right with the rest of the pack, except the MP600 Pro XT. The exception is with QD1 read, where the SN770 pulls away and the MP600 GS falls a bit behind. As the MP600 GS is holding right with the P3 Plus and P41 Plus, here, it’s likely a facet of the Phison E21T controller. The InnoGrit IG5220 controller, used on the FX900 and P400, is faster.

The drive’s 4K performance is mediocre, particularly at a low queue depth. This is one area in which the IG5220 is simply superior, scoring well with reads, which remains an important benchmark. One standout here is the P41 Plus, which manages to have slightly better latency despite using QLC. Part of this is due to SLC caching and part possibly due to optimizations for reads with Solidigm’s 144-layer QLC, but it also suggests that Solidigm tunes its drives carefully for client use.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 2TB MP600 GS writes at about 4.8 GBps for over 14 seconds in pSLC mode, with a cache of nearly 70GB. After that it writes at 1.78 GBps in TLC mode, which is about half the maximum speed of drives with this flash that use the Phison E18 controller, like the MP600 Pro XT. This is logical because the MP600 GS’s E21T controller has half the channels and therefore half the interleaving. This cache layout is different from the one on the FX900 and P400. On those drives, the cache is larger with an eventual decline to a slower folding mode after writing to TLC.

Only some E18-driven drives, like the MP600 Pro XT, opt for a smaller cache in return for better sustained performance. This design choice makes the MP600 GS unique, as it’s capable of very fast sustained performance. This will be better for certain prosumer tasks, but also will maintain performance with a fuller drive or after a lot of writes, scenarios you might face when using it as a caching drive. The cache is still large enough to manage random writes and, because it is smaller in absolute terms, it will maintain its size more consistently across the drive as it is filled.

The drive is capable of recovering some cache relatively quickly, which can help absorb random writes, but otherwise it does not seem in a hurry to move data over. This is a reasonable design decision because random writes tend to incur higher write amplification and are ideally cached in pSLC. Meanwhile typical usage, that is if the drive is used in a normal gaming computer, will likely offer plenty of idle time to restore the full cache. This potentially makes the MP600 GS more attractive than the alternatives in this price bracket for specific cases.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

Temperatures are gauged at both idle and load states, via sensor and an infrared thermometer.

One great thing to come out of the age of new PCIe 4.0 DRAM-less controllers, as well as 176-layer flash, is the preponderance of efficient drives. The MP600 GS is no exception, beating everything here except the FX900. The SN770 and P3 Plus come pretty close, though.

Not shown are the Team MP44L and Silicon Power UD90, drives with similar hardware as the MP600 GS. These have displayed better efficiency in our prior testing, which is likely a consequence of differences in pSLC caching. It’s possible then, that a longer or more variable test would have the MP600 GS win out.

The 2TB MP600GS idled at around 40C. Under load it reached a maximum of 80C after approximately 600GB of writes. Write performance was not perfectly consistent past this point, demonstrating some dips. This drive could benefit from a heatsink if you are purchasing it for anything heavy duty, and in fact we would recommend that for some applications and hotter environments.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Conclusion

The Corsair MP600 GS is your run-of-the-mill mid-range PCIe 4.0 SSD. Its performance is not amazing but certainly good enough, as is its power efficiency. In some areas it does stand out, specifically with sustained writes. This can make it useful for some tasks and also more consistent in edge cases. It’s nice to see it available at up to 2TB, too, but it has stiff competition from excellent drives like the WD Black SN770. The Corsair name will only get it so far, as it doesn’t offer anything special like an option for a heatsink, although the 5-year warranty is decent.

Our recommendation would be to wait for a price drop, as this drive is at least competent within its class. There’s nothing wrong with similar offerings from Silicon Power or Team Group, but Corsair perhaps has a bit more cachet in some circles. This may or may not be deserved, although we would point out that the drive's cache design does suggest Corsair was aiming at a more reliable experience. This could make the MP600 GS more broadly appealing, but the differences in daily use would be negligible.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Amazon Web Services hosts some of the world's most valuable data for commercial companies in its cloud, but when Russia invaded Ukraine earlier this year, it faced challenges that it probably had not faced before. AWS had to secure vital government data hosted in Ukraine and then bring all the digital services the country offered back online, reports Business Insider.

It was not particularly easy to move petabytes of crucial data using secure Internet channels, as Ukraine was heavily bombed and usage of the web increased significantly. To get physical copies of Ukraine's critical infrastructure and economic information data and preserve it, AWS shipped custom-built Snowball Edge solid-state drives to Poland and then drove them to Ukraine.

Unfortunately, AWS does not show the Snowball Edge suitcase-sized SSDs used to physically transfer data from Ukraine to safety. We know that datacenter Snowball Edge machines offer up to 80TB of usable storage space, and we presume these are all-flash arrays of extreme storage density that are also designed to be extremely reliable.

AWS currently hosts over 10PB (petabytes) of vital data from 27 Ukrainian ministries, 18 Ukrainian universities, the largest remote learning K–12 school (serving hundreds of thousands of displaced children), and dozens of various private sector companies, the cloud giant said. There are now 61 government data migrations to AWS with more to come.

PrivatBank, the largest bank in the country, has migrated 270 applications and 4PB of client data residing on 3,500 Ukraine-based servers to AWS. Interestingly, PrivatBank plans to keep using the AWS infrastructure even when the war is over, which is a departure from banks' usual approach to use on-premises machines to run their mission critical software.

Digitization of these data and services ensures the continuity of the Ukrainian state and allows people to access information they need from whenever they are. This data is important even today when students want to continue their education elsewhere, but it will also be crucial tomorrow, when the country is rebuilt. "Government in a box" servers and storage allows Ukraine to avoid the chaos associated with war and post-war periods, something that's hard to overstate.

"We used to assume that this is just how it is in war — everything gets destroyed and you have to rebuild from nothing," said Liam Maxwell, director of the Government Transformation team in the AWS public sector business and a volunteer in AWS’s efforts to aid Ukraine. "But by migrating to the safety and security of the cloud, the government and its citizen services prevail."

The Samsung 990 Pro is the much-anticipated successor to the popular 980 Pro. It is poised to be the very fastest drive we’ve tested, outdoing even the SK hynix Platinum P41 and the Western Digital (WD) SN850X. Initially available at 1 and 2TB for $169 and $289, respectively, the 990 Pro will also have a 4TB model out next year. Available bare or with an RGB heatsink, the 990 Pro is using Samsung’s new Pascal controller and it's new V7 TLC NAND flash, a combination that promises to be faster and more efficient.

The 990 Pro has the rest of the Samsung trimmings, including software support with the updated Magician application and the option for encryption. The drive is designed to fit in a wide range of devices, from desktops to laptops to the PlayStation 5 (PS5) console. As Samsung’s new top-of-the-line consumer PCIe 4 SSD, the 990 Pro is designed to take a place among the best SSDs in terms of performance and, according to our tests, it delivers. It also takes the crown of the fastest SSD on our list of Best SSDs for the PS5. 

Samsung 990 Pro Specifications

The Samsung 990 Pro arrives in the 1TB and 2TB capacities at launch. The 4TB model is coming later, in 2023; it will be nice to see a larger capacity option from Samsung. The 990 Pro promises up to 7450/6900 MBps, sequential read and write, with up to 1.4 million / 1.55 million read and write IOPS. This is more than competitive, exceeding the previous-gen, 980 Pro on all counts but also promising numbers higher than the SK hynix Platinum P41 and WD SN850X.

The Samsung 990 Pro does support TCG Opal encryption - an unsupported feature for many consumer SSDs - and offers 600TB of warrantied writes per TB of capacity. The endurance rating follows the JEDEC JESD218 standard which is largely irrelevant. The TBW ratings are unexceptional in any case, but should be sufficient for the drive’s intended use.

Samsung is offering a variant at each capacity with a heatsink and RGB, taking a page out of the WD SN850X’s playbook. It appears that the 4TB model will also have this option which was not the case with the SN850X. The Platinum P41 is therefore perhaps a bit less attractive by not having a heatsink variant. The Samsung 990 Pro’s heatsink meets the PCI-SIG D8 standard, which means that it comes in at less than a 8.8mm height to ensure it fits into a wide variety of devices, including the PS5.

The Samsung 990 Pro launches with a price tag of $169.99 for 1TB and $289.99 for 2TB. The heatsink and RGB add another $20 to each price, which matches the premium of the SN850X. One complaint we had about the SN850X is that its launch price was too high - something that has since changed drastically. Samsung also had a hefty MSRP for its T7 Shield launch but the drive was never actually sold for launch prices. We therefore expect the 990 Pro to be discounted, which is reasonable given the current real-world pricing of its direct competitors.

Software and Accessories of the Samsung 990 Pro

Samsung released version 7.2 of its Magician storage software on October 4th, in advance of the 990 Pro’s release. This software now allows for RGB control for the heatsinked versions of the drive. It also enables data migration and offers diagnostics, firmware updates and driver updates. Other functions are also present such as a PSID revert, which performs a secure erase and unlocks an encrypted drive. Samsung’s SSD software is, in general, the standard for the industry.

This software also has a Performance Optimization panel which allows for a Full Performance Mode. This is reminiscent of WD’s original Game Mode, which effectively disabled lower power states. In our tests, we benchmarked with the “Full Power Mode” both on and off.

