Traditionally, GPUs are connected through PCIe slots or the Thunderbolt/USB4 interfaces, which offer PCI Express tunneling support. As such, external GPU solutions rely on the aforementioned interfaces, which limits their support for older systems and laptops. Unlike Intel-based Macs/MacBooks, Apple Silicon based devices do not support external GPUs, mainly due to the lack of driver support and architectural differences. So, despite their efficiency compared to traditional x86-based systems, users have reported challenges in AI workloads, especially when it comes to prompt processing.

Requirements for running an eGPU through a USB3 interface at this time include the use of an ASM2464PD-based adapter and an AMD GPU. For its tests, Tiny Corp used the ADT-UT3G adapter, which uses the same ASM2464PD chip, but out of the box, it only works with Thunderbolt 3, Thunderbolt 4, or USB 4 interfaces. The team likely employed a custom firmware to enable USB3 mode that works without direct PCIe communication. Technical details are murky, however, the controller appears to be translating PCIe commands to USB packets and vice versa.

The solution is quite hacky, as it bypasses kernel-level GPU drivers, requires specific hardware, and uses USB3, which was not originally intended for GPU communication. It essentially offloads the computation part, referring to kernel executions, from your system to the eGPU. The constraint here is that data transfer speeds are capped at 10 Gbps due to the USB3 standard used, so loading models into the GPU will take much longer than if you were to use a standard PCIe connection.

Since it uses custom user-space drivers to avoid tinkering with the kernel, the feature is limited to AMD's RDNA 3/4 GPUs, although there's a hint of potential RDNA 2 support in the future. USB3 eGPU functionality has been upstreamed to Tiny Grad's master branch, so if you have an AMD GPU and a supported adapter, feel free to try it out. We can expect Tiny Corp to provide a more detailed and technical breakdown once its developers done tidying up the code.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

We don't have much in the way of technical details about the Stargate controller at this time, but we can read between the lines of the CEO's statement, and ponder over previous announcements and roadmaps. "We have a new architecture coming out in the next couple of quarters that we call Stargate, new ASIC, clean sheet design and then, with BiCS 8 QLC … we just think that’s going to be a dynamite project," ComputerBase attributes this statement to Goeckeler, taken from a recent investor Q&A session.

Based on the above, it may well be that Sandisk is going to leverage the Stargate controller and BiCS8 NAND in 128TB SSDs in the new DC SN670 data center drives, starting from Q3 this year. Moreover, roadmaps indicate that Sandisk aims to market 256TB drives the next year, and double up again with 512TB drives in 2027. The DC SN670 enterprise series SSDs were previously flagged as PCIe 5.0, but there is also the possibility that Stargate will be a technology that renews the series with up to a PCIe 6.0 interface.

For these expansive data center drives with new BiCS8 QLC NAND, with capacities of 2Tb (256TB) per IC, it is expected that the new controller will be scalable to 64 dies per channel. As ComputerBase notes, that will likely mean 32 channels would be required on the 512TB drive (coming 2027). Meanwhile, current client SSDs typically only offer up to eight channels.

With Computex around the corner, we hope that Sandisk can bring us up to date with what look to be rapidly developing plans for its capacious new SSDs. It would also be great to see larger capacity consumer SSD drives gain momentum via new levels of performance and affordability.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

This is a new idea, so only one component is available right now for download. The piece happens to be a 3mm comb for one of their shavers, but Philips assures there will be more components made available for more of their devices over time. This isn't the release of a grand library of parts by any means, but it does showcase a shift in supporting communities in search of businesses that support repairable hardware.

The program has been primarily released to the Czech Republic. Both Prusa Research and LePub are partnering up to help support the program which they hope will be easily adopted by the local maker community. One of the main goals of Philips Fixables is to encourage consumers with sustainable repair options instead of seeking a replacement device altogether.

The official Philips Fixables web page has a link for anyone in the general public to submit a request to add a specific component. Philips will notify customers with a download link if the component they suggested is able to be shared to Philips Fixables. It's not clear what sort of turnaround time to expect for these requests and whether there are limitations on what components will be made available.

According to Philips, consumers must adhere to the recommended print settings for their components to get the best results. This is the only way to ensure the replacement part is sturdy enough to stand in for a repair. Compromising on fill space for time could make or break your user experience, for example, but if done correctly, a replacement 3D print can be a useful long term solution.

Other companies have had similar programs that support right to repair initiatives. Logitech recently unveiled a repair store in partnership with iFixit, with kits available for a handful of their products that can be repaired at home rather than replaced. Only time will tell how far the open source aspect of right to repair will be taken, but for now it looks like it's at least on the table for Philips.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

The headsets on our list of best wireless headsets are all excellent in several ways — extremely comfortable, impressive audio, premium materials, and lifestyle-worthy aesthetics. The Void Wireless v2 plastic, mesh, and memory foam, and it looks like a gaming headset that's trying to be, well, a gaming headset.

It's comfortable, sure, but while it can connect to both 2.4GHz wireless and Bluetooth, it's not trying to be an Airpod replacement, or anything of the sort — it's unapologetically part of the Void lineup. Perhaps that's not such a bad thing, though. Because while the Void Wireless v2 may not be a headset you can wear outside without looking like you forgot it was on your head, it's a great headset for when you're actually at your PC. It's light and easy to wear, with breathable mesh fabric and a handy flip-to-mute mic, and it sounds good across all types of media: games, music, movies, and TV shows.

Design and Comfort of the Void Wireless v2

The Void Wireless v2 is a lightweight, wireless over-ear headset with a non-detachable boom mic and built-in lighting. It has a flexible plastic frame with large, angular earcups and mesh-covered foam padding on the headband and earpads (which are replaceable — like they are on all of Corsair's newer headsets). It comes in both black and white; our review unit was white, with a matte white finish, silver accents, and light gray mesh fabric.

The Void Wireless v2 is almost entirely plastic, and it feels like it — the headset weighs a mere 10.58 ounces (300g), which makes it one of the lighter headsets we've tested. It's still heavier than a few headsets — the wireless Logitech G735 weighs 9.6 ounces (273g), while the wireless SteelSeries Arctis Nova 5 weighs 9.5 ounces (265g) — but it is almost an ounce lighter than the most comfortable headsets we've tested: the HyperX Cloud III (11.29oz / 320g) and the HyperX Cloud III Wireless (11.64oz. / 330g). That said, the Void Wireless v2 feels pretty sturdy — not like you can run over it with a semi, but like it will last a few years even if you throw it across the room in a gaming-induced rage (try not to do that, though).

The headset has plastic headband with mesh-covered memory foam padding at the center. Though the padded area was a little smaller than I expected, I found it to be perfectly comfortable — I was worried I might end up hitting the sides (the padding sort of deceptively starts about an inch in from where the mesh starts), but I had no issues. The earcups are attached to the headband via silver plastic yokes, which each extend about 1.5 inches (38mm) for a total inner measurement of approximately 11.5 inches (292mm). The yokes simply slide out, so you aren't beholden preset notches when you're adjusting.

The earcups are large and angular, and in addition to being height-adjustable, also swivel (flat) and tilt. They feature Corsair's ship logo on the side in mirrored silver, and have dual-zone RGB light strips on the front, so anyone who's looking at you (such as your streaming audience) will see the lighting. I mostly don't see the point of having lighting in headsets in the first place, but at least this way the lighting will be seen by someone if you're looking at your webcam, as opposed to no one (e.g. the lighting on the sides of the Corsair Virtuoso Max). The earcups have mesh-covered foam earpads, which are replaceable. The foam is soft and seems like it should be thick enough, but I did find my ears bumping up against the insides of the earcups every so often (I have two flat piercings on my right ear, and a helix and forward helix piercing on my left ear, so I'm always instantly aware of when my ears touch the inside of an earcup).

Minus the earcup issue, which may not be annoying if you don't have four cartilage piercings like I do, the headset was fairly comfortable. It has a low-moderate clamping force (just enough that it didn't fall off my head if I tilted back slowly) and a flexible frame, and I didn't feel like I needed to take it off of my head, even when I wasn't actively testing it. The mesh fabric is breathable and otherwise comfortable, and while I haven't been the biggest fan of this light gray mesh Corsair has been putting on all its headsets recently, I didn't mind it on the Void Wireless v2 — it looks a lot less out of place with the plastic frame.

The headset has a handful of on-ear controls and ports, all of which are located on the left earcup. On the face of the earcup are two buttons: a power button, which you'll also use to switch between connections, with a textured multi-function button below it.

When the headset is turned on, tapping the power button will switch between the 2.4GHz wireless and Bluetooth connections, if you're connected to two devices. The multifunction button works like a pretty typical headset multifunction button: you can press it to answer/hang up phone calls or play/pause music, double-tap it to skip a track, or triple-tap it to go back one track. You can also use it to enter Bluetooth pairing mode by holding it down until the status LED (located along the bottom edge of the earcup) blinks rapid blue and red.

Along the edge of the earcup, starting from the back, you'll find a volume rocker, which can also be pressed ("scroll press") to access the headset's onboard EQ presets. The scroll press can also be remapped in iCue to, well, basically do whatever you want. I didn't have any issues with the volume rocker, but the scroll press felt mushy and inconsistent — though it seemed to work consistently enough. Moving forward along the edge of the earcup, there's a USB-C charging port, a small LED status light, and the flip-up boom mic.

The Void Wireless v2 comes with a short (approx. 1.8ft / 50cm) USB-C to USB-A charging cable and a 2.4GHz wireless USB-A dongle. Corsair's website lists the included cable as 1.8m (it's 1.8ft).

Specs

Audio Performance of the Void Wireless v2

The Void Wireless v2 has 50mm dynamic drivers with a frequency response range of 20 Hz - 20,000 Hz — fairly standard for a gaming headset, though we've recently seen a lot of gaming companies trying to push their headsets a little outside of that range. But frequency response range isn't the only measure of audio quality, and the Void Wireless v2 does a good job of demonstrating that: while not perfect, this headset sounds very good in just about every situation, especially at its price point.

I was very pleasantly surprised by the Void Wireless v2's audio quality — games, music, movies, and TV shows all sounded great. The headset sounded especially good in games — it separated out detail nicely in my favorite layered environmental scenes (Uncharted 4, God of War: Ragnarok, Baldur's Gate 3). It also did a very good job with directional sound (though the preloaded "FPS Competition" preset in the EQ was a little too keyed up for my liking), and it also supports Dolby Atmos spatial sound for an even more impressive surround sound experience.

Music also sounds pretty good on the Void Wireless v2 — it's not a headset I'd buy for pure listening, but it's good enough that you won't feel like you need to switch headsets. The headset has a mostly-balanced sound profile, with deep but slightly restrained bass, solid mids, and clear, comfortable highs. My favorite mid-range-heavy song, K-391's Lonely World, sounded great on the Void Wireless v2, thanks to the full, balanced range and impressive dynamics. While I did notice some distortion in David Guetta's Titanium (ft. Sia), the well-tuned mids still made it fun to listen to.

Voices — in games, movies, and meetings — also sounded good on the Void Wireless v2 (though I haven't come across many gaming headsets where vocals are much of an issue).

Microphone of the Void Wireless v2

The Void Wireless v2 has a non-detachable omnidirectional boom mic, which sits at the end of a flexible gooseneck arm that you can angle toward your mouth. It's a flip-up mic, with flip-to-mute — which is by far the best way to implement hardware muting on a headset, in my opinion.

Flip-up flexible boom mics are always a little awkward: what do you grab to flip them up/down? Grabbing the gooseneck arm doesn't really work (and it also looks awkward, if you happen to be streaming or otherwise on camera). The mic does have a decently robust attachment point, which is easy to grab once you know what you should be grabbing. This doesn't solve the issue of trying to flip up the mic and having it smack into your forehead if you forget to bend it away from your mouth, however.

The mic is omnidirectional, and sounds decent for a headset mic — vocals are loud, crisp, and clear, but my colleagues could tell I was talking on a headset. My voice sounded relatively full (though not particularly warm, but perhaps that was just me), but it still had enough hollowness that there was no question I was on a headset and not using a standalone USB mic. There's also not much you can do about it, either, without third-party software:

Corsair includes Nvidia Broadcast in its iCue software, but this only works if you have an Nvidia GPU (and is mostly for noise/room echo removal). Otherwise, all you can do in iCue is turn up the mic's volume (gain) and sidetone. The sidetone is, unfortunately, terrible — I couldn't hear anything until I turned it up to around 80%, and when I finally heard it, my voice was extremely staticky/fuzzy.

Features and Software of the Void Wireless v2

The Void Wireless v2 can be customized through Corsair's universal peripheral software, iCue. iCue is one of a couple of peripheral softwares I have to re-install every time I test something that uses it, because it's too annoying to leave on my PC when I don't actively require it (the other is Armoury Crate).

You can use iCue to customize the Void Wireless v2's RGB lighting, which has two zones. You can also remap the "scroll press" button to whatever you like and turn on Nvidia Broadcast for some microphone adjustments if you have an Nvidia GPU. Finally, there's a 10-band equalizer with a couple of basic presets already loaded (flat, movie, FPS, and bass boost); you can tweak these or add your own.