A Closer Look at the Samsung 990 Pro

The 990 Pro has unexceptional looks, with a label on both front and back providing basic drive information. The heat spreader back label helps reduce load temperatures. Under the top label there are two NAND packages, a DRAM package, and the controller. The controller is nickel-plated to improve heat dissipation, but the 990 Pro also uses Samsung’s Dynamic Thermal Guard (DTG) to improve thermal consistency. We’ve seen this before in the 980 Pro but this is also similar to the SN850X’s adaptive thermal management feature. This technology is nothing new, but may help with gaming workloads.

The Samsung 990 Pro sports a new controller, Pascal, as a successor to the 980 Pro’s Elpis. The Pascal is produced in Samsung’s 8nm process node and is still ARM-based. There does not appear to be much difference between the controllers superficially, but the Pascal has significantly higher performance specifications. Samsung states that this is thanks to an optimized NAND data path through “hardware automation technology” with reduced processing time via a “cache algorithm.” When I asked about the changes, Samsung stated that this architecture was meant to have more effective low-power mode coverage and that it was capable of taking advantage of read caching.

Hardware automation involves the storage data path: the flow of data from the host interface to the flash. As SSDs scale up to greater levels of parallelization with more complex addressing, IOPS bottlenecks begin to arise. Hardware automation helps overcome these and also offers power savings which is becoming more important as drives get faster. One example is hardware-accelerated flash map management which can increase IOPS versus managing the flash translation layer (FTL) - which translates between physical and logical memory addresses - in software. I/O queues and data transfers also benefit from acceleration. Separately, parallel processing improvements in firmware can also improve garbage collection/scheduling as with Phison’s I/O+ Firmware.

Volatile memory like SRAM is used by the controller to cache mapping information and buffer data to commit to the non-volatile NAND flash, so optimization of algorithms can improve overall performance. Write performance often benefits more from such advances, but changes may also be relevant for DirectStorage. Some manufacturers, like Solidigm, have also decided to introduce a form of caching to hold specific data in the cache via a specialized NVMe driver. In that case, knowledge about the type of data and how it is utilized - metadata - can bolster performance with intelligent caching. Samsung did not elaborate on what it meant by “read caching,” but it is not uncommon to keep some hot data in pSLC to improve forthcoming reads.

A typical cache algorithm would be least recently used (LRU), where the least recently accessed data would be evicted first from a full cache. Such an algorithm is constrained if it does not take into account spatial locality, that is, knowledge of adjacent and nearby memory locations. The inevitable increase in FTL overhead with more complex algorithms can create a performance bottleneck. Specific improvements to the controller can increase maximum IOPS by offloading or automating some of this work. This can also generally lead to lower latency, although the full benefits are unlikely to be realized on a consumer device.

Samsung has certainly made advances with the controller but other components are more familiar. The memory, or DRAM, is Samsung’s LPDDR4 which has been used on many of its SSD products. It offers a bit of power savings over traditional DDR. This module is 2GB in size, which matches the 2TB of flash with an ideal ratio. It’s expected that the 1TB model would have 1GB. The power components are also standard for Samsung. We would expect the promised power savings, in comparison to the 980 Pro, to come from the new flash plus controller optimizations.

The flash packages are labeled K9DVGY8JRD-DCK0 which denotes that it is Samsung’s 176-layer TLC NAND. With the 2TB SKU, this flash has 512Gb dies as indicated by “V” - although the upcoming 4TB may require 1Tb “X” dies. This is V7 in Samsung’s V-NAND nomenclature which was explored in some detail at ISSCC 2021. Samsung showed off its 8th generation TLC at ISSCC this year but recent trends in the flash market have generally left manufacturers in a holding pattern with regard to the production of higher-layer flash. In any case, this is the first time we’ve seen this flash, a direct upgrade to the V6 128-layer TLC found on the 980 Pro.

Samsung made many improvements with this generation of flash, the most significant being the use of a four-plane design and a Cell-on-Periphery (COP) implementation. More planes means more parallelization which translates at least to higher bandwidth. COP is similar to Micron’s CMOS-under-Array (CUA) but this is the first time Samsung is utilizing such technology. This involves moving control circuitry away from the peripheral or side of the array. This can greatly improve power efficiency and also reduces die surface area by placing peripheral circuitry under the data cell array.

Placing this circuitry under the array introduced some problems for Samsung who solved it by using innovative capacitor design. The result is improved power delivery with reduced surface area requirements. Samsung also uses a dual-scheme termination approach that allows for increased power efficiency when full I/O speeds are not required. This should help Samsung back up its claim of better power efficiency through the use of a low-power-aware architecture.

Samsung is the one manufacturer that was able to avoid string-stacking, that is the use of multiple NAND array decks, with higher flash layer counts. Its approach comes with many challenges but also avoids having to deal with melding decks. Etching so many layers leads to an increased aspect ratio which, among other things, can increase the level of voltage threshold deviation between cells in different layers. More layers in the same space does translate to an effectively smaller cell capacity which can impact performance and endurance, as well.

Samsung’s solution introduces an additional latch - a type of dynamic buffer, as with page buffers - to overcome this issue. The extra data is utilized with bit and word line forcing schemes to produce a refined cell charge, essentially having an extra verify stage to keep bit values tighter. While a manufacturer could simply keep adding decks - and it’s theoretically possible to hit 800 or more layers with that technique - it’s not without its own challenges. Samsung has instead decided that its 176-layer NAND provides the optimal balance between performance and power efficiency.

MORE: Samsung Promo Codes

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The 2TB Samsung 990 Pro must be compared to its immediate predecessor, the Samsung 980 Pro, but also to its two largest likely rivals: the WD Black SN850X and the SK hynix Platinum P41. The Crucial P5 Plus is another popular proprietary drive. Other competitors use InnoGrit’s IG5236 controller, as with the HP FX900 Pro, or Phison’s E18 controller, as with the Kingston KC3000 or updated Sabrent Rocket 4 Plus. These final two drives have minor differences, such as with pSLC caching. These last three drives are all using Micron’s 176-layer TLC. We tested the 990 Pro in both the default and Full Power modes.

Trace Testing - 3DMark Storage Benchmark

3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files, and recording gameplay video streams.

The 990 Pro delivers the best results we’ve seen in 3DMark to date, beating the SN850X and Platinum P41. Full Power Mode improves performance even more.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The 990 Pro also brings astonishing results in PCMark 10 with an even greater lead than it had in 3DMark. Full Power Mode doesn’t offer any gains this time, but the 990 Pro simply devastates its 980 Pro predecessor in this test.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

In DiskBench we can see that high-end PCIe 4.0 drives are essentially limited by the interface. The 990 Pro stil does well, but it’s basically on par with the SN850X in the copy test. This is one test where the Platinum P41 remains on top.

Synthetic Testing - ATTO / CrystalDiskMark

ATTO and CrystalDiskMark (CDM) are free and easy-to-use storage benchmarking tools that SSD vendors commonly use to assign performance specifications to their products. Both of these tools give us insight into how each device handles different file sizes.

There are no big surprises with ATTO. We do see some variation in the 990 Pro at higher block sizes, as we did with the 980 Pro. These variations are reversed at 1MB and 2MB, likely due to differences in the architecture of the flash. Samsung has moved to a four-plane design at 176-layer while at 128-layer it remained dual-plane; SSDs actually open a superpage across all planes to improve performance. Controller handling here can lead to some unwanted inconsistency, although the real world impact is debatable.

CDM’s sequential results indicate that the 990 Pro is better than the 980 Pro and generally it fares well. The exception is at QD1, although performance is still relatively good. Random performance is excellent with reads, particularly in Full Power Mode, but write performance is lackluster in both modes. The former result is not surprising given the relatively low read latency of this flash at 40µs, an area in which Samsung excels. Random read, low queue depth performance tends to be relevant for a drive’s “real world” feel, and the 990 Pro is the best we’ve yet tested here.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The 2TB Samsung 990 Pro writes at up to 6.2 GBps for 39 seconds, suggesting a pSLC cache size of about 240GB. Officially, the 990 Pro utilizes TurboWrite 2.0 - like the 980 Pro - which at 2TB is rated for 226GB. Of this, 10GB is static while 216GB is dynamic. The drive may engage in some folding in the background which can create an effectively larger cache. This drive may have different caching characteristics when run in PCIe 3.0 mode. Hybrid caches are becoming more common and that’s generally a good thing as such caches offer a good balance between performance and consistency.

After the pSLC cache runs out, the 990 Pro maintains around 1.4 GBps in a direct-to-TLC mode. This is slower than expected and, indeed, the rest of the drives in the test will all eventually out-write it. Post-pSLC performance may not seem super relevant in practice, but you are more likely to hit this state when the drive is fuller.

Samsung’s rated write performance of this flash has it faster than any other on the market, but there are reasons to limit write speeds. The most obvious reason is to provide better performance consistency, that is sustained performance, particularly with a flexible pSLC cache. Lower writes speeds may also induce less wear on the flash. 