Battery Life of the Void Wireless v2

Corsair rates the Void Wireless v2's battery life at "up to 70 hours" over a 2.4GHz wireless connection, and up to 130 hours over a Bluetooth connection. It also features quick-charging — 15 minutes of charging will get you six hours' worth of playtime, which is decent if you find yourself out of juice (and also necessary, as the Void Wireless v2 does not offer a wired connection).

Of course, there are plenty of things that can make the battery go faster, such as listening at a high volume or keeping the headset's RGB lighting turned up to the maximum. Still, I didn't have to charge this headset more than once during the week when I was testing it (over 2.4GHz wireless, with the lighting on), so you shouldn't find yourself too tethered.

Bottom Line

The Corsair Void Wireless v2 is a solid addition to Corsair's Void lineup — it's lightweight and comfortable (unless you have piercings), with a sound profile that will work across all sorts of media. It also looks pretty good: it's still got the Void aesthetic, which is a little more gamer-y than a lot of headsets we've seen lately, but, hey — sometimes you want a gaming headset (complete with forward-facing RGB lighting). It's also got decent battery life and a flip-to-mute mic — I still love flip-to-mute, even if flipping up the Void Wireless v2's mic meant it kept smacking me in the face and getting stuck in my hair.

The Void Wireless v2 is very similar to SteelSeries' Arctis Nova 5, which is also lightweight, comfortable, and features dual-wireless connectivity. The Arctis Nova 5 is geared a bit more toward the Bluetooth side of things, with its discreet retractable microphone and its companion app full of EQ presets, but the Void Wireless v2 supports Dolby Atmos, which is more appealing in a gaming headset than a companion iOS app. The Arctis Nova 5 is also a little pricier, at $129.99.

Human 13, Fugaku 442,010,000,000,000,000,000,000,000

The Fugaku Face Off is pretty simple, and kind of stacks things even more in the supercomputer’s favor. According to the instructions we saw, a competition entrant must solve as many numerical additions as possible in 10 seconds. Most seemed to achieve 10 or 11, but Tomooo_108, for example, managed to solve 13 sums in the allotted time. This pales into insignificance vs the supercomputer, though. Fugaku managed to complete 442,010,000,000,000,000,000,000,000 sums during the 10 second battle. That’s four sextillion four hundred twenty quintillion one hundred quadrillion calculations.

Competitors mostly found the scale of their defeat against Fugaku quite amusing, in social media comments we have seen. Some joked that if you could beat Fugaku, you wouldn’t need to use it for the free day. Others pondered over the cosmic collapse that would occur if a human could move a pen fast enough to record four sextillion sums in 10 seconds. Good news - even if you lost against Fugaku, there were consolation prizes such as a miniature Fugaku to take home (expand the Tweet embedded above to see them).

At the Nico Nico Super Conference, which seems to be dominated by tech entertainment phenomena (e.g. VTubers and Cosplay), the Fugaku Face Off is another fun diversion, but with a serious twist. The supercomputer makers remind entrants that Fugaku, and its brethren, are used for tasks like research on disaster prevention – flood damage simulation, optimal evacuation plans, and so on.

Fugaku by the numbers

Fugaku, currently ranked sixth place in the TOP500 supercomputer list, delivers 442 PFlop/s of compute. It wields 158,976 nodes (each powered by a Fujitsu A64FX (48+4 core) microprocessor), packs 4.85 PiB RAM, and runs a custom Linux kernel.

The last time we reported on Fugaku, was a story largely about its successor. The collaborative work between RIKEN and Fujitsu should start this year and the goal is to deliver a ZetaFLOPS-scale supercomputer by 2030.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Nvidia's Titan-class GPUs occupied a unique space among enthusiasts, outfitted with the fastest dies and configurations of their respective generation. While being powerhouses for workstations, they also managed to beat their consumer equivalents in gaming as well. The Titan RTX, revealed in 2018, was the last of its kind and these GPUs have since been superseded by Nvidia's 90-class GPUs: RTX 3090, RTX 4090, and RTX 5090.

The prototype carries a massive quad-slot cooler design, shadowing even the RTX 4090 Founders Edition. The card's shroud is stamped with the "Titan" designation. Der8auer defines it as "The biggest GPU I've ever held in my hand." Internally, the PCB has been mounted vertically to the side, as reflected by the I/O ports at the rear. Specifications-wise, we're looking at a fully-enabled AD102 chip with all 144 Streaming Multiprocessors functional, equating to 18,432 CUDA cores, 12.5% more than the RTX 4090. This is complemented by 48GB of G6X memory, achieved by placing 24x 16Gb modules in clamshell mode. We've already seen images of this GPU's PCB previously.

Der8auer says this GPU has a 450W TGP, despite the dual 12V-2x6 design, and that's probably a limitation of the vBIOS used. The driver used dates back to 2023, but the YouTuber didn't provide further details, likely to protect their sources.

In 3DMark Time Spy Extreme, the RTX Titan Ada beats the RTX 4090 by 15% (139.3 vs 121.3), while consuming 14% (443W vs 388W) more power. Meanwhile, the RTX 5090 is 11% faster than the RTX Titan Ada, but consumes 21% more power. Near-linear scaling at this power draw is quite impressive for the Titan Ada. This suggests the GPU still has some performance headroom available and might be able to match the RTX 5090 in some tests if it benefitted from proper vBIOS and driver support.

3DMark Speedway was less impressive, scoring 7% faster than the RTX 4090, with the RTX 5090 ahead by 33%. Remnant 2 at 4K Max sees the Titan Ada pull ahead of the RTX 4090 by 10%, still 14% behind the RTX 5090, but that's still a commendable result. Much of the same story repeats in Cyberpunk 2077. Other games were accompanied by crashes and weird anomalies due to the drivers. In efficiency (FPS per Watt), the Titan Ada beats both the RTX 5090 and the RTX 4090, albeit by a small margin.

Overall, this was an impressive showcase for what could've been the fastest RTX 40 series GPU. Manufacturing costs were likely the primary reason this GPU didn't see the light of day, as there can only be so many perfect AD102 dies (609mm²) with no defects. The RTX Titan Ada was never a practical GPU, considering the workstation-oriented RTX 6000 Ada, which used a partially disabled AD102 chip (142/144 SMs), cost between $7,000 - $10,000. Either the RTX Titan would've been more expensive than this, which is unlikely since the RTX Titan (Turing) launched at $2,500, or it would've hurt Nvidia's workstation GPU sales.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

The Iskur V2 makes a good first impression with its racing chair-inspired design, green accents, and a wide array of adjustments. From the lockable 4D armrests to the multi-way lumbar adjustment to the removable headrest pillow, the Iskur V2 has a lot going for it among the best gaming chairs.

Razer Iskur V2 Specifications

Razer Iskur V2 Design

If you’re familiar with Razer’s computing products and accessories, the aesthetic of the Iskur V2 should come as no surprise. Our review unit was finished in black "ePU" synthetic leather with green accent stitching. The stitching is most prominent along the backrest bolsters and seat base bolsters, and there's also a green Razer logo on the seatback. If the green accents are a bit too "ornate" for you, Razer also offers the Iskur V2 in black or gray without contrast stitching.

The chair has a familiar gaming/racing style, which I find appealing. I'm a car nut at heart, and the Iskur reminds me of racing bolsters that you might find in 1990s JDM sports cars. Even with the green accents, the Iskur wouldn't look out of place in an office setting. This would be especially true for the solid black and gray versions.

Synthetic leather is a more eco-friendly alternative to leather, and it looks as though it should have good wear characteristics for extended gaming sessions. One downside I noticed is that there are no perforations/ventilation for the seat, so my legs stuck to the bottom cushion when wearing shorts. It's starting to creep into the high 80s and low 90s in North Carolina, and the heat combined with my adrenaline pumping from gaming sessions isn't a good combination. I don't have any such ventilation problems with my personal gaming chair, which has a taut, mesh fabric on the seat base and backrest.

With that said, the synthetic leather makes it easy to clean up in case you have a food or drink accident on the chair. This is bound to happen for gamers at some point, and the Iskur V2 can be quickly wiped down with a wet cloth. Dealing with drinks and food mishaps with fabric is often more of a hassle; just trying removing barbeque sauce (and the smell) from a gaming chair with fabric surfaces.

Razer includes a headrest pillow with the Iskur V2 – which also has a green Razer logo -- but it doesn’t solidly attach to the seatback. Instead, there’s an adjustable loop that fastens with a plastic quick-release buckle that allows for a friction fit. Depending on how you adjust the strap, you can raise or lower its position on the seatback. Although I attached the headrest pillow, I didn’t end up using it daily. It would be nice if there was a more semi-permanent attachment point for the headrest pillow, because I could see it slipping out of your preferred position if you are inclined to use it frequently.

All the knobs, levers, and buttons on the Iskur V2 have a high-quality feel. I especially liked the knurled adjustment knobs on either side of the chair. The right knob moves the lumbar support in or out, while the left knob adjusts the lumbar support up or down.

Once fully assembled, the Iskur V2 is sturdy and creek-free. I’ve previously encountered gaming chairs that are very “talkative” when reclining or leaning on the armrests. The Iskur V2, on the other hand, is tight as a drum, letting out nary a peep even when going to full recline. The only noises that emanate from the chair are the satisfying clicks from turning the lumbar dials or adjusting the armrests.

Assembling the Razer Iskur V2

The Iskur V2 arrived on my porch in a massive box that weighed over 80 pounds. I managed to "walk" it into the house by flipping it end over end, then scratched my head as to how I would manage to transport the box upstairs by myself (my stairs leading to the second floor also have a 90-degree turn, making the ascent even more tricky). Given how cumbersome (35 x 28 x 15 inches) and heavy the box was, I opted to open it up and carry the large components upstairs, piece by piece.

The Razer Iskur V2 parts list includes:

• Backrest and seat base (left and right arms pre-attached)
• Headrest pillow
• Left and right armrest
• Casters 5x
• Gas lift, tilt mechanism
• Allen wrench, screws, and plastic covers

When I opened the box, I was surprised to see a pair of work gloves in a bag taped to the inner flap, which is a nice touch from Razer. The bottom cushion, seatback, and head cushion were wrapped in plastic to protect them during shipping. The other major components, like the base, were wrapped in flexible foam, with foam inserts interspersed for added protection.

Razer included a multitool to assemble the Iskur V2, and the whole process took no longer than ten minutes. I first had to affix the wheels to the heavy metal base, then insert the gas lift. The next step was to remove four screws from the seat base with the included Allen wrench tool, then attach the metal base for the seat and reinsert the four screws. Next, you'll need to affix the entire seat base to the gas lift, then fasten the seatback to the backrest with four Allen screws. The next step is to install a plastic cover plate on either side of the chair and a small round plastic plug to hide the screw that holds the cover plates.

The final step was to strap the head pillow to the seatback.

Razer Iskur V2 Comfort and Adjustments

The Iskur V2 comes with a bevy of adjustments to fit a wide range of body types. For starters, Razer goes all-in on comfort by using dual-density foam cushions for the seat base and the backrest. I must say, my butt felt immediately at home as soon as I sunk down into the chair.

One feature I only used once is the ability for the chair to recline back to a maximum of 152 degrees. That's not quite the 180-degree maximum recline of the Cooler Master Caliber X2, but it was extreme enough to leave me feeling a bit unsettled. Even though I know that the chair can support my weight at full recline, I couldn't quite shake the feeling that I might embarrassingly topple completely backward and spark an unforced injury on my middle-aged body. So, I was one-and-done with the 152-degree recline.

The chair supports height adjustment (floor-to-seat height ranges from 435 to 535 mm), which is controlled using the right stalk that juts out from the seat base. As is common in today's gaming chairs, the Iskur V2 features 4D armrests, which are highly adjustable. I found the perfect height and horizontal position to accommodate my arms while seated at my desk. The armrests can also pivot up to 25 degrees in or out, but I found them most comfortable pointed straight ahead.

The lumbar is fully adjustable and marks one of the most significant changes from the first-generation Iskur. The previous chair featured a lumbar pad that was only adjustable for depth, while the Iskur V2's is adjustable in three planes. It can be adjusted horizontally, vertically, or in what's called a "cross-plane path." This means the entire lumbar pad can swivel left or right. The left and right movement is beneficial for people with unorthodox sitting styles (i.e., me — I like to sit with my legs crossed in the chair), or for people constantly looking at extreme right or left viewpoints with multi-monitor setups or the best ultrawide gaming monitors. The swivel ensures the lumbar follows your body movements, maintaining pressure points for maximum comfort.

Razer says the Iskur V2 can comfortably support people weighing up to 300 pounds. I weigh roughly half of that maximum, and could easily find a comfortable seating position after about 10 minutes of tinkering with the controls. Whether I sat with my feet on the floor or with my legs crossed, my comfort levels were high. The adjustable lumbar was also much appreciated, as my primary monitor for work is a 49-inch ultrawide. I sit at the mid-point of the monitor and spend a lot of time moving my viewpoint from left to right during the day as I multitask throughout the day across macOS and Windows machines via my monitor’s built-in KVM switch.

My primary complaint about the chair relates to seat space/depth. Although Razer says that it has improved the amount of seat space compared to the original Iskur, I still felt that it was a little lacking in thigh support — even though I’m not overly tall at 5’ 10”. My “daily driver” gaming chair has a seat depth of 500mm compared to the Iskur V2's 450mm, and this was immediately noticeable when sitting with my feet on the floor.

The Iskur V2 includes a removable headrest pillow, but I don’t typically rest my head all the way back on the seatback when I'm gaming or performing everyday work tasks. So, while I attached the headrest pillow when I first set up the chair, my head only touched it once, when I tested the 152-degree recline.

Bottom Line

The Razer Iskur V2 is an excellent gaming chair, offering plenty of adjustability, somewhat restrained styling, and high-quality construction. The 4D armrests should allow even the most discerning gamers to find a comfortable position, and the highly adjustable lumbar support should do wonders to keep your back from screaming out in pain after an hours-long gaming session (although you may be screaming into your headphones for another reason).

However, the Iskur V2 isn’t perfect. The pleather seating surfaces caused my back and legs to sweat after gaming for a few hours. In addition, the relatively short seat cushion resulted in less-than-ideal thigh support.

There’s also the issue of price. The Iskur V2 retails for $649, although Amazon currently offers the chair for under $620. That’s a lot of change for a gaming chair, but the trick lumbar adjustments might be worth it for people with back problems and/or who have an unorthodox way of sitting in their gaming chairs. The chair is also well-built and looks good sitting in your gaming lair. However, if the price is too high for you, consider something like the Cooler Master Caliber X2, which has a street price of under $300.

Today, we’re looking at the new Jiushark JF800 Diamond dual-tower air cooler, available in white and black, with the company claiming cooling capabilities of up to 280W!

Will Jiushark’s latest make our list of best CPU coolers on the market? There’s no question that it deserves a shot because, spoiler, this is the strongest performing dual-tower cooler we’ve ever tested, on both Intel Raptor Lake and Arrow Lake CPUs. But it’s also pretty hard to come by in the U.S., unless you’re willing to order it directly from China.

Let’s take a look at the specifications and features of the cooler. Then we’ll go over thermal performance and noise levels.

Cooler specifications

Packing and included contents

Opening the compact box reveals the accessories, neatly packaged above the cooler.

Included in the box are the following:

• Two 120 mm fans
• Dual-tower heatsink
• Thermal paste
• Mounting accessories for modern AMD and Intel platforms
• Installation manual

Features of Jiushark’s JF800 Diamond

*️⃣ Six copper heatpipes

The Jiushark JF800 Diamond moves heat away from the CPU using six copper heatpipes – but as you’ll see in our benchmarks, it offers performance superior to similar air coolers we’ve tested.

*️⃣ Heatsink Design

The towers of the JF800 feature what appears to be a fairly basic fin design.

The tops of the heatsinks have a grille reminiscent of cattle guards, adding some nice aesthetics with the ARGB lighting shining through them.

*️⃣ Two 120mm fans with ARGB lighting

There’s more to a cooler than just the heatsink or radiator. The bundled fans significantly impact cooling and noise levels, as well as how the cooler looks in your case. The included fans here have nine blades, and there are models with or without ARGB lighting, according to your preference.

As you’ll see in the benchmarks below, the fans help enable superior thermal performance, but can be a little noisy.

Testing configuration – Intel LGA1700 and LGA1851 platform

There are many factors other than the CPU cooler that can influence your cooling performance, including the case you use and the fans installed in it. A system's motherboard can also influence this, especially if it suffers from bending, which results in poor cooler contact with the CPU.

To prevent bending from impacting our cooling results, we’ve installed Thermalright’s LGA 1700 contact frame into our testing rig. If your motherboard is affected by bending, your thermal results will be worse than those shown below.

Not all motherboards are affected equally by this issue. I tested Raptor Lake CPUs on two motherboards. And while one showed significant thermal improvements after installing Thermalright’s LGA1700 contact frame, the other motherboard showed no temperature diffrence! Check out our review of the contact frame for more information.

I’ve also tested this cooler with Intel’s latest platform, Arrow Lake and the LGA 1851 socket.

LGA 1700 and 1851 installation

The installation of this air cooler is easy with the included mounting hardware.

1. You’ll first need to place the backplate against the rear of the motherboard after assembling it.

2. Next, you’ll need to take the rubber standoffs and slide them against the poles of the backplate.

3. Then, you’ll take the mounting bars and place them on top of the rubber standoffs, securing them with the included screws.

4. Apply the included thermal paste to your CPU. If you have any questions on how to do this properly, please refer to our handy how to apply thermal paste guide.

5. Remove the middle fan from the cooler, then place the cooler on top of your CPU. Use a screwdriver to secure it using the screws in between the two towers, shown below.

6. Once complete, connect the PWM and ARGB cables to the corresponding motherboard headers. Installation is now complete.

CPU-only thermal results without power limits

Without power limits enforced on Intel’s Core Ultra 9 285K and i7-14700K CPUs, the CPU will hit its peak temperature (TJ Max) and thermally throttle with with almost all air coolers and even most liquid coolers on the market. For the best liquid coolers on the market, the results of this test will be shown using the CPU’s temperature. However, when the CPU reaches its peak temperature, I’ve measured the CPU package power to determine the maximum wattage cooled to best compare their performance. It’s important to note that thermal performance can scale differently depending on the CPU it’s being tested with.

We’ll start by looking at the performance of this cooler with Intel’s Arrow Lake Core Ultra 9 285K, as these are rather interesting!

Normally, when I’ve tested air coolers with Intel’s Core Ultra 9 285k, they run with the CPU operating at peak temperature and experience small amounts of thermal throttling as a result. Every air cooler I’ve tested thus far has demonstrated this behavior – until now! When I tested Jiushark’s JF800 did momentarily peak at maximum temperature, but operated without throttling – the first air cooler I’ve tested to accomplish this.

That said, Arrow Lake’s thermal characteristics have changed a bit since the launch due to both Windows and BIOS updates, which have altered its behavior. I’m going to need to retest coolers like Thermalright’s Peerless Assassin 140 as a result. It’s quite possible that other coolers can also manage this feat with Arrow Lake now.

Again, CPU coolers can scale (and perform) differently depending on the CPU they are paired with due to differences in manufacturing processes and the location of hotspots on the CPU. We’ve also tested Jiushark’s JF800 Diamond with the harder-to-cool Raptor Lake Core i7-14700K to see if its amazing performance would hold up with other CPUs.

With 254W cooled during the course of testing, Jiushark’s JF800 Diamond is again impressive! This is better cooling performance than I’ve seen from any other air cooler and competitive with many AIOs!

To achieve these levels of thermal performance, the fans of this cooler reach up to 48.9 dBA. This is much louder than it needs to be. I don’t understand why Jiushark allowed these fans to run so loudly, especially since – as the next section will show – they perform well even when limited to low noise levels.

CPU-only thermal results with noise normalized to 38.9 dBA

Finding the right balance between fan noise levels and cooling performance is important. While running fans at full speed can improve cooling capacity to some extent, the benefits are limited and many users prefer a quieter system.

With this noise-normalized test, I’ve set noise levels to 38.9 dBA using the i7-14700K system. This level of noise is low, but slightly audible to most people. Jiushark’s JF800 continued to offer the best performance we’ve seen from any air cooler here, with 236W cooled during this test.

253W results

My recent reviews have focused more on tests with both the CPU and GPU being stressed, but many of y’all have indicated that you would like to see more CPU-only tests. In response, I’ve started testing Intel’s “Arrow Lake” Core Ultra 9 285K with a 253W limit.

Testing with the default power limits of 253W shows very good performance, on par with Thermalright’s Peerless Assassin 140!

135W CPU + 290W GPU results

Testing a CPU Cooler in isolation is great for synthetic benchmarks, but doesn’t tell the whole story of how it will perform. I’ve incorporated two tests with a power limit imposed on the CPU, while also running a full load on MSI’s GeForce RTX 4070 Ti Super 16G VENTUS 3X.

The CPU power limit of 135W was chosen based on the worst CPU power consumption I observed in gaming with Intel’s Core Ultra 9 285K, which was in Rise of the Tomb Raider.

In this test, the JF800 Diamond continues to impress. With a temperature recorded of 66 degrees Celsius, it again takes the spot for the best air cooler, outperforming rival options from ID-Cooling, Thermalright, and Ocypus! Noise levels weren’t bad, but they were near the loudest of the results we’ve recorded thus far.

85W CPU + 290W GPU results

Our second round of CPU + GPU testing is also performed with Arrow Lake. The power limit of 85W was chosen based on typical power consumption in gaming scenarios using the Core Ultra 9 285K CPU. This should be fairly easy for most coolers. The main point of this test is to see how quietly (or loudly!) a cooler runs in low-intensity scenarios.

With a CPU temperature of 54 degrees C, Jiushark retains its lead over other air coolers. Its noise level was also relatively low in this test, averaging 38.2 dBA.

Conclusion

Jiushark’s JF800 Diamond is the best air cooler on the market that you can’t (easily) purchase. It outperforms its competitors in every thermal benchmark, even when the fans are noise-normalized. My only complaint is the lack of availability, at least in the U.S. If you want this amazing CPU cooler, you’ll need to import it from AliExpress or another Asian distributor, at least until there is better availability stateside.

Along the way, we spoke with the company’s internal display specialists: Senior Color Expert Dr. Chris Bai and BenQ’s Director of Integrated Marketing, David BW Lai. Given the level of access offered to us, which is quite rare among usually secretive tech companies, we’ll walk you through our experience visiting BenQ’s headquarters and share exactly what we saw during our time there.

Towering in Taiwan

BenQ’s headquarters are located in an impressive skyscraper in the Neihu District of Taipei, right beside fellow giant Nvidia, an equally imposing neighbour. Upon arrival, BenQ outlined a brief history of the company. The firm started as a contract manufacturer in 1984, producing products as part of Acer. In 2001, Acer sold BenQ, which became the brand we know today.

But BenQ wasn’t always just about displays. The company also dabbled in other product categories since 2001, including laptops. But in 2025, BenQ’s sole focus was on display products, including monitors and projectors, where it has found success. In 2019, BenQ became the first manufacturer of a Pantone-validated display, a breakthrough for colorists and artists alike.

The talents that BenQ has amassed could be a part of the reason behind the company’s recent success. Dr. Chris Bai, BenQ’s resident Color Expert, holds a doctorate in Color Science awarded by Taiwan’s NTUST (National Taiwan University of Science and Technology). But, as well as his senior position at BenQ, Dr. Bai serves as the Vice Chair of the ICC (International Color Consortium) and is an Expert Member on ISO TC-130. Many of you will inevitably have experience with monitors and printers using ICC profiles, so you will have some grasp on the challenges of color management.

BenQ’s AQCOLOR tech is groundbreaking

One of BenQ's main missions is to deliver "colors you can trust,” and the company seeks to do that with its AQCOLOR monitors for professionals. Dr. Bai explained that the Designer PD series and Photographer SW series stand upon four pillars of quality to achieve this. These are commitments to 10-bit color, user calibratable software (Palette Master Ultimate), industry standards validation, and providing a factory calibration report to users of each product. This is backed up with an out-of-the-box guarantee of color precision, so users won’t have to tweak for hours before getting an accurate color profile.

Beyond the talk of calibration, validation, and monitor hardware and software working together, what was most impressive was BenQ's ability to deliver a consistent color appearance via its Universal Color Communication Platform.

Dr. Bai reasons that ‘color mismatch’ is the biggest problem in the world of displays and content creation. Color mismatch occurs when photos or digital art don't look as intended when they get to a consumer monitor, projector, TV screen, or even print.

As someone who worked in reprographics for several years before writing about tech, seeing such color consistency between the diverse output media was quite eyebrow-raising. It's been years since I was in the business, and happily, the color quality and accuracy of monitors, TVs, printers, and projectors have all been raised significantly in the interim.

The game developer gambit

But it’s not just professional displays enjoying the new AQCOLOR tech. BenQ is bringing it to the gaming segment with a handful of new monitors from the company’s performance-focused PD series of panels.

Specifically, we were shown the recently-launched BenQ PD2730S and PD3226G. As well as appealing to some of BenQ’s traditional color-accuracy sensitive professional customers like designers and photographers, the new PD series has ramped up resolutions and/or refresh rates.

BenQ has done this in response to seeing an increased demand for its PD series monitors among game designers. These creative professionals also want to ensure that visual assets are communicated to players as they see them. Therefore, developers may be lured in by the accurate, consistent visuals. Other elements, such as improved pixel density and refresh rates, offer not only accurate visuals but smooth-as-butter performance, too.

To get an idea of why game developers are going wild on high-end monitors, I got my hands on one of BenQ’s latest models: the new PS2730S.

Dr. Bai had set up a color test on the monitor, and a 15-inch MacBook Pro (M2 Pro), driven by the device’s Thunderbolt port. We were told to play an odd-one-out color game on each display. However, we weren’t told the purpose of the game.