Other reasons include thermal and power management. The former can improve lifespan and performance consistency by avoiding throttling. The latter can mean better drive efficiency, particularly in laptops and consoles. Samsung has emphasized improved thermal and power management with this drive, so this makes sense. Additionally, the DirectStorage API may hit drives harder, so with faster drives we are seeing more effort being put into power and thermal controls.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

We also monitor the drive’s temperature via the S.M.A.R.T. data and an IR thermometer to see when (or if) thermal throttling kicks in and how it impacts performance. Remember that results will vary based on the workload and ambient air temperature.

The 990 Pro is quite power-efficient in both power modes, easily besting the older 980 Pro but not setting any new records. It falls somewhere in-between the SN850X and Platinum P41, which is not a bad thing. It’s possible, if not likely, that this drive would show strong efficiency in more nuanced testing. This flash is significantly more efficient than what Samsung delivered on the 980 Pro and the improved low-power states on the 990 Pro should allow it to generally do better in laptops. Enhanced thermal management should also help it maintain efficiency with sustained workloads.

The Full Power Mode is a bit less efficient, but the distance between it and the normal mode would probably be much larger in some cases. For example, laptops tend to be much better at utilizing idle power states, so ignoring those would likely ramp up battery usage quite a bit. This also means that if you’re plugged in with a workstation-class laptop that the Full Power Mode could be beneficial. Of course, most drives ramp up power modes with heavy workloads anyway, so this may be more about response time as it corresponds to power state enter and exit latencies.

While it is possible to see that the 990 Pro pulls less power during idle than with the Full Power Mode in our test, it is important to keep in mind that this test is designed for typical desktop use. These drives as a whole will pull less power with lower idle states on laptops. Thus, the Full Power Mode differential will be more significant in that case.

We also tested temperatures to see how effective Samsung’s thermal optimizations work in practice. The drive idled around 40C - 36C by sensor, 45C by IR gun - which is well within reason. Over 1.6TB of writes saw the drive reaching around 72C which is comfortably below the first throttling point. This is a significantly lower load temperature than with the bare SN850X and the Platinum P41. Our 2TB 990 Pro sample did not come with a heatsink and it is likely that a heatsink would lower temperatures considerably.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless noted, we use active cooling for all SSDs.

Conclusion

The Samsung 990 Pro is an exceptionally powerful PCIe 4.0 SSD. It did not disappoint, giving us some record-breaking results and performing well where it mattered. Sustained write performance seemed a bit relaxed, but the overall consistency of the drive makes it worthwhile. Samsung updated its software in time for the release of this drive and that includes customizable options for power modes. This drive proved to be efficient and thermally-sound, but it is possible to push it for higher performance if desired.

This drive is a fine successor to the 980 Pro and, perhaps more importantly, a strong rival for the Platinum P41 and SN850X. Those two drives shone brightly earlier this year by pulling away from the ever-growing pack of high-end PCIe 4.0 drives. In particular, the Platinum P41 impressed with its efficiency and record-setting performance results. The SN850X got pretty close and is now often priced very competitively, but it also has the option for a heatsink that the Platinum P41 lacked.

Samsung focused on power efficiency and thermal management with the 990 Pro so that its high performance would work well in a laptop or a PS5, in addition to a high-end desktop. This puts it somewhere between the other two drives in terms of efficiency but with the potential for the highest performance we’ve seen to date. It’s likely this drive would be even more efficient inside a laptop with proper idle states. It also makes for a very compelling high-end gaming drive given its High Power Mode and other features that should work well with DirectStorage.

While many lamented the lack of a heatsink option on the Platinum P41, Samsung took from WD’s playbook and offers it along with RGB, as with the SN850X. This supposedly applies even to the 4TB SKU which won’t arrive until next year. The 4TB SN850X lacks this option, however it is actually available and has been as low as $399.99 at the time of review. That means Samsung still has a hole at that capacity for at least a few more months, although it can ably rival the Platinum P41 at 1TB and 2TB.

The top criticism we had for the SN850X was its pricing: launch MSRP prices were far too high given the other options available on the market, especially given the downward-trending pricing in the NAND and SSD industries. Samsung’s MSRPs for the 990 Pro are likewise too high. WD quickly reduced its prices well below MSRP, more so than did SK hynix with the Platinum P41, making the SN850X a very compelling option. Samsung will likely sell this below MSRP and it really has to in order to effectively compete. It is certainly capable of carrying at least a bit of a premium, though, as it appears to be the all-around champion at this point in time.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Acer Predator GM3500 is your run-of-the-mill PCIe 3.0 SSD. It utilizes older hardware and does not attempt to do anything new. Its limited capacity range makes it ideal for a cheap build or upgrade where you don’t need anything fancy. It has enough performance to get the job done, but is not going to be as fast or as efficient as newer drives. Acer backs it with a fair warranty and some software, which makes it a good option if the price is right. But otherwise, there is nothing too exciting here.

The “gold” standard for PCIe 3.0 SSDs remains the SK hynix Gold P31, with the possible exception of the Samsung 970 Evo Plus. The Evo Plus has seen some hardware shifting in the last year or so and has also had some steep sales that have put it back into contention. It still feels cumbersome when compared to efficient, four-channel drives like the Gold P31 and less-powerful options like the Sabrent Rocket or this GM3500, drives that often run cooler. Of more immediate impact is the availability of newer, DRAM-less drives, particularly those that are PCIe 4.0-capable, which have largely made DRAM-equipped PCIe 3.0 SSDs feel obsolete.

Of course, the allure of PCIe 3.0 SSDs like the Gold P31 is the fact they are often a bit cheaper. This makes drives like the GM3500 a nice alternative when on sale, if you just want something to upgrade an old machine or fill out a cheaper build. It’s mature, proven technology that is “fast enough” for most of what matters. The GM3500 doesn’t take any risks, which can be a good thing. You don’t always need the very newest technology, especially when it comes to NVMe storage. SATA options exist, of course, but innovation there has stagnated. And those who expect the best possible reliability should look to technologies that are still actively being developed and improved upon.

Specifications

The Acer GM3500 is available in only two capacities: 512GB and 1TB. While this may seem restrictive, it’s sufficient if you just want a basic drive for your OS. This is a PCIe 3.0 drive and, as such, is limited to around 3.4/3.0GBps for sequential read and write speeds. IOPS are lackluster, as expected from the use of older hardware. However, Acer backs it with a 5-year warranty and 600TB written per TB of capacity.

The real issue for this drive is that pricing at the time of review places it very close to mid-range PCIe 4.0 drives. Those choices, such as with the WD SN770, HP FX900 and Silicon Power UD90, are DRAM-less. Larger capacities are offered by drives like the Crucial P3 and P3 Plus, also DRAM-less but with QLC. This puts the GM3500 in a bit of an awkward position as all of these drives use newer hardware that easily compensates for the lack of DRAM.

Software and Accessories

Acer offers an OEM version of Acronis True Image with this drive. This is convenient and nice to have for cloning, something people often do during a storage upgrade.

A Closer Look

This is a basic drive without any fancy cooling or styling. Under the top label we see two NAND packages, the controller, and one DRAM package. This layout is a bit different than usual, especially as it is single-sided.

This controller, as used in the past, had a different physical configuration with regard to its components and was typically double-sided even at lower capacities. The GM3500 therefore may be easier to cool and would work better in some machines, such as laptops restricted to single-sided drives.

The controller is an SMI SM2262EN, an outdated but effective design. In some respects this was improved on with the SM2267, as used in the Adata S50 Lite, but that controller never saw wide use. The SM2262EN and its earlier SM2262 precursor, along with Phison’s E12 (later E12S) controller, were important in opening up the consumer NVMe market. The SM2262EN is now not as efficient or as powerful as newer controller designs, but offers a mature platform for someone who does not need the very fastest tech.

The DRAM is 1GB of DDR4, which is a nice pairing for 1TB of storage.

The flash is labeled BW29F4T08EMLCE. This is managed by Biwin, who traditionally has worked with HP for its SSD products. The rest of the code, in order, tells us that this Micron NAND is arranged in 4Tb or 512GB modules, with an 8-bit configuration. The last part is similar to but not exactly the same as the flash we saw on the HP FX900 Pro. This indicates a different generation, specifically two when comparing “C” to “E” - 96-layer here versus 176-layer there, although Micron’s 128-layer generation was never super common.

Micron actually had more than one configuration of this 96-layer TLC, but the most common has been the later-arriving B27B. Some early drives with the SM2262EN controller had B27A instead or, more typically, 64-layer B17A. B27B has a few changes over B27A, such as a faster I/O speed and a smaller block size. This is not massively relevant, but does suggest some improvements to endurance and performance with the flash the GM3500 is using.

There are also drives with this controller using older 64-layer TLC, as made famous with the Adata SX8200 Pro hardware swap. It’s more likely that those drives are using 96-layer or newer now, as 64-layer stock is depleted.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

Comparison Products

The GM3500 may seem outdated, but let’s see how it stacks up against some drives that remain popular. High-end PCIe 3.0 drives with DRAM include the Crucial P5, the SK hynix Gold P31 and the Sabrent Rocket. Two popular DRAM-less drives here are the Samsung 980 and the WD Blue SN570. The Inland Prime is a bit of the odd one out as it’s newer with a relatively rare controller and fast flash. Rounding out the mix is the HP FX900, a mid-range and DRAM-less PCIe 4.0 drive that represents an increasingly popular market segment.