After completing a fairly simple set of tasks twice, there was a significant score difference in spotting accurate colors on the panel compared to the M2 MacBook Pro’s display. On the PS2730S, the score hit 760, whereas the MacBook Pro’s core was 560 points, meaning that on this (very simple) test, colors were more accurate to the tune of around 30%.

BenQ is also eyeing up the Mac market

BenQ has been pleased with the take-up of its MA Series monitors, which target Apple Mac users. The latest MA Series displays improve on the previous generation’s M-book mode with the more expansive iDevice Color Sync, which was launched in January 2025 and delivers ICC profiles on a per-model basis. Understandably, going from Mac to Mac across models and generations, there is some visual variation, despite Apple's efforts.

Going forward, the new color syncing will deliver fine-tuned profiles to match the subtle color differences of the Apple hardware across models and generations. Thus, your external display will be a much better complement to the built-in one thanks to iDevice Color Sync.

We also saw Display Pilot 2 software for Mac, which makes display and color management, as well as brightness syncing, very easy with a few mouse clicks or keyboard shortcuts. BenQ also teased that it has more new MA Series monitors to come, but has no further details to share publicly yet.

The gaming market has two clear segments

Regular readers will be well aware of BenQ’s gaming-purposed monitor sub-brands: the FPS and esports-focused Zowie line and the AAA and open-world-purposed Mobiuz monitors.

The former still utilizes TN technology displays for high-end eSports performance. We recently reviewed the BenQ Zowie XL2566X+ 400 Hz 25-inch gaming monitor, which has blisteringly-fast refresh rates for those who require it.

Meanwhile, the Mobiuz series of gaming monitors embraces other screen technologies to achieve the best balance of quality and performance, tailored for AAA and open-world gaming. We reviewed the BenQ Mobiuz EX321UX 144 Hz gaming monitor with Mini-LED technology last year, and it fared rather well thanks to its bright and accurate color, low input lag, and contrast enhanced by 1,152 local dimming zones. However, as you will read later in our chats with the BenQ execs, the days of Mini-LED are already numbered, with OLED likely to be the next key technology for this segment.

Inside BenQ’s Color Fidelity lab

In BenQ’s Color Fidelity Lab, we were pleased to see some real-world demonstrations of color management pipelines and the results that can be achieved with vastly different mediums like monitors, TVs, signage, projectors, and print.

Please note that my smartphone camera photos from the lab don't do it justice, due to the smart processing shenanigans that my Asus Zenfone 9 is surely applying to the scenes. Nevertheless, the shots aren't bad at communicating the consistent colors achieved in these demo workflows. Thus, we’ve also used BenQ’s lab photos in our report.

The first lab setup we observed was a purposely complex scene which included a standard Macbeth ColorChecker Chart (center), and a warm-yellow household lamp (also center) to add a bit of natural chaos to the studio lighting. The first image of the two shows the monitor in the scene before the BenQ color consistency technology has been applied. The second image in the sequence has the tech applied, so that the image displayed on the monitor does correspond to the actual scene. In our experience inside the lab, the ColorChecker Chart allowed for easy color comparison vs standards, and the scenes viewed in real life and via the BenQ monitor were impressively consistent.

There was also an additional demo, which showed a professional BenQ monitor and a 4K projector showing uncannily similar imagery. The color characteristics and quality looked very similar between the two devices. This would be a great setup for creative professionals targeting or checking content for projectors.

As someone with a background in the printing industry, I was also pleased to see how similar the countryside scene print was to the image on the monitor. Balancing paper media CMYK output (reflective) to what someone sees on screen (emissive) has been important since the dawn of color digital workflows. So, users today are very lucky to have a wide choice of great devices and support for color management technology.

We also saw a demo in which some non-color-managed displays were compared with several color-matched and calibrated displays. This demo showed that designers could accurately target TVs and signage to communicate their original creative vision, which might also include important color features like properly reproduced brand logos.

The last part of the lab shared a technology BenQ is working on to convey textures more accurately on displays. Getting colors right in eTail environments is important to minimize the return rate of their products. In this use case, the displays showed off accurate carpet sample textures, which could help customers narrow down choices before ordering a small subset of physical samples. The tech could also apply to fabrics, among other product categories.

Q&A with BenQ

We had the chance to sit down with two senior figures at BenQ, Dr. Chris Bai (Senior Color Expert, Color Technology Lab) and Timothy Y Lee (Director of the Work-Tainment Division, covering PC gaming), to quiz them on where they envision monitor panel technology, color accuracy and panel types are going next.

Our first query was about BenQ's plans and observed trends in the highly competitive gaming monitor segment. Lee answered this, saying the gaming market seems to be moving to more color-accurate displays, and a shift of preference to OLED. The firm's gaming monitors should benefit from color accuracy from the R&D done on the professional side of the business.

Gamers are also gravitating toward larger and higher-resolution displays, so we expect BenQ to follow suit.

However, Mini-LED has not and (it seems) will not be a technology that will gain favor at BenQ. Lee mentioned that BenQ has only released a single Mini-LED gaming monitor in 2024, and it didn't have the greatest impact. Dr. Bai commented that inherent Mini-LED display factors like halo effects and uniformity issues meant Mini-LED wasn't much more than a stopgap on the way to OLED.

The BenQ Mobiuz EX321UX 144 Hz gaming monitor with Mini-LED technology garnered some praise in our review, thanks to its bright and accurate color, low input lag, and contrast that’s enhanced by 1,152 local dimming zones. However, with OLED shipments ramping up, quality improving, and prices coming down, it is reasonable to expect that more new Mobiuz gaming displays will adopt this type of screen.

On the topic of favorite products, it was surprising to hear Lee point to the MA Series for Macs. He admitted he wasn't a gamer, but the Zowie and Mobiuz product managers were. His reasons for loving the MA series seem to be simply down to its success and great feedback from the target market. Dr. Bai agreed that the Mac market, in general, was more interested in color-accurate workflows.

Previously, we mentioned the close proximity of BenQ and Nvidia HQs, so we thought it reasonable to ask if there is much collaboration with the neighbors in gaming display tech.

In brief, there are some minor collaborations between the companies, but Nvidia prefers to push its proprietary technologies while BenQ prefers open industry standards. So, while it might be easy to get new BenQ monitors validated for G-Sync compatibility thanks to its neighbor, for example, BenQ prefers to adopt tech like Adaptive Sync, and leaves closer Nvidia gaming display collaboration to rivals like Asus and MSI.

As for where this all leaves the monitor market, BenQ has shown that it’s deadly serious about pushing color accuracy and new panel technologies.With segmentation for just about anyone looking for a new display, it’s clear that BenQ’s not only one of the biggest players in the segment but is also pushing it forward in clear and tangible ways that people connect with.

Deep Link was introduced in late 2020. It allows you to harness the combined power of your Intel CPU and Arc GPU to improve streaming, AI acceleration, and overall efficiency. To utilize Deep Link, you needed an Intel 11th, 12th, or 13th Generation CPU and a dedicated Arc Alchemist GPU. The suite offered four key utilities: Dynamic Power Share, Stream Assist, Hyper Encode, and Hyper Compute.

Dynamic Power Share optimizes performance and power by intelligently shifting power resources between the CPU and GPU. Stream Assist improved streaming by offloading the task from the dedicated GPU to the integrated GPU. Hyper Encode accelerated video encoding using multiple Intel processors. Lastly, Hyper Compute leveraged your Intel CPU and GPU to accelerate AI workloads in OpenVINO.

These features boosted performance in apps like OBS, DaVinci Resolve, and Handbrake. The user who originated the thread at GitHub could not get Stream Assist up and running with OBS using the latest Arc B580 paired with the Core Ultra 7 265K. Following a month-long wait, a representative relayed that Intel had discontinued software development.

It turns out that even Alchemist users had a hard time getting these features working in Handbrake and OBS. It's possible that Intel considered Deep Link a niche feature and deemed the ongoing effort and investment not worthwhile. Besides, most of these features require per-vendor validation. Development was likely dropped a while back, as Meteor Lake, an architecture that dates back to late 2023, is also not among the supported CPUs.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Still, having the Wi-Fi driver already pre-installed on the motherboard makes setting up a new computer more convenient, especially as Microsoft now requires you to create an online Microsoft Account (or log into your existing one) when setting up Windows 11 for the first time. A workaround still lets you start your new Windows PC offline, but this will likely be unfamiliar to most users.

The company did not just stop at having the Wi-Fi driver already pre-installed on the motherboard. It also added other conveniences, like the Wi-Fi EZ-Plug, which integrates the antennas into one adapter for an easier and cleaner install, various EZ-Latch designs, making it easy to add and remove PCIe components and M.2 drives and their corresponding heatsink, as well as the EZ Debug Zone, which helps you troubleshoot your motherboard if you run into any error. It also introduced pressurized thermal pads for M.2 SSDs, reducing operating temperatures by up to 12 degrees Celsius.

These innovations help make Gigabyte’s high-end motherboards stand out from the competition. Hopefully, they will eventually trickle down to mid-range and entry-level motherboards, especially the Driver BIOS feature.

With Computex 2025 just around the corner, we’re excited to see if we spot any of these on the motherboards that Gigabyte will show off at the expo. Maybe this could even start a trend where these components already come with a Wi-Fi driver out of the box to make both hardware and software installation and initial setup easier.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

We haven't had the opportunity to review the 2024 edition with an M4 chip, but we got our hands on an M3 in 2023. We noticed a significant improvement over the M1 chip and rated our overall experience with the 2023 iMac at 4 out of 5 stars. The M4 is even more powerful and capable of speeds as high as 4.4 GHz.

Apple's M4 chip is packed with processing needs and graphics support, eliminating needing a separate GPU. The M4 has an 8-core CPU alongside an 8-core GPU, which outputs to a 24-inch Retina display. This screen supports up to 1 billion colors and is illuminated by a maximum possible brightness of 500 Nits.

Regarding memory, you get 16GB of RAM and a 256GB Internal SSD for storage. It has four Thunderbolt 4 ports and can support two external 6K monitors. It also supports wireless peripherals using Bluetooth 5.3 and features WiFi 6E.

So far, the discount has not been specified as expiring, so we're not sure how long it will be listed at this price. Visit Amazon's Apple 2024 iMac product listing for more information and purchase options.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Both cards boast a boost clock of up to 3,320 MHz, slightly higher than the leaked reference specs of just 3.2 GHz. Popular hardware leaker momomo_us also shared a screenshot from a Swiss retailer showing Acer Nitro RX 9060 XT GPUs on X. The 16 GB OC variant is priced at CHF 556.70 (around $670), with the cheaper 8 GB OC costing CHF 508.30 (around $612).

It might seem that the 9060 XT is significantly more expensive than the RTX 5060 Ti series. After all, Team Green’s 16 GB version has an MSRP of just $429, with the 8 GB model set at $379. But if you check where you can buy these graphics cards, you’ll find that the $400 to $500 price range is about right, especially as many AIB partners offer models that sell at much higher prices than the manufacturer has recommended.

Furthermore, these could also be just placeholder pricing—after all, AMD has yet to officially launch these cards, and we do not have a whiff of the GPU’s MSRP. Still, we hope that the 9060 XT family will be competitively priced and that there’ll be many options at MSRP during launch, which is expected to be May 18.

What we’re not excited about, though, is the existence of an 8 GB variant for the 9060 XT. Many experts say that 8 GB of VRAM is no longer viable for modern AAA games, with Grok AI saying it’s “like bringing a butter knife to a gunfight”. We’ve even seen this with the RTX 5060 Ti 8 GB, which loses up to 10% of performance when plugged into a PCIe 4.0 slot.

We do not know how the 9060 XT 8 GB will perform yet, as AMD has not released any information about the RX 9060 XT family yet. However, we do hope that reviewers can publish their findings on the GPU’s release date, allowing us, consumers, to determine whether these GPUs are worth the asking price.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

AMD's current-generation Zen 5-based server EPYC 9005 CPUs, codenamed Turin, slot into the SP5 socket, similar to EPYC 9004 (Genoa), which used Zen 4 cores, adopting a design similar to Arm's big.LITTLE. AMD also launched these CPUs with denser Zen 4c/5c cores, with Turin Dense reaching up to 192 cores/384 threads. Similarly, AMD also offers the cost-effective SP6 socket, which currently supports AMD's EPYC 8004 (Siena) offerings with up to 64 Zen 4c cores, also rumored to be compatible with the speculated Zen 5-based EPYC 8005 (Sorano) family.

AMD is rumored to introduce new SP7 and SP8 sockets for its next-generation server offerings, succeeding SP5 and SP6, respectively. The larger SP7 platform is expected to accommodate EPYC 9006 (Venice) CPUs, which reportedly feature as many as 256 Zen 6c cores, divided across eight CCDs. This lands us at 32-core Zen 6c CCDs, which are also purported to feature 128MB of L3 cache, for 1GB of total L3 cache (128MB x 8) for a 256-core chip.

While the top core count for standard Zen 6-based Venice CPUs on SP7 remains unknown, the leak suggests a significant update in the CCD design. Each Zen 6 CCD allegedly carries 12 cores and 48MB of L3 cache, a notable increase from the eight cores and 32MB seen since Zen 2.

The lower-cost SP8 platform will allegedly support four 32-core Zen 6c CCDs for 128 cores and 512MB of total L3 cache. That's a solid increase over Siena and likely even Sorano. Standard Zen 6 options on SP8 are projected to offer 96 cores (eight 12-core chiplets) and 384MB of L3 cache. This significant improvement in core density is likely a direct consequence of adopting TSMC's 2nm (N2) process node. Although specifics have not been detailed, we can expect similar improvements in I/O capabilities, including more memory channels and increased PCIe lane counts.

The rationale behind pursuing more cores and larger caches is evident in light of upcoming Diamond Rapids and Clearwater Forest, which are Xeon processors, anticipated to be some of Intel's most sophisticated and advanced designs in recent history. Likewise, AMD is also expected to shift to advanced packaging options from TSMC with Zen 6, potentially employing silicon interposers (TSMC's CoWoS-S) or silicon bridges (TSMC's InFO_LSI or CoWoS-L) to link its CCDs and IODs. Following the typical two-year cadence between generations, we can expect the first Zen 6-based products to arrive by H2 2026.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

There have been multiple reports of students trying the challenge, which sometimes results in their laptop smoking or catching fire. And because there’s a high risk of explosion with damaged lithium batteries, these incidents resulted in classroom evacuations, class cancellations, and, in some cases, deployment of firefighters or other first responders. One incident had a student sticking a pair of scissors into a laptop, causing it to release a lot of toxic smoke. This led to one student being sent to the hospital for inhalation, while the offender is now facing charges in juvenile court.

This isn’t the first time a dangerous viral trend has started circulating on popular social media. We’ve seen several “challenges” in the previous years, including the Kia Challenge, which showed how easy it was to steal certain Kia and Hyundai model years and led to a rise in car thefts just for clout on TikTok.

There was also the Blackout Challenge, where kids would strangle themselves on live video until they passed out, which, unfortunately, has led to several accidental deaths. Finally, there was the Penny or Outlet Challenge in early 2020, where those trying it would insert a plug into a power outlet and then slide a penny onto the exposed prongs, leading to short circuits.

TikTok reacted quickly when the news started going viral and shut down the challenge. We got no results when we searched for Chromebook Challenge on the social media platform. Instead, we got this warning: “This phrase may be associated with behavior or content that violates our guidelines. Promoting a safe and positive experience is TikTok’s top priority. For more information, we invite you to review our Community Guidelines.”

We did get some results when we searched for Chromebook durability, which showed some people purposefully throwing their laptops against hard surfaces or bending them to see how long they would last. We even saw one clip of a person scratching the laptop’s port with a metal pin, but none showed a Chromebook smoking or catching fire.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Thermals often constrain the performance of high-end SSDs in an M.2 form-factor, as both the controller and memory packages tend to heat up under high loads. Most high-end drives come with a pre-installed heat spreader, and many high-end motherboards come with their thermal solutions for SSDs. However, these thermal solutions are mainly designed to cool down the controller and the upper side of the drives, whereas backplates are tailored for single-sided or double-sided drives. Apparently, Gigabyte has found a flexible solution. Literally.

Gigabyte’s EZ-Flex solution, introduced on its B850 Aorus Stealth Ice motherboard, addresses these shortcomings by combining pressurized thermal pads with a spring-loaded, flexible backplate. This setup adapts to the thickness of both single- and double-sided M.2 SSDs, maintaining uniform pressure and ensuring better thermal contact on both sides of the drive. As a result, SSD temperatures can be reduced by up to 12°C, helping to prevent thermal throttling and sustain high performance during intensive workloads.

Gigabyte’s EZ-Flex design not only improves thermal performance and enhances compatibility with both single- and double-sided SSDs, but it also ensures more consistent mechanical pressure for better stability and reduces the risk of installation errors due to its adaptive fit. As a bonus, such a design could reduce SSD bending after prolonged usage. In addition, lowering temperatures helps extend the lifespan of SSD components. It minimizes performance throttling under sustained workloads, making it especially beneficial for users handling large data transfers or demanding applications.

It remains to be seen whether EZ-Flex will now be a part of all Gigabyte’s motherboards, or will only be featured on premium solutions. Considering that the B850 is not a premium platform, EZ-Flex will likely be a widespread feature on Gigabyte mainboards. After all, just two springs are not expensive, right?

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Apple's data center processor is reportedly codenamed Baltra; it was developed in partnership with Broadcom and is expected to be completed by 2027. All Broadcom-designed AI processors typically use systolic arrays of proprietary processing elements (usually tensor or matrix units) and HBM memory stacks. However, the report says that Apple is studying different configurations for its server processor, including models that have 'double, quadruple or eight times the number of main processing and graphics cores as today's M3 Ultra.'

The M3 Ultra is a formidable CPU regarding core count: it packs 24 high-performance and eight energy-efficient cores. However, data center processors do not tend to use hybrid designs, so we can only wonder whether the report meant eight times more general-purpose CPU cores (i.e., 256 cores) or eight times more high-performance CPU cores (192 cores). Quadrupling M3 Ultra's graphics clusters will probably result in performance comparable to that of Nvidia's GeForce RTX 5090 or better.

But there is a catch: modern AI applications do not use general-purpose CPU or GPU cores for training or inference; instead, they use tensor or matrix units. Apple's NPU relies on matrix multiply logic optimized for on-device AI, for example. If Apple is developing an AI accelerator for inference, it would rather need to scale its NPU engine than the number of CPU cores. Still, AI servers need a CPU to schedule data flows and feed data to accelerators so that Apple might be developing both CPU and AI accelerators under the project Baltra moniker.

On the Mac front, the roadmap includes several new processors. The report says that the M5 system-on-chip will be updated for iPad Pro and MacBook Pro models by the end of 2025, which is no surprise. Beyond that, Apple is developing system-on-chip devices referred to internally as M6 (codenamed Komodo) and M7 (codenamed Borneo), intended to power subsequent generations of iPad and Mac devices. Another high-end chip, Sotra, is also under development, though no details about its target devices are known.

In addition, Apple is developing custom processors for its wearable devices, such as augmented reality smart glasses.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Elgato is trying to make the introduction to stream decks simpler and less intimidating with its Stream Deck Neo — the stream deck part of its budget-friendly, plug-'n-play, sustainably-packaged Neo line of streaming gear. The Stream Deck neo is a small, customizable stream deck with eight tactile LCD keys, an info bar, and two touch-sensitive buttons for scrolling through pages of said LCD keys.

It's designed to be as plug-'n-play as possible: it comes with a built-in stand and cable, it's preloaded with handy actions, links, and tutorials, and it's priced... sort of in the budget range, at $99.99. That's not exactly what I'd call budget, but it does make it the second-cheapest stream deck you can pick up, after the $60 Stream Deck Mini.

One thing that always makes a Stream Deck more versatile is a mix of input types — tactile buttons, touch buttons, knobs, dials, swipes, etc. But while the Stream Deck Neo does have both tactile and touch-sensitive buttons, the touch-sensitive buttons are, confusingly, not customizable — they're for scrolling through pages of LCD keys, and that's it. The Stream Deck Neo is impressively beginner-friendly for a stream deck (thanks mostly to Elgato's effort to make it so, not the hardware itself, however), and it is a good entry-level option. But it would be a great entry-level option if you could actually customize most of it.

Design of the Stream Deck Neo

The Stream Deck Neo is a small, white rectangular stream deck with rounded edges, a built-in stand, and a non-detachable braided USB-C cable. It's one of the smaller Stream Decks you can get, measuring 4.2 inches (107mm) wide by 3.1 inches (78mm) deep, and Elgato lists it as being 1 inch (26mm) high, but that's only when it's lying flat. It features a built-in stand with two positions.

The Neo comes with everything built-in — the stand, the cable... the software. The braided, white USB-C cable is around 3 feet (90cm) long, which is a little too short for a device that's supposed to be plugged directly into your PC and not into a USB hub, but at the same time it's also a USB-C cable so you might just end up getting a USB extension cord, anyway.

The back of the Neo does feature a channel for the cable so it doesn't interfere with the stand (or something); this does little in terms of cable management other than ensure the cable sticks awkwardly out of the back of the device. However, this does make it so that you can use the Neo lying flat, without the stand, which I suppose is probably the point.

The Neo is one of the smaller stream decks you can get, measuring 4.2 inches (107mm) wide by 3.1 inches (78mm) deep, and Elgato lists it as being 1 inch (26mm) high, but that's only when it's lying flat. It features a built-in stand with two positions — the lower position puts the deck at 2.5 inches (63.5mm) tall, while the higher position puts it at just under 3 inches (76.2mm). The Neo isn't much bigger than the Stream Deck Mini — the smallest stream deck you can get — which measures 3.3 x 2.4 x 2.3 inches (84 x 60 x 58mm) and has six keys. It's a little smaller than the Stream Deck MK.2, which measures 4.6 x 3.3 x 1 inches (118 x 84 x 25mm) without the stand and has 15 keys.

The Neo is also fairly lightweight, weighing just 7.41 ounces (210g), not that other stream decks are heavy — the MK.2 weighs 5.11 ounces (145g) without the stand, but weighs 9.52 ounces (270g) with it. The Neo's built-in stand is convenient — it folds up flat and would make a great little travel accessory for someone who needs the functionality of a stream deck when they're traveling.

The Neo is a bit sleeker-looking than other stream decks, not just because it's white — but it is white, designed to fit in with the rest of Elgato's budget-friendly streaming line of peripherals, which include the Elgato Wave Neo microphone we reviewed a couple of months ago. The Neo has a slightly concave face and rounded corners, and while it's not actually much thinner than other stream decks (the MK.2 is one inch thick without its stand — the same as the Neo when it's folded flat), these little details make it look smaller and sleeker. (To be fair, the stream deck lineup has mostly been blocky black plastic, so almost anything looks smaller and sleeker.)

All you'll find in the box is the Stream Deck Neo with its built-in stand and built-in USB-C cable.

The Neo comes with a protective plastic cover on the bottom; at first I thought this was because the Neo had a microsuction pad to help it stick to your desk, like the Razer Stream Controller X, but it appears to be normal, non-microsuction anti-slip material. It didn't really stick very well to my desk, but it also didn't move too much as long as I wasn't jabbing it at an angle. (Plus, the non-microsuction bottom meant that I could at least move it around easily.)

Specs

Buttons on the Stream Deck Neo

The Stream Deck Neo has eight LCD keys: each of which measures approximately 0.5 x 0.5 inches (12.7 x 12.7mm) and is highly customizable depending on what you're doing at any given moment. The keys are tactile and made of glossy, clear plastic. They feel like the LCD keys on other stream decks we've tested — tactile, yes, but a little mushy. Still, they're more tactile than the keys on the Loupedeck Live S or the Logitech MX Creative Console, and tactility is important when you're streaming and pressing without looking.

Underneath the LCD keys, there's a small info bar (248 x 58 pixels) that displays the date and time, which is flanked by two touch-sensitive "buttons" for scrolling through pages of LCD keys. The info bar is slightly customizable — you can either have it display the date/time, or you can have it display a page indicator, but that's it (there are, however, several different date/time and page indicator designs to choose from). The touch buttons are not customizable — their only purpose is to scroll through pages of LCD buttons.

The LCD keys are bright and colorful, and can display icons, text, and images — including GIFs (in case you need more tiny, gif-displaying screens positioned around your desk top). The LCD keys can also show real-time stats or updates, such as the weather, your current CPU usage, an audio meter, how many Twitch followers you have, etc.

Software of the Stream Deck Neo

A stream deck is only as good as its software — and stream deck software is only as good as its integrated plugins and apps. Elgato might just have the most robust stream deck ecosystem and marketplace at the moment — the Neo integrates with all of Elgato's other products, which include webcams, capture cards, mics, and lights, while Elgato's marketplace offers plugins for social/streaming platforms (Twitch, YouTube, Discord, OBS, Streamlabs, etc.), smart home and lifestyle tech, other streaming devices, and various PC/system controllers.

Of course, sorting through Elgato's marketplace to find the plugins and controllers you need for your specific workflow takes some time — and Elgato's Neo line is designed to be as plug-'n-play as possible. So, when you first plugin the Stream Deck Neo, there are already a couple of screens of buttons set up for you to give you an idea of what you can do with a stream deck. The buttons include a "Tutorials" folder, which takes you various stream deck setup videos. It's pretty basic, but it's surprisingly useful — and much less intimidating than a stream deck full of blank buttons.

In addition to plugins and hardware/software integrations, the Elgato marketplace also features a number of resources, including pre-made profiles for different types of creators, as well as for specific programs and games.

Elgato has uploaded several Neo-specific profiles (e..g "Play," "Work," "Stream," "Control Neo," etc), for a more curated plug-'n-play experience — I'm not sure a stream deck can really be plug-'n-play, but Elgato has definitely gone out of its way to make the Stream Deck Neo as beginner-friendly as possible.