Trace Testing - 3DMark Storage Benchmark

3DMark’s Storage Benchmark focuses on real-world gaming performance. Each round in this benchmark stresses storage based on gaming activities including loading games, saving progress, installing game files and recording gameplay video streams.

The GM3500 performs a bit above average in 3DMark, which is a fair showing. The newer, DRAM-less offerings - the 980 and FX900 - do better, as does the popular Gold P31.

Trace Testing – PCMark 10 Storage Benchmark

PCMark 10 is a trace-based benchmark that uses a wide-ranging set of real-world traces from popular applications and everyday tasks to measure the performance of storage devices.

The GM3500 does a bit better here than in 3DMark, managing to best the Gold P31. The SM2262EN controller has always done well with consumer workloads, particularly with the updated B27B flash.

Transfer Rates – DiskBench

We use the DiskBench storage benchmarking tool to test file transfer performance with a custom, 50GB dataset. We copy 31,227 files of various types, such as pictures, PDFs, and videos to a new folder and then follow-up with a reading test of a newly-written 6.5GB zip file.

The Predator GM3500 performs exactly as expected, close to the Rocket and P5. The PCIe 4.0 FX900 easily pulls away from the rest of the pack. The newer flash in the 980 and Gold P31 drives may help them do a bit better here.

Synthetic Testing - CrystalDiskMark

CrystalDiskMark (CDM) is a free and easy-to-use storage benchmarking tool that SSD vendors commonly use to assign performance specifications to their products. This tool gives us insight into how each device handles different file sizes.

CDM’s sequential performance tests are largely limited by a drive’s interface, so it's no surprise the FX900 easily comes out on top. The GM3500 does not stand out in any particular way. With random performance we can see why the Gold P31 is so popular and also how newer drives - those with faster controllers and flash - can really dominate in these traditional metrics. The GM3500 remains unexceptional, but its performance is still “good enough” when compared to other PCIe 3.0 SSDs.

Sustained Write Performance and Cache Recovery

Official write specifications are only part of the performance picture. Most SSDs implement a write cache, which is a fast area of (usually) pseudo-SLC programmed flash that absorbs incoming data.  Sustained write speeds can suffer tremendously once the workload spills outside of the cache and into the "native" TLC or QLC flash. We use Iometer to hammer the SSD with sequential writes for 15 minutes to measure both the size of the write cache and performance after the cache is saturated. We also monitor cache recovery via multiple idle rounds.

The GM3500 writes in three common modes: pSLC, direct-to-TLC, and folding. The drive writes at around 2.8GBps for almost 60 seconds for a pSLC cache around 160GB. After that, TLC writes are around 700MBps, a bit slower than what is possible which is evident by looking at the Sabrent Rocket. The GM3500 could do better at 2TB if such a capacity option existed. B27B has twice the density of the original 64-layer flas used with this controller at this capacity; fewer dies means potentially slower speeds outside of pSLC.

This is also reflected in folding performance, which in this case is around 400MBps. It takes a good while to hit this state and the performance there suggests the drive does a good job of shuffling data from pSLC to TLC in the background in the earlier write states. The write response is therefore fairly consistent even though it is not otherwise impressive. This is perfectly suitable for a cheap, primary drive at 1TB. For sustained writes, as one might want with caching on NAS, the Sabrent Rocket or Gold P31 would be more sensible options.

Power Consumption and Temperature

We use the Quarch HD Programmable Power Module to gain a deeper understanding of power characteristics. Idle power consumption is an important aspect to consider, especially if you're looking for a laptop upgrade as even the best ultrabooks can have mediocre storage.

Some SSDs can consume watts of power at idle while better-suited ones sip just milliwatts. Average workload power consumption and max consumption are two other aspects of power consumption, but performance-per-watt is more important. A drive might consume more power during any given workload, but accomplishing a task faster allows the drive to drop into an idle state more quickly, ultimately saving energy.

We also monitor the drive’s temperature via the S.M.A.R.T. data and an IR thermometer to see when (or if) thermal throttling kicks in and how it impacts performance. Remember that results will vary based on the workload and ambient air temperature.

Another reason users may jump up to newer drives, particularly PCIe 4.0 SSDs, is that they can prove to be more efficient. SSD controllers manufactured in smaller process nodes with newer flash - more layers often means power savings, including from the general trend towards putting the CMOS under array - means higher horsepower and less power draw for a given level of performance. Power efficiency therefore goes up.

This is not necessarily a huge consideration even for laptops, as a properly functioning NVMe drive with adequate PC power states should mostly sit at idle with minimal draw. Consumer workloads aren’t particularly heavy. Of course, this is changing with the advent of DirectStorage, and some power savings are always welcome to mitigate heat generation. Electricity usage is also becoming more important over time.

The GM3500 can’t hope to touch the newer drives or the famously efficient Gold P31 when it comes to power efficiency. However, it does beat some of the other PCIe 3.0 drives, even the budget-oriented SN570. Still, we would generally recommend something newer for laptops and HTPCs if you can afford it.

We measured the GM3500 at 37C when idle. Extensive sustained writes got it up to a maximum of 81C, which is just shy of its first throttling point. This drive probably does not require a heatsink like many high-end PCIe 4.0 SSDs, but is also not the coolest-running drive within the budget segment.

Test Bench and Testing Notes

We use an Alder Lake platform with most background applications such as indexing, Windows updates, and anti-virus disabled in the OS to reduce run-to-run variability. Each SSD is prefilled to 50% capacity and tested as a secondary device. Unless otherwise noted, we use active cooling for all SSDs.

Conclusion

The Acer Predator GM3500 is a basic drive, utilizing old hardware to rival an already over-saturated market. PCIe 3.0 SSDs have not seen a lot of innovation lately, aside from perhaps the SK hynix Gold P31. But the argument could be made that not a lot of progress is necessary. The GM3500 gets the job done, minimally and with few frills. Its performance is good enough for most use cases. The warranty and manufacturer support are sufficient, and you know what to expect from this hardware.

We would still recommend something like the Gold P31 for laptops and HTPC usage for both energy efficiency and potentially reduced heat generation. The GM3500 was not horribly inefficient or hot. In fact, the single-sided design makes it competitive with the Sabrent Rocket and similar drives, but there may be better options available if those factors matter to you. This drive’s best place would be the same as with the Sabrent Rocket: in a basic build or for a simple NVMe upgrade. It’s predictable and consistent. It just has to be priced right for it to be compelling.

The flash is from Biwin,  a newer partner for Acer, which previously made similar flash for drives like the HP EX950. And so hopefully we won't see the same swap issues as the Adata SX8200 Pro. The controller is mature. Even the pSLC caching scheme is the standard fare. No huge capacities or massive heatsink - just a basic drive, and that’s okay, but at its current price point, there are better choices.

Chat said, considering the Gold P31 has had availability and/or pricing issues in many places, the GM3500 could be a good choice for laptops and HTPCs--partularly if and when its price drops.

MORE: Best SSDs

MORE: Best External SSDs and Hard Drives

MORE: How We Test HDDs And SSDs

MORE: All SSD Content

The Intel Arc A770 graphics card has finally arrived, along with its little brother, the Intel Arc A750. After a rather disappointing Arc A380 review last month, Intel has a lot to prove with the bigger and far more potent A770. And it mostly succeeds! While there are certainly caveats — mostly about drivers, XeSS adoption, and long-term support — Intel clearly wants to prove it can compete with the likes of AMD and Nvidia, perhaps even laying claim to a seat at the table among the best graphics cards.

It's been a long road to Intel re-entering the dedicated graphics card market. It first attempted to make a dedicated graphics card with the i740 back in the late 90s — that would be "Gen1 Graphics" if you're wondering — before giving up. Larabee was another attempt at a potential GPU, but it got mired in internal politics. Now, despite claims of Arc being dead in the water, we have cards in hand and can finally see what Intel's best-ever GPU brings to the table.

How does Arc A770 stack up against the AMD and Nvidia competition? Is XeSS a true DLSS competitor, and what about all the AV1 video hype? Could Intel fix its drivers to the point where it won't be one of the first things people warn you about? We'll look to answer all these questions and more, but let's start with the high-level overview.

Arc Alchemist Architecture Recap

We've had plenty of details on Intel's Arc Alchemist architecture, which we've known about since last year. We now have final clocks and specs, as well as pricing, but it's worth revisiting the road Intel has traveled. The first time we heard about Arc GPUs, we anticipated a late 2021 or early 2022 launch. By that metric, Intel is nearly a year late to the party — blame Covid, supply chain issues, and even the Russian invasion of Ukraine if you're looking for reasons.

On one level, Arc Alchemist would be the 13th iteration of Intel Graphics. At the same time, we'd also mark this down as generation one. There are enough major overhauls of the fundamental design, features, and functionality that a card like the A770 has very little in common with Intel's current 12th-Gen integrated Xe Graphics. There are several new additions to the GPU that provide a clear demarcation between the pre-Arc and the post-Arc world of Intel GPUs.

Starting with the main building block, the Xe-Core, Intel is apparently waffling a bit on naming conventions. The "Xe Vector Engines (XVE)" are still referred to as "Execution Units (EUs)" at times — a rose by any other name would be as fast, I suppose. Intel groups 16 XVEs into a single Xe-Core, which also includes other functionality like the Xe Matrix Engines (XMX) blocks, Load/Store unit, and L1 cache. Attached to each Xe-Core (but not directly a part of it) are the Thread Sorting Unit (TSU), BVH Cache (for ray tracing), and the other ray tracing hardware for BVH ray/box intersection and ray/triangle intersection.

Each XVE can compute eight FP32 operations per cycle. That gets loosely translated into "GPU cores," though we prefer to call them "GPU shaders," and each is roughly (very roughly!) equivalent to an AMD or Nvidia shader. Each Xe-Core thus has 128 shader cores and sort of maps to an (upcoming) AMD RDNA 3 Compute Unit (CU) or an Nvidia Streaming Multiprocessor (SM) — both of which will also have 128 GPU shaders. They're all SIMD (single instruction multiple data) designs, and Arc Alchemist has enhanced the shaders to meet the full DirectX 12 Ultimate feature set.

The XVEs also have additional functionality, like INT and EM support. There are a lot of memory address calculations in graphics workloads, which is where the INT functionality often helps, though such things can also be used for cryptographic hashing. EM stands for "Extended Math" and provides access to more complex stuff like transcendental functions — log, exponent, sine, cosine, etc. The INT execution port is shared with the EM functionality, which tends to be used less frequently than INT and FP calculations.

Meanwhile, the XMX blocks are comparable to Nvidia's Tensor cores. Each XMX unit can handle either FP16/BF16 (16-bit floating point/brain floating point), INT8 (8-bit integer), or INT4/INT2 (4-bit/2-bit integer) data. These blocks are useful with deep learning workloads, including Intel's XeSS (Xe Super Sampling) upscaling algorithm. They can be used in any workload that just needs a lot of lower-precision number crunching, and each XMX block can do either 128 FP16, 256 INT8, or 512 INT4/INT2 operations per clock.

The processor can co-issue instructions to all three execution ports — FP, INT/EM, and XMX — at the same time, and all three execution blocks can be active at the same time. Again, that's loosely similar to Nvidia's architectures, though it's important to note that Nvidia has a shared INT/FP port, so if the INT aspect is active, it cuts the potential FP throughput in half.

Stepping up one level, Intel has what it calls a render slice, which is sort of analogous to Nvidia's Graphics Processing Cluster (GPC). The render slice consists of four Xe-Cores, and then adds texture units ("Sampler" in the above image) and Render Outputs (ROPs, the "Pixel Backend" in the image), plus some other hardware.

Intel has two main designs with Arc Alchemist, one with two render slices and up to eight Xe-Cores, and the other is a larger design with up to eight render slices. The Arc A770 represents the fully enabled larger design and thus has the full 32 Xe-Cores, while the Arc A380 uses the fully enabled smaller design — with other variants (including mobile models) using partially enabled chips.

Here's the full block diagram for the larger ACM-G10 die, and the A770. Along with all the render slices and other functionality, there's a large chunk of L2 cache and the memory fabric linking all of the slices together, up to 16MB in the A770. That might not seem that large, considering Nvidia's Ada Lovelace and the new AD102 has up to 96MB of L2 cache, but keep in mind the different performance and price levels, plus the fact that Nvidia's soon-to-be-previous-gen GA102 chip (used in the RTX 3090 Ti and other cards) only had up to 6MB of L2 cache.

The block diagram also shows the Xe Media Engine, which handles various video/image codecs, including AV1, VP9, HEVC, and H.264. There are two full media engines (MFX) in both the larger and smaller Arc GPUs. These can work on separate streams or combine their computational power to double their encoding throughput.

On the left of the block diagram — and it's important to note that this is not a representation of the actual chip layout — Intel shows the four display engine blocks, PCI Express link, Copy Engine, and Memory Controllers. The GDDR6 controllers are actually positioned around the outside of the GPU, linking up to the external memory, and the ACM-G10 has up to eight 32-bit memory controllers.

It's not clear if Intel can disable those individually on the larger GPU or if it needs to be done in pairs, though we do know on the smaller ACM-G11 that the three 32-bit controllers can be individually disabled. Meanwhile, the mobile A730M has a 192-bit interface with two of the controllers fused off, while the mobile A550M has a 128-bit interface with four controllers disabled.

Intel Arc A770 Specifications

With that overview of the architecture out of the way, here are the full specifications for the Arc A770, A750, (upcoming) A580, and A380. Performance potential here is theoretical teraflops / teraops (trillions of operations per second), but remember that not all teraflops and teraops are created equal. We need real-world testing to see what sort of actual performance the architecture can deliver, but we'll get to that soon enough.

Intel's Arc A770 looks quite potent on paper, and clearly, Intel is gunning for its share of the midrange market with aggressive pricing on the A750 and A770. Both end up competing against the Nvidia RTX 3060, with the A770 coming at a similar theoretical price, while the A750 clearly undercuts the 3060 — though Nvidia's GPU still tends to sell for closer to $370 at retail. On the AMD side of the fence, things are a bit different. The RX 6700 XT potentially delivers more performance at a higher $400 price point, while the RX 6650 XT still sells for around $300 (give or take).

That's where GPU prices stand right now, at least — and we expect everything to continue to fall in the coming months, though probably not too much further before parts start getting discontinued and replaced by newer models.

In terms of specs and features, Intel's Arc GPUs end up looking a lot more like Nvidia's GPUs than AMD's offerings, with the extra matrix cores and a much bigger emphasis on ray tracing hardware. Intel is also the only GPU company that currently has AV1 and VP9 hardware accelerated video encoding. AMD and Nvidia will add AV1 support to their upcoming RDNA 3 and Ada architectures, but those may not come to the midrange market any time soon. If you're interested in AV1 and don't want to spend more than $350, the Arc A770, A750, and A580 might be your best and only options this side of 2023.

The three midrange Arc cards deliver theoretical compute performance of 10.4 to 17.2 teraflops of FP32. (Note that Intel uses "typical" game clocks, though, in our testing, we've seen a lot of cases where the clock speed is far higher than the nominal 2.1 GHz listed above.) By comparison, Nvidia's RTX 3060 offers 12.7 teraflops of graphics compute, while AMD's RX 6650 XT sits at 10.8 teraflops. However, we're definitely getting into the realm of apples and oranges, as — spoiler alert! — the RX 6650 XT tends to lead the RTX 3060 by 10–20 percent in traditional gaming benchmarks. Ray tracing games flip the tables, however, with the 3060 beating the 6650 XT by around 25 percent.

Arc's ray tracing capabilities have been a bit difficult to pin down up to now. The A380 did deliver better RT performance than the RX 6500 XT, but that's hardly praiseworthy. With four times the cores and hardware, we're expecting a lot more from the A770 and A750 — and Intel has even shown benchmarks where the A770 clearly beat the RTX 3060 with ray tracing enabled.

That's a nice change of pace from AMD, which to date hasn't done much with ray tracing and tends to downplay its importance. And to be fair, AMD has a point: the visual fidelity gains that come from enabling ray tracing are often far outweighed by the loss of performance. Especially on AMD's GPUs.

Intel's RTUs have dedicated BVH traversal hardware and a BVH cache that reduces memory bottlenecks, and it says each RTU can do up to 12 ray/box BVH intersections per clock, along with one ray/triangle intersection. By comparison, AMD's RDNA 2 GPUs only do up to four ray/box intersections per clock, and that's done using enhanced texture units, which means the hardware gets shared and ends up with some resource contention.

Nvidia's RT cores are more like a black box that Nvidia doesn't want to discuss in too much detail. Turing could do one ray/triangle intersection per clock, and an indeterminate number of ray/box intersections. Nvidia apparently determined the BVH side of things was running ahead of the ray/triangle hardware, so in Ampere, it added a second ray/triangle intersection unit per RT core. Looking forward, Ada doubles the ray/triangle throughput and includes other BVH enhancements, and it seems unlikely that any contemporary GPUs will match Ada in ray tracing performance. But Intel does seem to be at least relatively competitive with Ampere in terms of RT capabilities.

The A770 maxes out at 32 RTUs, which is more than the RTX 3060's 28 RT cores but less than the RTX 3060 Ti's 38 RT cores. Meanwhile, AMD's RX 6650 XT has 32 ray accelerators, while the RX 6700 XT has 40 ray accelerators… but again, AMD's ray tracing hardware appears to be the weakest of the three GPU vendors. Check our ray tracing performance results a few pages on to see where the chips fall in real-world testing.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

Intel might be new to the graphics card market. Still, it's had motherboards, NUCs, and numerous other products over the years and obviously has plenty of expertise when it comes to manufacturing. That's definitely apparent when you look at the Arc A770 Limited Edition and consider this is the first consumer graphics card Intel has shipped in over two decades.

Intel will sell cards under its own label under the "Limited Edition" brand. However, think of them as vehicles with an "LE" suffix because these won't be limited production runs. As long as Intel continues to sell Arc GPUs, it intends to sell Limited Edition cards. The A770 Limited Edition represents the current halo product from the Arc family, with the fully enabled ACM-G10 die along with 16GB of 17.5 Gbps GDDR6 memory. There will also be partner cards, but those may opt for either 16GB like Intel's card, or 8GB — with a presumably lower price. Frankly, considering Intel only recommends a $20 difference in pricing between the 8GB and 16GB A770 cards, we see little reason to bother with an 8GB model.