Tom’s Hardware Editor-in-Chief Avram Piltch also tried out the Stream Deck Neo and found customizing both simple and powerful. He assigned the eight keys to web tools and apps he opens every day, including Google Drive, Gmail, our editorial calendar, Photoshop Elements and Slack.

The Bottom Line

The Stream Deck Neo is surprisingly accessible — Elgato has gone out of its way to try to demonstrate just how useful a stream deck can be, no matter what you're doing on your PC, but you'll still need to spend some time customizing and getting to know the Neo if you're serious about using it. It's an attractive, sturdily-built customizable controller with bright, colorful, tactile LCD keys, and it doesn't take up too much space on your desk — but it does cost $100, which is a lot if you're not sure what you'll use it for.

As for your other options: the Stream Deck Mini has dropped from its original $100 price to $60, but it only has six keys — two of which will need to act as navigation keys if you have more than one "page" of buttons. The Stream Deck MK.2 has 15 tactile LCD keys and that's it — no info bar or touch buttons. But the Neo's info bar and touch buttons are largely uncustomizable (though, I suppose, this could be changed with a firmware update), so they don't really count. The MK.2 retails for $150 but can often be found for under $130 (and close to $100 around holidays). The Stream Deck Neo is a pretty good option for an entry-level stream deck / macro pad, but it would be a much better option if the info bar and touch buttons were customizable.

The Trident Z5 CK is the latest addition to the Trident Z5 family. The CK designation, which stands for clock driver, reflects G.Skill's initiative to distinguish the new CUDIMMs from the brand's standard DDR5 memory kits. Regarding design, the Trident Z5 CK memory modules retain the bold appearance characteristic of the Trident lineup. The notable difference with the aluminum heat spreader lies in its surface. The Trident Z5 CK features a black, glossy, mirror-like finish that readily attracts dust and fingerprints, but that won't be much of a consideration after installation and cleaning.

Like the numerous other Trident Z5 memory modules, the CK variant still stands at 1.65 inches (42mm) tall. G.Skill offers both regular and RGB versions of the Trident Z5 CK memory. Whether you prefer the RGB flair or not, the heights of both variants remain the same. Only black is available, but we're not complaining, as the memory modules look stunning in piano black.

All Trident Z5 CK memory modules are equipped with a black PCB. This 48GB memory kit comprises two 24GB modules designed with a single-rank configuration. It features eight SK hynix H5CGD8MGBDX021 (M-die) ICs arranged in a single-sided layout, each with a capacity of 3GB. The power management IC (PMIC) is the Richtek 0P=BA CG3, while the client clock driver is the Rambus DR5CKD1GC0.

It is important to note that AMD's current 800-series platform does not fully support CUDIMMs. They will operate, albeit in bypass mode, which results in the loss of the CKD functionality.

Tailored for Intel's Core Ultra 200S processors (codenamed Arrow Lake) and the latest 800-series platform, the Trident Z5 CK memory modules default to DDR5-6400, the native data rate supported by Arrow Lake. The timings are set at 52-51-51-102. This kit is designed specifically for Intel chips, so it only features XMP 3.0. The primary profile is optimized for rated DDR5-8800 at 42-55-55-104 timings with 1.45V DRAM voltage. In contrast, the backup profile caters to DDR5-6400 at 32-39-39-102 timings, operating with a 1.35V DRAM voltage. See our PC Memory 101 feature and How to Shop for RAM story for more timings and frequency considerations.

Comparison Hardware

The Intel system comprises the Core Ultra 9 285K and the MSI MEG Z890 Unify-X (7E20v1A41 firmware). In contrast, the AMD system employs the Ryzen 9 9900X and MSI MPG X870E Carbon WiFi (7E49v1A23 firmware). The Corsair iCUE Link Titan 360 RX LCD CPU liquid cooler effectively maintains the operating temperatures of both the Arrow Lake and Zen 5 processors under optimal conditions.

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 balances performance and storage capacity, with 2TB of high-speed storage up to 7,400 MB/s for our Windows 11 24H2 installation, benchmarking software, and games.

Meanwhile, the Corsair RM1000x Shift ATX 3.0 power supply delivers a reliable and ample power source to our test systems, directly feeding the GeForce RTX 4080 with a native 16-pin (12VHPWR) power cable. Finally, the Streacom BC1 open-air test bench accommodates all of our hardware.

Intel Performance

The Trident Z5 CK DDR5-8800 memory kit showcased solid performance, ranking below the Xtreem DDR5-7200 C34. Regarding gaming, G.Skill's memory kit performed comparably to Corsair's DDR5-8400 C40.

During the bandwidth test, the Trident Z5 CK stood out by surpassing all competitors with a peak of 84.30 GB/s. Additionally, it achieved the highest score in the V-Ray 6 benchmark.

AMD Performance

Regrettably, DDR5-8800 exceeds the limitations of our Ryzen 9 9900X sample. The integrated memory controller (IMC) of the 12-core Zen 5 processor cannot accommodate memory of this speed. Consequently, our system would not post with DDR5-8800 memory. Unfortunately, AMD's existing platform doesn't fully leverage CUDIMMs' potential.

Although we had better luck with the Ryzen 7 9700X sample, our system remained unstable and frequently encountered BSOD errors. Due to the extent of the system instability, the Ryzen 7 9700X could not satisfactorily complete our benchmarking tests.

Overclocking and Latency Tuning

It goes without saying that SK hynix M-die ICs are among the easiest ICs to overclock. A slight increase in voltage can go a long way on these ICs. Sometimes, the memory functions perfectly with the exact timings, as seen with the Trident Z5 CK DDR5-8800 C42 memory kit.

The memory kit operates at 1.45V for DDR5-8800. Consequently, we elevated it to 1.5V. This additional voltage gave us the headroom to run the memory stably at DDR5-9200 with the same 42-55-55-140 timings as DDR5-8800.

Lowest Stable Timings

SK hynix M-die ICs can occasionally be a little stubborn with tight timings. However, by increasing the DRAM voltage to 1.5V on the Trident Z5 CK, we successfully lowered the timings from 42-55-55-140 to 40-52-52-134!

There might be a bit more leeway on the memory modules, but that would involve running the memory above 1.5V, and we generally suggest avoiding that for long-term use.

Bottom Line

The G.Skill Trident Z5 CK DDR5-8800 C42 ranks among the top tier of performance DDR5 memory. It boasts aesthetics that can impress anyone and performance that your system cannot complain about. Unfortunately, one significant limitation lies outside G.Skill's control. Intel's Core Ultra 200S is currently the only consumer chip to support CUDIMMs and has the highest possibility of playing nice with this memory kit, leaving AMD fans out in the cold. However, that doesn't mean you can't use this memory kit on an AMD platform; the DIMMs can run in bypass mode. They just require an AMD chip with an exceptional IMC to hit top speeds, which is not easy to find unless you have access to multiple samples.

Naturally, the G.Skill Trident Z5 CK DDR5-8800 C42 comes with a hefty price tag of $319.99, likely more than most regular users would consider spending on RAM. The CUDIMMs came out recently, leading to an early adopter premium similar to the initial launch of DDR5. Therefore, we don't anticipate a drop in the price of G.Skill's memory kit anytime soon. If you're determined to get DDR5-8800 memory with a 48GB capacity, the G.Skill Trident Z5 CK DDR5-8800 C42 is an excellent choice that you shouldn't overlook.

Mission Center, created by Romeo Calota, is a more user-friendly system monitoring tool that covers CPU, GPU, RAM, disks (including SMART functionality), network and fan speeds for Linux machines.

We’ve installed it onto our test machine, an 8th gen Intel Lenovo X390 laptop and we’ll walk you through installation and configuration. The user interface is very “Windows 11” but that is not a backhanded compliment. The UI is clear and easy to use, versus terminal applications which while useful, can be a little overwhelming for new users.

Mission Center Key Features

• Monitor overall or per-thread CPU usage
• See system process, thread, and handle count, uptime, clock speed (base and current), cache sizes
• Monitor RAM and Swap usage
• See a breakdown how the memory is being used by the system
• Monitor Disk utilization and transfer rates
• Monitor network utilization and transfer speeds
• See network interface information such as network card name, connection type (Wi-Fi or Ethernet), wireless speeds and frequency, hardware address, IP address
• Monitor overall GPU usage, video encoder and decoder usage, memory usage and power consumption, powered by the popular NVTOP project
• Monitor system fans
• See a breakdown of resource usage by app and process
• Supports a minified summary view for simple monitoring
• Use hardware accelerated rendering for all the graphs in an effort to reduce CPU and overall resource usage
• Uses GTK4 and Libadwaita
• Written in Rust

Installation

There are many ways to install Mission Center on your Linux machine. The Flatpak can be installed via Flathub, there is a self-contained app image release that can be used on many Linux distros. For this how to, we’ll focus on installing a Snap package on Ubuntu 25.04. Snaps are Canonical’s sandboxed application installer which enables apps to talk to the underlying OS, but not make any major changes.

1. Press the Super (Windows) key and search for “app”. Click on the App Center icon.

2. Search for Mission Center and select it from the drop down list.

3. Ensuring that the latest stable release is selected, click on Install.

4. Enter your password to authorize the installation. Installation should take just a few seconds.

5. Close the App Center.

6. Press the Super (Windows) key and search for Mission. Click on the Mission Center icon to run the app.

Those who prefer the terminal can install the Mission Center snap package via the terminal using the following command. When installed, go to step 6 above and launch the app.

sudo apt update
sudo apt install snapd
sudo snap install mission-center

Using Mission Center

Mission Center, by default, will show all of the attached hardware that it can monitor. In the case of our Lenovo X390 laptop it could see:

• CPU
• RAM
• NVMe SSD
• Wired Networking
• Wi-Fi
• GPU
• CPU Fan Speed

But, if you have the hardware it can also see your Nvidia GPU.

Monitoring Performance

Clicking on each device, found on the left hand side of the app, will show the performance of said device. By default, the graph will update every second but this can be changed.

1. Click on the hamburger icon in the top right corner and select “Preferences”.

2. Under General (top entry) change the interval as you see fit.

3. Click on the X to close the popup dialog.

The graph should now update according to your interval, but it may appear “jumpy”, if you would prefer a smoother, sliding graph.

1. Click on the hamburger icon in the top right corner and select “Preferences”.

2. Under Performance Page change set Smooth Graphs and Sliding Graphs to On. This will smooth out the lines drawn for the graph, and animate the graph’s progress using a slide.You will see a performance hit though, so enable this at your own risk.

3. Click on the X to close the popup dialog.

App Monitoring

Mission Center can also monitor the running apps and processes. We can see the RAM, CPU and Drive usage of each app and process. And in the same manner as htop or top etc, we can stop or force stop running apps / processes

1. Click on the Apps tab at the top of the Mission Center window.

2. Open another application, we opened Google Chrome.

3. Click on the app in Mission Center. You can click on the dropdown next to the app, this will show all of the running instances for the app.

4. Click on Details.

5. The new dialog box will tell you all of the resources used by the app / process. Click on the X to close the dialog.

6. In the top right corner, click on Force Stop to close the application (Google Chrome). This will, abruptly, force close the application. It is sudden and quite hostile, so only do this to a real application if it has truly crashed.

Monitoring Services

Mission Center’s third tab shows all of the running services that work behind the scenes. These can be services for Bluetooth, system security and mounting drives. By examining the details of the selected service, we can see who is the owner of the service, its status and enable / disable a service (if we have the correct permissions, or run the application with sudo).

1. Click on the Services tab at the top of the Mission Center window.

2. Select a running service.

3. Click on Details.

4. Service details are now displayed, including a means to enable / disable a service should you have the permissions to do so. Click on the X to close the dialog.

Asus Headquarters

Asus’ open-book HQ

Our visit to the Asus LiGong building was in mid-March, while several key components were still a secret. Two of the company's business units welcomed us on the day: The Asus NUC team and the Asus ROG gaming laptop technical marketing team.

As you might expect, Asus has quite an impressive headquarters in Taipei's Beitou District. The LiGong building we visited is its newer headquarters, styled like an 'open book' (the logo you see is on the rear, the spine of the book). However, certain divisions like VGA and motherboards have remained at the longer-established Asus Li-De building across the road.

Inside Asus LiGong, there seemed to be little reason for employees to leave other than to go home to sleep. There was a whole floor dedicated to a modern, glistening food court with a wide range of cuisines, a 7-11 convenience store, a juice bar, chains like Starbucks and MOS Burger, hang-out spaces, and I even noticed an on-site travel agent. After regular work hours, people also stick around for the on-site gym, clubs, and social events — we noticed the Asus orchestra had started practicing as we left.

It is also interesting to note that Asus manufacturing spinoff Pegatron remains very close to Beitou. Even though it is now an independent company, Beitou remains a good site for the ODM firm's headquarters.

Asus NUC

What’s going on with NUC?

It was incredibly interesting to spend time with the NUC team and learn how the different teams from Asus and Intel will work together in 2025. We also got to analyze the latest product stack and ponder the team's plans for the future.