The A770 LE has two custom 85mm fans with 15 blades each. The card weighs 1080g and measures 268x110x38mm, a traditional 2-slot design. Compared to some of the other monster cards we've seen in the past generation, the A770 LE is relatively compact, though the competing RTX 3060 and RX 6650 XT are similarly sized and use less power. We'll get to power and noise measurements later, but the fans and cooler do appear to do a good job.

The A770 LE comes with a decent amount of RGB lighting, including a USB cable that connects to a motherboard port that allows control of the lighting effects. In addition, there are glowing rings around the fans, a lighting strip that extends around the outside edge of the card, and an additional lighting strip on the side opposite the fans. The default color scheme is blue, with purple/pink segments lit up that move around the card.

Intel's card comes with 8-pin and 6-pin PEG power connectors. Note that on the review samples, these are colored gray and black (or at least darker gray), but the retail products will have two black connectors. The card runs with a 225W TBP (Total Board Power) at stock, with a decent amount of headroom for overclocking should you want to boost the power limit.

Internally, the heatsink uses a large vapor chamber to cool the GPU and memory, which is a higher-end cooling solution and not something we normally see on midrange and lower cards. Even though the card isn't particularly large and has a higher power draw than competing GPUs, the overall results during testing have been good.

Video outputs consist of the typical triple DisplayPort and a single HDMI that's become almost universal among modern GPUs. Note that this is the first GPU with DisplayPort 2.0 support, sort of, which should allow for higher refresh rates and color depths over a single cable. Meanwhile, the HDMI port is 2.0b, though a DisplayPort to HDMI 2.1 adapter may be provided with some cards.

It's important to note that Intel classifies the DisplayPort as "1.4a, DP 2.0 10G Ready" — likely due to the lack of DP 2.0 displays. Intel Arc can support up to UHBR 10 data rates, but not UHBR 13.5 or UHBR 20. That makes the DP 2.0 10G a modest 24% improvement in bandwidth over DP1.4a. However, UHBR 10 and above switch from 8b/10b encoding to 128b/130b encoding, which increases the maximum data rate from DP1.4a's 25.92 Gb/s up to 39.38 Gb/s — a 52% increase in usable bandwidth.

The important thing (for some) will be that Arc can support up to 8K 60 Hz resolutions, using DSC (Display Stream Compression) — 4:2:2 (maybe) or 4:2:0 (for sure) encoding. The display engines can handle up to two 8K 60 Hz displays, or four 4K 120 Hz displays. 1440p at up to 360 Hz is also supported.

Overall, the Arc A770 Limited Edition is an attractive design that should please anyone who's disgruntled with the overly gaudy and bulky graphics cards that have become so common. And if you don't like Intel's particular design, Arc A700-series cards from ASRock and Gunnir are apparently on the way, and Acer's first graphics card ever will be an Arc A770 with a unique radial blower plus axial fan design. 

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

Our 2022 GPU testbed consists of a Core i9-12900K processor, MSI Pro Z690-A DDR4 WiFi motherboard, and DDR4-3600 memory (with XMP enabled). We're running Windows 11 with the 22H2 update to ensure we get the most out of Alder Lake. We're also running the latest game patches and Intel drivers at the time of testing: Intel 30.0.101.3435.

In case you're wondering, Intel has released a bunch of different drivers for Arc GPUs since late March — seven non-beta drivers and another seven beta releases. So there were at least six different drivers released in the past two months, and we probably missed some. The 3433 and 3435 drivers are pre-release drivers for the A700-series launch, and we expect updated versions that continue to fix problems will be made publicly available next week.

What sort of problems? We have a modest test suite that's public knowledge. It consists of 14 widely known and benchmarked games — just the kind of stuff Intel should be optimizing around for a new GPU launch. Still, we ran into a few problems and idiosyncrasies during testing. For example, most of our standard test suite worked fine, but Red Dead Redemption 2 would run in a window when selecting full screen. The workaround was to change the desktop resolution to whatever resolution we wanted to use in the game, at which point things worked more or less properly (and performance improved).

Our ray tracing test suite didn't fare quite so well. First, the Bright Memory Infinite Benchmark seemed to experience a memory leak that caused degraded performance over time. About midway through the first run, performance dropped on the A750, and subsequent runs got worse until, eventually, the only real option was to quit and restart the game. Fortnite had a bug with the 3433 drivers that prevented it from launching; 3435 fixed the situation. Finally, Minecraft Bedrock (the Windows Store version) still hasn't whitelisted the Arc GPUs to allow the use of ray tracing, which means we can't run that particular benchmark right now.

I also experienced a few other glitches. The sound cut out during testing a few times (the sound was playing through my monitor speakers via DisplayPort), and the only way to get it back was to restart the PC. Actually, that might not be correct, as I didn't try restarting the Windows sound service, but I couldn't determine the exact sequence that caused the problem, which made replicating it difficult.

Finally, as with the Arc A380, my older MSI MEG Z390 Ace WiFi motherboard running the latest BIOS (with Resizable BAR enabled) still won't POST without some other graphics card installed. I put in a GTX 1650 Super and it booted right up, so there's still some sort of VBIOS or firmware problem. Other Z390 boards may work fine, or they might not, but every other graphics card I have works without any complaint in the PC.

For this review, we've tested at 1080p with medium settings, and 1080p, 1440p, and 4K using ultra settings on our standard test suite. For our ray tracing tests, we dropped 4K ultra from the list, as 1440p (at least without XeSS) was already struggling. We've tested the A770 and A750 Limited Edition cards and will show the results here, though the discussion of the A750 will be primarily in our separate Intel Arc A750 review.

Intel Arc A770 Overclocking

The only support for overclocking on Arc A770 right now comes from Intel's own Arc Control utility. Due to time constraints (Nvidia's RTX 4090 is also launching next week, in case you've been living under a rock), we've omitted overclocking tests for this initial review. We feel most will just run stock clocks for a midrange card, so that's the best baseline for performance. Once time permits, we'll revisit the Intel cards and see how far we can push things with overclocking.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

We're going to go in order of the most meaningful settings (in our opinion) for the Arc A770. That means 1080p and 1440p ultra first, then 1080p medium, and finally 4K ultra. Not all cards were tested at all resolutions, particularly at 4K. As you'd expect, 1080p means higher fps, often more than a lot of monitors need. 1440p represents more of a sweet spot, while 4K will generally need more GPU power than the A770 can muster, at least in more demanding games.

Also, note that we're focusing solely on the A770 for this review, so while the A750 numbers are present, we won't comment on them. Head over to that review for the A750 analysis.

Starting with 1080p and "ultra" settings, the Arc A770 puts up some good results. It averaged 85 fps across our test suite and outperformed the RTX 3060 by 13%. That puts it on equal footing with the RX 6650 XT (2% behind, if you want to be exact). That's a bit disappointing, as the A770 LE comes with twice as much VRAM, though 8GB of memory doesn't usually hold back performance at 1080p.

Flipping through the individual game results, things become a bit more chaotic. Against the 3060, there's only one game in our test suite where the Arc A770 doesn't claim the lead: Horizon Zero Dawn. Interestingly, that's a DX12 game, but Intel apparently hasn't fully optimized for it yet. Elsewhere, the A770 leads by anywhere from 4% (Flight Simulator) to as much as 35% (Red Dead Redemption 2). Finally, the A770 and 6650 XT split the games 50-50, with Arc trailing by as much as 19% (Horizon Zero Dawn again) and leading by as much as 28% (Red Dead Redemption 2 again).

If we were to extend our test suite to dozens more games, perhaps with more older titles that don't use DX12 or Vulkan, we expect the overall A770 performance would skew downward. At the same time, assuming Intel continues to put a serious effort into improving and optimizing its drivers, we could see some impressive performance increases over time. Take the best results (Borderlands 3, Forza Horizon 5, Red Dead Redemption 2, and Watch Dogs Legion); if Intel can get similar relative performance in a wider test suite, overall performance could easily improve by 10–20 percent.

But if you're thinking about picking up an Arc GPU, it's important to know that right now, there are going to be situations where the drivers hold Arc back.

Stepping up to 1440p Ultra drops the A770's performance by 22% compared to 1080p ultra, while the competing AMD and Nvidia cards lose 29% and 25%, respectively. That drop is also consistent with the Arc A750, so the extra VRAM on the A770 LE isn't really coming into play. Overall, the A770 still averages more than 60 fps, and it's 18% faster than the RTX 3060 and 8% faster than the RX 6650 XT — but still 13% slower than the RX 6700 XT and 11% slower than the RTX 3060 Ti.

Looking at the individual games again, the A770 gets a clean sweep of the RTX 3060 now, with leads ranging from 6% to as much as 32%. It also basically ties or beats the RX 6650 XT across the suite, trailing by 1% slower in Far Cry 6 and leading by up to 26% (RDR2). Also, note that only Flight Simulator and Total War: Warhammer 3 failed to average 60 fps or more.

If you're primarily interested in 1440p gaming, especially if you're willing to drop a few settings from ultra to high, the Arc A770 LE ends up as the better option against its direct competitors — again, assuming you don't want to play a game where Intel hasn't yet optimized the drivers.

As you might expect, given what we've already seen, 1080p medium actually decreases the Arc A770's relative performance. Arc performance improves by 56% compared to 1080p ultra, but the RTX 3060 is 68% faster and the RX 6650 XT is 73% faster. Any of these cards easily handle 1080p medium, but if you have a 144Hz monitor, you might prefer one of the non-Arc cards — at least for the time being.

What's likely happening is that Arc's drivers are becoming more CPU limited at lighter settings, meaning they can't scale as much as the AMD and Nvidia GPUs. This is something Intel has already talked about, how using the integrated graphics drivers as the base for Arc hampered overall performance, and it's something the driver team continues to work on.

With integrated graphics, they're so slow that it's often beneficial to move work from the GPU to the CPU. The CPU spends so much time waiting on the GPU that it has room for other work. But that's no longer the case with a faster dedicated graphics card, so the drivers need to give as much work as possible back to the GPU so the CPU can generate frames faster.

4K ultra naturally represents the reverse of 1080p medium. The graphics workload is so high that the GPU becomes the major bottleneck. We didn't test the RTX 3060 or RX 6650 XT at 4K ultra, but if you look at the cards we did test, like the RTX 3070 and RX 6700 XT, you can see that Arc once again closes the gap a bit.

Where the RX 6700 XT was 15% faster than the A770 at 1440p ultra, that lead drops to 8% at 4K. (Of course, much of that likely stems from AMD's Infinity Cache being less effective at higher resolutions.) Meanwhile, RTX 3070 still holds a sizable 20% lead at 4K, but that's less than the 25% lead it held at 1440p and 1080p ultra. 

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

Ray tracing can be extremely demanding, and Nvidia GPUs have generally held the lead at similar price points compared to AMD's GPUs. With Arc, we now have a third player, and as we hinted earlier, it looks as though Intel put a lot more effort into the ray tracing aspects of the hardware.

If you're used to seeing Nvidia demolish AMD at ray tracing, Arc should be a nice breath of fresh air. The A770 LE soundly beats the RTX 3060 by 25% on average at 1080p medium, though Arc couldn't run Minecraft. That's apparently something Microsoft needs to fix, not a driver bug — the game looks for "known" GPUs and only allows you to enable DXR on certain models. That's… annoying, and Microsoft, of all companies, ought to know better. Query the hardware for its capabilities and then expose the functionality, rather than limiting support to known quantities! But I digress.

AMD's RX 6650 XT doesn't do nearly as well with our DXR test suite. The A770 is 50% faster at 1080p medium and is still 21% faster than the costlier RX 6700 XT. That's a nominal win for Intel and Arc over the competition… except we need to consider Nvidia's DLSS.

Intel offers a reasonable competitor to DLSS in its XeSS, but DLSS has been promoted for four years now. XeSS is just getting started. Right now, there are only a handful of games with XeSS support, and we can imagine Intel basically pleading and cajoling game developers into adding it. Theoretically, anything with DLSS 2 could be made to support XeSS (and FSR 2) with minimal effort, but getting that to happen is another matter.

Depending on the game, DLSS in Quality mode typically improves performance by 20–40 percent. Quality mode is nearly indistinguishable from native rendering and sometimes looks superior, so it's a great option. That would easily put Nvidia's RTX 3060 back in the lead, but if a game also supports XeSS, we'd be back to an Arc victory.

More or less maxing out the ray tracing settings at 1080p can cripple performance, particularly without DLSS, XeSS, or FSR 2 support. Comparing apples to apples, though, Intel's A770 LE continues to deliver impressive results. It clearly wasn't kidding when it said it put a lot of effort into the RTUs (ray tracing units) and related hardware.

At 1080p ultra, the A770 still leads the 3060 by 23%, beats the 6700 XT by 19%, and demolishes the RX 6650 XT by 50% again. The RTX 3060 Ti is still 9% faster, but that's also a more expensive card. The only real blemish on Arc is the currently buggy (memory leak?) Bright Memory Infinite result. Even in the very demanding Cyberpunk 2077, the A770 still manages to go toe to toe with Nvidia's 3060 — and let's not even look at the RTX 3050, please.

What about 1440p ultra with DXR? Well, that's a bit more than Arc A770 can manage, at least with native rendering. To be fair, the same is true of most other graphics cards that cost less than $500. Only the RTX 3070 and above can break 30 fps across our DXR test suite at 1440p, or AMD's RX 6800 XT. If you prefer putting it in older terms, the A770 LE basically ties the RTX 2080 Super at these settings. Not bad for about half the price and twice the VRAM, three years later.

Of the five games we tested, only Control and Metro Exodus Enhanced reached playable framerates of 30 fps or more. With XeSS Performance mode, you could probably double a lot of the numbers and that could make everything playable in a pinch, but at present, the confirmed XeSS titles are relatively few in number. Which brings us to our XeSS testing on the next page. 

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

Intel Arc A770 XeSS Performance and Quality

XeSS has been heralded as a more open alternative to Nvidia's DLSS. Both use machine learning to train an AI network that will then intelligently upscale and anti-alias lower resolution input frames into a higher resolution result. Both should improve over time as the network and algorithms get tweaked. In theory, it seems as though XeSS should be able to match DLSS 2 quality — but doing something more involved like DLSS 3 would be another matter.

Right now, we have a few publicly available games: Shadow of the Tomb Raider, Monster Hunter Rise, Death Stranding, and… I'm not sure which of the others are actually out. You can check Intel's XeSS page for further updates, but Intel also provided beta access to XeSS in Hitman III, The DioField Chronicle, Ghostwire Tokyo, and Arcadegeddon. Unfortunately, we didn't have time to test all of those, so we just did some testing with Shadow of the Tomb Raider and Death Stranding.

One thing you need to understand with XeSS is that it has support for both Intel's XMX cores, but for non-Arc GPUs, it can utilize DP4a instructions (basically 8-bit integers). Here's the kicker: The two algorithms are not the same! That makes sense, as the XMX cores provide far more computational power than any GPU using DP4a, so the quality will inevitably be lower with DP4a mode. And DP4a is what you'll get on any non-Arc GPU for now.

It's also important to note that even in XMX mode, on a significantly slower GPU like the Arc A380, XeSS may not show as much of a benefit. Instead, it's probably running into bottlenecks of some form, or at least that's what it looked like in our testing last week. Anyway, we'll have to wait and see how the adoption and quality of XeSS changes over time, but there's still AMD's FRS 2 (now 2.1), which tends to work much faster than XeSS DP4a mode while potentially providing equal or better image quality. That's okay since Arc can run FSR2 just fine, but it does make it more difficult to get developer buy-in for XeSS.

Due to time constraints, we've only tested the A770 and A750 cards in two games, using native, Quality, and Performance modes — at three resolutions. Quality looks relatively close to native, while Performance mode uses 4X upscaling and can definitely look worse than native (for now, anyway).

Quality mode, which likely uses 2X upscaling (if it's the same as DLSS and FSR2), improves performance by up to 28% in Death Stranding, but we're clearly hitting CPU bottlenecks as the gains are far lower at 1440p and non-existent at 1080p.

The gains are more pronounced for Shadow of the Tomb Raider, with maxed-out settings including ray-traced shadows. The A750 sees up to a 44% improvement at 4K but very little change at 1080p. Interestingly, the A770 improves by 51% at 4K and 18% at 1080p. There does appear to be some impact due to the reduced amount of VRAM on the A750, though it may also be driver related.

Performance mode helps even more, as you'd expect. In Death Stranding, 1080p remains CPU bottlenecked, but 4K performance improves by 53% while 1440p gains 20–26%. On the other hand, Tomb Raider performance basically doubles at 4K, increases by 45–55% at 1440p, and is 25–35% faster at 1080p.

Those are far bigger gains than when we tested Shadow of the Tomb Raider XeSS on a collection of GPUs that didn't include Arc A700-series parts. Again, DP4a might be okay in a pinch if no other upscaling algorithm is available, but it's not all that great for now.

Intel Arc A770 Video Encoding Performance and Quality

One of the bigger selling points for the Arc A770, and Intel's Arc Alchemist GPUs in general, is the support for hardware accelerated AV1 and VP9 video encoding. Current AMD and Nvidia GPUs only support hybrid decoding of VP9 and AV1, where some of the work is done by the CPU, and both codecs can prove quite taxing on lesser CPUs. Ada and RDNA 3 will add AV1 support, but those will likely start with high-end and extreme performance (and price) cards, not midrange options that cost less than $350.

We extensively tested the Arc A380 video encoding performance and quality, and things should be effectively unchanged with Arc A770 and A750 as the video engine is a fixed functional unit — all the extra Xe-Cores don't matter. If you're interested in learning more about Arc's video engine and encoding quality, we refer you to the previous article for now. The summary is that Arc GPUs could be great for streaming AV1 once support for that codec becomes more widely available. 

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

For our power, temperature, clock speed, and fan speed testing, we use Powenetics hardware and software. We capture in-line GPU power consumption by collecting data while looping Metro Exodus — at 1440p ultra with midrange and higher GPUs. We also test with the FurMark stress test at 1600x900. Our power testing PC uses an open testbed, as that's required for all the extra wires and riser card, and it's the same Core i9-9900K that we've used for the past several years.