The lead NUC executive we met was Mavila Wu, whose current full title is Deputy Division Director of Product Division, NUC Business Unit, at Asus. An Asus employee for approaching 20 years, Wu hosted the presentation, device showcase, and Q&As. However, other Asus executives (including some ex-Intel NUC folks) were there to share insight and field questions.

◼ NUC team transition

Our discussions began with a talk about the NUC product transition from Intel to Asus, which officially began about 18 months ago. The change meant a big cultural and technological transition. However, Asus has maintained the geographically broad expertise from the merger, with teams still collaborating between offices in Oregon, Arizona, and Taipei.

The transition occurred during the NUC 12 to 13 era, with Asus firmly holding the reins on the recently announced NUC 14 line. Of course, the emerging NUC 15 Pro, NUC 15 Pro+, NUC 15 Performance, and the closely related ROG NUC 2025 are now well-established Asus products. The Expert Center (AMD-powered) NUC-a-like products are maintained and benefit from similar thermal, power management, multiphase designs, and other advancements that we have seen come to the (Intel-powered) NUC machines recently.

◼ Asus NUC improvements – noise profile and thermals

We asked the team to highlight some of the biggest differences a user might notice between an old Intel NUC and a contemporary Asus NUC. It turns out that some design targets were changed due to Asus’ influence.

The NUC team pointed out that Asus is stricter about what is considered an acceptable level of device noise. We were told that an Intel NUC might have run at up to 45dB under load, but Asus makes sure the active cooling systems in new devices won't exceed 38dB—a noticeably quieter profile.

With the latest generation of NUCs we see they typically have their Lunar Lake or Arrow Lake processors running at TDPs configured from 25 to 40W. For a long time, these standard NUCs have been roughly 4x4-inch devices of varying heights.

◼ Asus NUC 15 Pro+

During the visit, we were pleased to see the new NUC 15 Pro+, which continues the 4x5-inch form factor that we saw debut with the NUC 14 Pro+. Specifically, these are 144 x 112 x 42mm (0.7 liter). Side by side, the Nuc 15 Pro+ body seemed negligibly bulkier than its standard compact NUC siblings (which start at 0.48 liters), but the benefits might be well worth it if you want more processing power.

The slightly larger chassis Asus uses for the Pro+ can be used effectively. According to the technical team at the presentation, the new rectangular chassis enables the use of Intel's new Arrow Lake H processors configured at up to 65W this generation. Remember, they still have to keep the noise levels below 38 dB.

Opening up the Pro+ and comparing it with its brethren shows the significantly reconfigured cooling in this model. Instead of all the cooling apparatus being mostly on one side of the PCB, the Pro+ features an array of heatpipes extending over the side of the PCB with the CPU, with the motherboard flanked by fin stacks connected to the heatpipes. Last but not least, we have what’s likely to be the largest fan you could fit (approx 90mm), attached to the hinged lid of the device, directing air over the left/right fin stacks.

◼ Asus ROG NUC 2025 and NUC 15 Performance

Another highlight of the NUC team meeting was getting to check out a sample of the ROG NUC 2025. This 3-liter device packs a lot of CPU and GPU processing power, but the Asus team assured us that it is very quiet, with its triple fan design and what looked like more than half the volume of the chassis used by the dual vapor chamber enhanced cooling system.

The ROG NUC 2025 system was powered up during the visit, but we didn't run any programs on it. But, you wouldn't know it was even on, if it were not for the ROG RGB lighting zones being lit.

Inside this high-performance yet compact machine is a potent mix of an Intel Core Ultra CPU (Arrow Lake HX) plus one of Nvidia's new Blackwell architecture GeForce RTX 5070 Ti or 5080 laptop GPUs.

We were told that the CPU was configured to run at 55W, and the GPU at up to 140W, by default. But power limits are dynamic, depending on system workloads, and the CPU could chew through significantly more than 55W if required, and the headroom was available. A sturdy 330W PSU keeps the power flowing to the ROG NUC's laptop processors, and is sufficient enough for some USB ports to function as charging/power sources.

Other attractions of the ROG NUC (2025) highlighted by the team were its support for up to five displays, its Intel Killer networking, and its onboard Thunderbolt 4 ports.

Asus's ROG NUC 2025 device will be available from H2 this year. Professionals who might need the same kind of power, without the RGB and ROG styling, should look at the near-identical NUC 15 Performance. The only thing it lacks is the typical ROG styling and RGB. A bonus for the pros is that there will be more configurations available, including a barebones option.

◼ More NUC conversation nuggets

We talked about cooling NUCs, or rather, mini-PCs in general. Interestingly, the team at the meeting insisted that NUCs aren't easier to cool or thermally manage than laptops. They might be thicker and deeper, but laptops have a large surface area for venting and heat convection, it was reasoned.

Asus also wanted to highlight the quality and reliability of its NUC line. It was claimed that the Asus NUCs (and Expert Centers) went through far more rigorous heat, humidity, and endurance testing than the "industry standard." Moreover, we were quoted a return rate of under 0.5% for NUCs. This compared well with the mini-PC average of over 3%, according to the Asus team.

◼ NUC is crucial for Asus’ grand AI plans

Going forward into 2026 and beyond, Asus sees AI as a driver for the NUC business unit. We've heard lots of this kind of thing before, but Asus isn't talking about Microsoft's lame Copilot. Instead, it has large customers who want to use Edge AI applications. We are talking about tasks like deepfake image protection, medical AI, customer service AI, and so on. With a quoted 80% of the unit's business with SMBs, governments, education, and other large organizations – this shows AI isn't just a tech talking point in this instance.

Asus 2025 Laptops

Asus’s high-end 2025 laptop range

The range of Asus laptops available in 2025 is quite staggering, with the company developing products for many niches, with additional geographic and retail partner exclusive models. So, at Asus LiGong, it was a relief to focus solely on the latest ROG designs for 2025 featuring the newest mobile processors from AMD and Intel, as well as (of course) the brand-new Nvidia GeForce RTX 50 series GPUs.

Specific Asus ROG devices we got a close look at, including a peek at the PCBs, were the ROG Flow Z13, ROG Strix Scar 16, ROG Strix Scar 18, ROG Zephyrus G14, the ROG Zephyrus G16, and the ROG XG Mobile GC34 – all 2025 models. For completeness, there are also non-Scar versions of the Strix models launching, the Strix G16 and Strix G18, which basically trim the “AniMe Vision” frills.

◼ How Asus designs its laptops

Before looking at the shiny new 2025 hardware, it was interesting to hear about the process Asus goes through from concept to finished laptop. Sascha Krohn, the firm's Director of Technical Marketing and a 10-year veteran at Asus, offered the most insight in this segment of our visit.

We were told that every design would start with digital renders, which sounds logical. Once things were narrowed down on the digital side, the earliest prototypes would be made in the physical world. So, the earliest physical sample could be simply made from cardboard or foam, or increasingly nowadays, 3D printed. It would still be pretty quick and cheap to iterate at this stage.

◼ Asus ROG XG Mobile and ROG Flow Z13 (2025)

Before the laptops, we got to see these two very interesting devices in the flesh and powered up. You've likely seen these devices in our news, but reviews may or may not have been published by the time you read this.

Firstly, let's look at Asus' latest iteration of its ROG-branded portable eGPU. The new ROG XG Mobile (2025) stands on the shoulders of its predecessor with two important new features: the upgrade to Nvidia GeForce RTX 50 series GPUs (laptop GPUs) and Thunderbolt 5 connectivity.

Both while under load and while idle, this eGPU was pleasingly quiet, as it should be, and the solitary RGB area was a ROG logo case cutout. We had mixed feelings about the smoked semi-opaque case, as from a distance the unit just looked like it needed to be wiped clean of fingerprints, or something.

The RTX 5090 laptop unit (24 GB) we had in hand was also very compact and light. It measures 8.2 x 6.1 x 1.2 inches (208 x 155 x 29.6 mm) and weighs just 1.3kg. That figure is more impressive because there's a built-in 330W power brick, so all you need is a standard IEC mains cable for power. There's a kickstand, too, providing some more placement or mounting flexibility.

Remember, this is more than an eGPU. It also has fast ports and an SD Express card reader. Asus will also be making an RTX 5070 Ti laptop GPU (12GB) variant.

Though we didn't have much time to play with the Asus ROG XG Mobile, we were pleased to see a rev. 2.0 PCB from the device. On this unpopulated sample, it is easy to see the various port connector pads to the right and the large GPU area flanked by locations for up to eight GDDR7 ICs. To the far left are the locations for all the power stage components. Putting this PCB in context with the assembled device shows how compact the internal 330W PSU really is.

We have reviewed earlier iterations of the ROG Flow Z13 and will be deep-diving into the 2025 edition, which beats a brand-new path with a Strix Halo APU at its heart, aka the AMD Ryzen AI Max+ 395.

We met the Z13 product manager, Albert Lee, who said that the scale of demand for this gaming tablet had been a (pleasant) surprise.

These tablet hybrid devices aren't very affordable due to the complex and large (hence expensive) AMD chips, plus premium supporting components. However, markets like China, which had at first seemed reticent, suddenly became very keen for stock, according to Lee. The well publicized AI processing power of these Strix Halo platforms might have piqued the interest of Chinese power users. This is good for Asus, and probably good for AMD's plans to develop and sell similar - or even stronger - APUs.

◼ ROG Strix and Zephyrus (2025) gaming laptops

We also saw and had some hands-on time with the Asus ROG Strix Scar 16, ROG Strix Scar 18, ROG Zephyrus G14, and ROG Zephyrus G16 – all 2025 models, of course.

Across the board, these models feature the latest CPUs from AMD and Intel, as well as RTX 50 series graphics. The family will initially launch for those with the appetite and budget for higher-end devices. Thus, we will see RTX 5080 and RTX 5090 variants of these powerful laptops come first.

For the best performance on the go, the thicker, heavier ROG Strix family is an easy choice. With one of these AniMe Matrix lid capped machines, you will enjoy CPUs up to 55W and GPUs with up to 175W TDPs. We were told that if a task required max CPU resources and wasn't interested in GPU acceleration, the CPU would be able to run at a sustained 140W, with boosting up to 175W, too (manual mode). When gaming, users should expect the CPU to run at up to 80W, and the GPU at 175W, for a total platform max power of 255W.

Our opened-up laptops and PCBs revealed some interesting features of the Strix laptops. These are really easy to get into, too, with a tool-less system reliant on a sliding latch which enables the bottom covers to be slid and popped off with very little effort.

It was evident that the two different sized ROG Strix models shared the same PCB design, but they fit into the two different sized chassis thanks to the floating position of the I/O daughterboard on the left (looking down on an opened unit).

The triple fan cooling system has the smallest fan approximately beneath the top of the trackpad/keyboard boundary to cool the chipset. To one side are the removable memory slots and an M.2 SSD slot, to the other another M.2 SSD slot (both are Q-latch). A sizable end-to-end vapor chamber unifies this cooling solution.

Another interesting observation in the ROG laptops was the intrusion sensor design. With the easy detach bottom cover of the Strix Scar models, this might be a particularly good idea to prevent electronic damage through carelessness. In the images you can see a black foam ‘obelisk’ on the PCB, this is hollowed out and it encircles a light sensor. When the bottom cover is on, no light can get to the sensor, but when the cover is popped off and the sensor exposed it will initiate a system shutdown – if the user has forgotten to do this, or left the machine asleep, inadvertently.

Moving along to the ROG Zephyrus refresh machines, and (as usual) these slim, attractive designs make a significant break from traditional gaming laptop aesthetics and builds. One could say they are more like a MacBook Pro or Dell XPS design, but they still offer gaming credentials – and this year’s Zephyrus machines pack in the latest and greatest silicon.

Talking to Krohn about the new Zephyrus, it was highlighted that to deliver gaming laptops so sleek, compact, and light it was necessary to keep the number of watts the CPU and GPU can consume in check. This has always been the case for this line, and if purchasers aren’t happy with this compromise, they are directed to the ROG Strix line instead.

Comparing the 16-inch Zephyrus and 16-inch Scar, for contrast, the Zephyrus is around one kilo (2.2 pounds) lighter, and less than half the volume of the Scar. Meanwhile, the Zephyrus G14 weighs about the same as an Apple MacBook Pro 14.

Asus has a number of tricks up its sleeve that mean investing in higher spec GPU-equipped members of this gaming laptop family is still worthwhile. The new Zephyrus G14, equipped with an RTX 5080 (the highest spec at this size) makes use of liquid metal on the CPU and GPU for better thermals, though it still uses a heatpipe-based cooling system.

The first new Zephyrus G16 machines with RTX 5080 or RTX 5090 laptop GPUs will feature a vapor chamber cooler. We aren’t sure whether the new RTX 5070 Ti machines will sit in the vapor chamber or heatpipe camp.

Opening up these new Zephyrus units, we see a similar sight to the 2024 models. The G14 has a replaceable M.2 2280 SSD and Wi-Fi card, and the G16 improves on this with a welcome extra M.2 2280 slot for storage expansion. The design is very clean, making best use of the limited space in these slim devices.