As noted earlier, we had to install a second graphics card in the PC to get the A770 to POST, but we disabled the card in Device Manager and our equipment doesn't include the power use from the other PCIe slot. GPU-Z meanwhile monitors the fan speeds, clock speeds, and other aspects of the cards, though it doesn't capture fan speed as a percentage. The GPU core clocks and fan speeds also seen to be getting slightly erratic readings — or perhaps they just expose far more granular data that causes some jitter in the charts.

For the A770 Limited Edition, average power use while running Metro Exodus was 217W, though peak power spiked as high as 253W. The 8-pin connector maxed out at 146W, the 6-pin at 63W, and the PCIe slot at 48W, so all of the connectors are within spec. While the GPU was under load during the Metro benchmark, power use ranged between 190W and 230W.

As usual, FurMark generated a far more consistent load, and the A770 LE stayed very close to its 225W TBP, averaging 223W during our test. It's interesting that the A750 LE actually consumed more power, indicating the silicon binning is quite real. The A770 can likely maintain lower voltages while running at higher clocks.

While we don't normally discuss idle power use, mostly because it's not much of a concern with modern GPUs, the Arc cards do warrant an extra footnote. Power use at idle with the Arc A770 LE was significantly higher than on competing GPUs. With our Powenetics hardware, we recorded average power draw over 30 seconds of 48.1W. The RTX 3060 meanwhile averaged 15.0W while the RX 6650 XT used just 7.3W. Intel says it's aware of the problem and is working on a potential fix.

While Intel advertises a 2100 MHz typical game clock, our experience suggests the cards will do far more than that on a consistent basis. For example, the A770 averaged 2350 MHz in the Metro test, but more importantly, it also managed to run at 2375 MHz during FurMark's stress test. Considering FurMark often represents something of a worst-case scenario, the fact that the GPU clocked well above Intel's official speed means you'll likely get better performance.

Fan speeds appear to be tuned for peak temperatures of around 75C, at least for our test environment. We saw a maximum temperature of 75C on the A770 in both tests, with peak fan speeds of 2032 RPM in Metro and 2041 RPM in FurMark. Average fan speeds were quite a bit lower, with 1763 RPM in Metro and 1806 RPM in FurMark.

For noise testing, we positioned the SPL (sound pressure level) meter between the two fans at a distance of about 10cm to minimize the impact of the CPU cooling fans — and we ensured the secondary card's fans were stopped. We had a noise floor of 31 dB(A), and even though the fans were spinning, that's what the idle A770 measured: 31.5 dB(A). Under load, after 10 minutes of gaming, the fan speed stabilized at around 2000 RPM, and the noise level increased to 46.0 dB(A).

As with the A380, we couldn't measure fan speed as a percentage of maximum, nor could we set a static fan speed. Still, from a few feet away, the GPU's fan wasn't really audible over other noise sources. 

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

While the Arc A380 technically arrived first, the Intel Arc A770 represents Team Blue's best foot forward for the graphics card market. We can't say it's perfect or even a better option than the competing AMD and Nvidia cards, particularly if you play a wider selection of older DX11 (or earlier) games, but compared to the A380 just one month ago, the overall experience is much improved. If Intel can keep iterating on the drivers for a few more months, it could be a serious competitor in the midrange market.

Given the Arc A770 Limited Edition represents the best-performing Arc Alchemist GPU Intel currently has available, outside of overclocking, Arc maxes out at mainstream levels of performance. It's faster than the RTX 3060 in our testing, even in ray tracing games, but it's clearly slower than the RTX 3060 Ti. The same goes for AMD's RX 6650 XT, which it can match or beat in standard games, and it easily surpasses AMD's GPUs in the realm of ray tracing.

Intel also has some other interesting ideas. One is Smooth Sync, which obscures screen tearing with vsync off by blurring the boundary between the current and previous frames. It wasn't quite ready for use when the A380 first started getting into reviewers' hands, but after running everything in the current test suite with Smooth Sync enabled, I can say that I never noticed tearing as a problem. Others might be more keen-eyed, but it's an interesting alternative to getting a more expensive adaptive sync display with a high refresh rate.

I also watched a bunch of videos with the A770 installed, and everything worked without a problem. That's not too surprising, but it's good to know that quirks during everyday use are likely far less common than driver incompatibilities with games. Intel's media encoding capabilities remain excellent and are another potential selling point, assuming you'll actually use them.

Where the Arc A770 truly shines is in the value department. Sure, AMD's RX 6700 XT has been on sale for as little as $360 in recent weeks, but it's now selling for $410 or more. The RX 6650 XT might cost $50 less, but it also has half the VRAM and substantially weaker ray tracing performance. As for Nvidia, there's no question the A770 has the potential to deliver superior performance for a lower price. Except when the drivers occasionally hold it back.

The safer bet right now would be buying an AMD or Nvidia GPU. They're known quantities, and both companies have been working on graphics drivers for decades. Of course, Intel has decades of driver experience as well… but it's only just starting to truly take such things seriously. Just check out what we said last year when we benchmarked the Intel DG1 or even our review of the Arc A380 last month. Things are improving, but there's still work to be done — and there will be more driver work to do as long as Intel keeps supporting its GPUs.

We can't help but wonder what Intel might have been able to do if Arc had been fully ready to ship earlier this year. Whatever the cause of the delays — and there were most definitely delays, as companies don't do a full deep dive into a new GPU architecture and reveal most of the secrets a full year before launch — the fact is we now have Intel Arc GPUs facing off against AMD and Nvidia GPUs. It's great to finally see some new blood in the dedicated graphics card market.

Now, Intel needs to take the learnings and successes of Arc Alchemist and execute better on delivering the follow-up Battlemage in a timely fashion. The A770 might be able to trade blows with current midrange RX 6000-series and RTX 3000-series cards, but when Nvidia's Ada Lovelace and AMD's RDNA 3 enter that market, we suspect the A770 and A750 will be in for some stormy seas. 

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content

All aboard the Intel Arc! Or Ark? Intel's first true dedicated graphics cards are here, and we've got the Arc A750 review along with the Arc A770 review. Intel has set its sights on the value midrange market with the A750, hoping to earn a spot among the best graphics cards. We've thoroughly tested the card and have come away far more impressed than we were with the budget-friendly Arc A380.

This is a companion review to the A770, so there's a lot of additional detail in that article. You can also check out our deep dive into the Intel Arc Alchemist architecture. For the A750, we'll skip straight to the important bits and render our verdict.

Intel Arc A750 Specifications

Here's the quick rundown of the full Intel Arc desktop card lineup. The A580 hasn't launched yet, so we're missing some information like the price, but everything else should now be available. That's not quite true, as the A750 and A770 will actually go on sale next week, on October 12 — right alongside the Nvidia RTX 4090, which promises gobs more performance for slightly more than five times the price of the Arc A750. It's going to be a tough decision, we know!

The Intel Arc A750 follows the familiar pattern of taking the same GPU and core design of a more expensive model and then trimming down some features. The A750 disables four of the potential 32 Xe-Cores and has a slightly lower boost clock, giving it about 85% of the theoretical compute of the A770. It also has half the memory of the 16GB A770 Limited Edition, clocked at 16 Gbps instead of 17.5 Gbps, so that's 9% less memory bandwidth. Memory capacity is our bigger concern.

8GB of VRAM was great back in 2016 when the GTX 1070 and 1080 launched. However, six years later, we're not quite as keen on only having 8GB of memory. Granted, the RX 6600-series cards from AMD are all packing 8GB, and everything from the RTX 3050 through the RTX 3070 Ti — with the exception of the RTX 3060 — also has 8GB. But one look at where games are heading with VRAM use, and we can't help but think the A770 16GB is probably worth the extra $60. The extra memory could definitely help big brother leave his little sibling sucking wind.

Looking at the direct competition, which, based on current GPU prices, would be the AMD RX 6600 or RX 6650 XT and the Nvidia RTX 3050, things are a bit messy. Let's just get this out of the way and say that the RTX 3050 ends up hopelessly outclassed. That was already true with the RX 6600, and the Arc A750 can pour some salt into the wound. But the AMD cards aren't going to roll over so easily.

In short, AMD promises good performance and excellent value for people that aren't worried about ray tracing or fancy schmancy AI upscaling technologies. On the other hand, Intel and Nvidia offer much better ray tracing performance along with matrix cores that can boost machine learning and AI performance.

With 28 Xe-Cores and a nominal 2050 MHz boost clock — and we say "nominal" because, as you'll see later, the Arc A750 easily exceeded that mark in our testing — Intel offers 14.7 teraflops of graphics compute performance. For deep learning and AI workloads, the A750 can perform 118 teraflops of FP16 calculations, or double that for INT8 work with 235 teraops of compute. Don't worry too much about the loss in precision, as companies like Google and Facebook have proven that 8-bits is sufficient for such work.

By comparison, AMD's RX 6650 XT has 10.8 teraflops of compute, while the RTX 3050 putters along with a meager 9.0 teraflops. Of course, that's on paper, and you can't just look at theoretical numbers to determine a winner — we've seen AMD's RDNA 2 chips punch well above their theoretical specs over the past two years. And that's why we run the benchmarks.

• MORE: Best Graphics Cards
• MORE: GPU Benchmarks and Hierarchy
• MORE: All Graphics Content