◼ Liquid metal shouldn't be user-serviced

We asked about the specter of liquid metal run-off, or dry spots causing issues. The internet is full of discussions of this phenomenon. In answer, Krohn said that this was incredibly rare to see in RMA samples.

Furthermore, we were told the appearance of tarnishing once the processor surfaces are exposed by a technician isn’t typically evidence of a dry spot, but a simple chemical reaction with some impurity that had found its way there. Asus insists it uses more than enough TIM to mean there will not be dry spots.

Last but not least, those who think they have liquid metal-based issues are encouraged to contact Asus support, and not to loosen or remove the laptop's heatsinks.

Asus Q&A

Asus ROG laptop executive Q&A

Sascha Krohn, the Director of Technical Marketing at Asus, was happy to share some insight into his background, education, career, and more.

Krohn was born in Germany, and you may be surprised to hear he didn't study any computer technology-related subject at university, concluding his studies after obtaining a joint degree in Biology and Spanish.

Sensing his future was in tech, Krohn thought firing out emails at an assortment of computer companies was worth a shot. As luck would have it, AMD called back with an offer of a customer support role at its UK HQ. After his grounding at AMD, the young man returned to Germany to work for a memory firm, which became his springboard to a career and a settled life in Taiwan.

The current Technical Marketing head at Asus moved to the sweet potato-shaped silicon island in the mid-to-late 2000s to work for Foxconn and set up its Quantum Force components division. His fast-paced career path saw him then move between Lucid Logix, Cooler Master (3 years), and BitFenix (3 years).

Thankfully, Krohn found an anchor at Asus and has been there for over a decade. Now, he enjoys his position as Director of Technical Marketing, which is a rather wide-ranging position, as you might expect.

◼ Krohn’s gaming PC’s of yore

Krohn's first remembered PC gaming experience was with a Compaq-branded desktop, on which Prince of Persia was installed. The PC had a turbo button that doubled the digital clock speed readout, so we suspect it was an i486DX machine. However, the young Krohn preferred his Sega Master system for casual gaming.

Later, PCs would become more attractive, and this was in the Pentium III era. Specifically, Krohn recalls purchasing a Pentium III 700e from MediaMarkt, which came with an ATI Rage graphics card of some kind. It wasn't long before the young enthusiast overclocked this pre-built from 700 MHz to 1 GHz+, and upgraded to an Nvidia GPU, likely a TNT2 card.

The first PC-DIY system that Krohn built himself from scratch was based on an AMD Athlon XP 1700+ processor, installed on an Epox motherboard. This system's CPU could be overclocked from 1.5 to 2.3 GHz without huge effort or esoteric cooling. An Nvidia GeForce 4MX was also added for extra gaming horsepower.

◼ Krohn on the future of mobile computing

It is always interesting to hear what a person in such a position at a major tech company identifies as future trends. Krohn wasn't cagey about his predictions. He reckons that the coming year will see iGPUs get even more powerful—and these are increasingly popular in hot segment devices like the ROG Flow Z13 and ROG Ally.

You can find this wireless mouse deal at Amazon, where the Glorious Series 2 Pro is on sale for just $38. The white version of the Series 2 Pro is still $98, so it looks like quite the bargain, seeing as this mouse only launched in Sept 2024.

The Series 2 Pro Wireless features Glorious' BAMF 2.0 26K sensor, which has a maximum resolution of 26,000 DPI, a maximum speed of 650 IPS, and can handle up to 50 G's of force. It's got 8K polling thanks to the included wireless dongle, and the mouse uses Glorious' own-brand optical switches.

We very recently reviewed the Glorious Series 2 Pro and considered it a worthy addition to the lineup of ultra-lightweight mice on the market. Although its generic shape doesn't stand out as very exciting, its 55-gram weight and shape are indicative of the ultralight category of mice. At just $38, however, it's one of the cheapest mice available with these high specs.

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

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

"This is a key enabler for the industry to be able to deliver the storage capacity that the market needs," shared Teh when asked about the demand for 100TB units. "There's no other technology that's able to produce this capacity of storage technology to meet the growth that the market needs."

Seagate's current largest hard drive by capacity is the Exos Mozaic 3+, a 36TB drive released this year and available only to select customers. The Moziac line uses HAMR (heat-assisted magnetic recording) technology, briefly heating recording regions and reducing the intensity required to write data to discs, greatly improving storage density.

The Mozaic contains ten glass data platters, which Seagate has produced in capacities of up to 6TB each, making Seagate's current theoretical size limit 60TB. But the current corporate roadmap does not stretch this far, with Seagate in January saying it expects its next platform, Mozaic 4+, to bring 40TB+ HDDs in 2026, and the following Mozaic 5+ platform to bring 50TB drives in 2028 or later.

This falls well short of Seagate's 2017 roadmap, which aimed for 50TB hard drives by 2026. As a result, it could be wise to take claims of 100TB by 2030 with a grain of salt, unless Seagate has plans beyond HAMR after Mozaic 5+.

Seagate will also have energy conservation interests breathing down its neck in its race to boost capacity. As the AI market continues to demand an ever greater share of the energy grid, energy efficiency will likely be a more serious concern than consolidating capacity for enterprise users.

Teh claims that Seagate's race to 100TB is keeping conservation on the mind. "With the product itself, we design it to have lower power per terabyte, or to have higher density of the device itself, such that when you actually integrate that product into your data center, you require less space, less power, less everything, because you're using your fewer drives to fulfill that capacity." He also claims that Seagate is pushing renewable energy sources to power its production lines.

Whether the fires of AI and enterprise needs are enough to push Seagate through the goalposts by 2030 is yet to be seen. While the firm consistently releases the market's largest capacity drives, its history of falling short of over-enthusiastic stretch goals casts some doubt on the bullish claims of its C-suite.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

SMIC ran into two yield and output-related issues recently. An unexpected incident during scheduled annual maintenance disrupted production lines and compromised process accuracy, leading to a drop in yield rates. On top of that, the validation of newly installed equipment uncovered performance issues that needed correction, causing additional yield fluctuations. These issues lasted for more than a month during the first quarter and are expected to affect output in the second quarter. SMIC co-CEO Haijun Zhao explicitly said average selling prices (ASPs) were stable, confirming that the revenue shortfall was not pricing-related — it was from lower sellable output due to yield loss.

All chipmakers perform annual maintenance. However, the U.S. government prohibits American makers of wafer fabricating equipment (WFE) from servicing their advanced tools in China. While SMIC engineers can perform some scheduled maintenance, they are not formally qualified, so unfortunate incidents may occur.

The same applies to newly installed equipment. Normally, tools are assembled at toolmakers like ASML, thoroughly tested, disassembled, and then reassembled on site. However, expedited deliveries of tools bypass assembly and tests at toolmakers, which may lead to issues that need a fix.

Under normal circumstances, operational issues like yield degradation or equipment problems typically impact profitability (e.g., gross margin), since they raise costs per good unit produced rather than directly cutting revenue. However, in SMIC's case, the operational issues lowered yields and disrupted production output. This likely meant fewer wafers or chips were completed and available for shipment in Q2, so revenue dropped due to reduced volume, even though pricing remained stable. So, while the root cause is operational, the effect cascades into profitability (via higher cost per unit) and revenue (via fewer units delivered).

Apparently, the installation issues were so severe that SMIC had to divert between $30 million and $75 million from its R&D budget to debug the new tools. SMIC usually devotes 8% to 10% of its revenue to R&D, which in the case of the company's Q1 earnings means between $180 million and $225 million. However, since resources were temporarily shifted to setting up and fine-tuning new tools, the company's R&D expenses in the first quarter totaled $150 million. As for capital expenditures, SMIC expects to allocate $7.5 billion to building new production capacity this year.

SMIC generated $2.247 billion in revenue in the first quarter of 2025, up 1.8% quarter-over-quarter from $2.207 billion. Sales from wafers made up 95.2% of the total and climbed nearly 5% QoQ, driven by an 18% increase in 200-mm wafer sales and a 2% uptick in 300-mm wafer sales. Factory utilization reached 89.6%, a 4.1% increase from the prior quarter. Usage levels at 300-mm sites remained strong, while 200-mm sites matched the utilization seen at the larger facilities. These gains helped counterbalance the effects of softer unit pricing and higher depreciation expenses on margins.

The co-CEO noted that SMIC evaluated tariff risks through internal reviews and close coordination with domestic and overseas partners. The estimated revenue impact was minor: under 1%. He stressed that ongoing dialogue with partners has helped the company stay ready for any trade-related disruptions.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

The NM-IB8 mounting bars only replace the NH-D15 G2's existing units; the rest of the mounting equipment (that comes with the cooler) including the washers and screws are reused. Noctua claims the adjusted mounting spot provides a 3-degree Celsius drop in temperature with the HBC version of the NH-D15 G2 and a 1-degree Celsius drop with the standard version.

The NM-IB8's new mounting location (and the chip's hotspot area) is allegedly 3.7mm north and 2mm east of the center of the LGA1851 socket. This location is directly above Arrow Lake's compute tile, which houses the P-cores and E-cores. Noctua claims this area is the hottest side of the package when running multi-core workloads, which makes sense. By contrast, the original mounting solution puts the contact point of the cooler directly in the middle of the CPU package, which resides directly over the right bottom edge of the compute tile, and nearer to the SoC tile and I/O tile.

All three versions of Noctua's NH-D15 G2 are compatible with the NM-IB8 mounting solution: the standard, HBC, and LBC versions. Other Noctua models are not compatible.

The NM-IB8 is Noctua's first offset mounting kit optimized for Intel processors. Previous generations of Intel CPUs (including LGA 1700 chips) took advantage of a monolithic die design, which always had a hotspot area very close to the center of the CPU package. As a result, they did not need any sort of mounting offset to improve cooling performance.

Noctua first started shipping offset mounts for AMD AM4 and AM5-based processors. These offset mounts also prioritize shifting a Noctua cooler's central portion toward the CPU's hottest part, namely the chiplet containing the CPU cores. However, on AMD Ryzen-based chips, the cores are located in a different location compared to Arrow Lake chips.

The NH-D15 G2 is the successor to the legendary NH-D15 that debuted all the way back in 2014. One of the most prominent upgrades with the G2 variant is the addition of two alternate variants of the cooler, tailored-made for specific CPUs. The HBC variant (High Base Convexity) has a modified contact plate optimized for LGA 1700 chips that bend heavily in their socket. LBC (Low Base Convexity) is aimed at relatively flat CPUs such as LGA 2011 or AM4 processors.

These alternate versions are primarily geared towards enthusiasts or overclockers who want to extract as much performance from the NH-D15 G2 as possible. There is still a vanilla version of the cooler that works with all types of chips and is tuned for balanced performance across a variety of chips.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

This PC is intended for those who want more customization potential, which perfectly suits this modular AI PC. Even at its core, it relies on a RISC-V AI SoC with SiFive P550 CPU cores. It can produce up to 50 TOPS for AI edge computing with 40 TOPS coming from its integrated NPU. The DC-ROMA RISC-V Mainboard II replacement is $50 more expensive than anticipated, but its release date is also much earlier.

This RISC-V-based CPU is a 64-bit processor with a total of 8 cores. It's hardly the fastest on the market, reaching up to 2.0 GHz, but it's plenty for handling various edge AI applications. There are a few options available to customize the storage options, with the cheapest being no storage at all and ranging in capacity from 256GB to 1TB.

As far as graphics are concerned, it features an integrated Imagination GPU which outputs to a 13.5-inch monitor, although a sold-out 14-inch option is available on the product page. The 13.5-inch display has a resolution of 2256 x 1504px with a refresh rate of 60Hz. Memory-wise, the DC-ROMA RISC-V Mainboard II replacement comes with 32GB of RAM, but 64GB varieties are listed, albeit sold out.

The DC-ROMA RISC-V AI Laptop comes with Ubuntu Desktop 24.04 LTS preinstalled, but you could run others in its place. It has built-in audio support with two 2W speakers and a 3.5mm jack for external audio devices. A USB-C hub provides four USB Type-C ports. A fingerprint scanner is available for added security, along with a 1080p integrated webcam. The unit is powered by a 55Whr battery.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Hardware-wise, you do get the best of the best, including the most robust power delivery for the platform, 10 GbE and Wi-Fi 7, a premium audio solution with an integrated DAC, ample storage with eight M.2 sockets (M.2 Xpander-z Slider Gen 5), a Thunderbolt 5 card, and an EZ Control Hub for fans and lighting. If you sift through the additional details, you’re almost certain to find the kitchen sink. In addition to the hardware, you get MSI’s updated Click BIOS X and several “EZ” features, including EZ Link and Ai Overclocking.

Below, we’ll examine the board's details and determine whether it deserves a spot on our Best Motherboards list. However, before we share the test results and discuss the details, the specifications from MSI’s site are listed below.

Specifications of the MSI MEG Z890 Godlike

Inside the Box…

Inside the large motherboard box, you'll find a bunch of extras. Some, like the M.2 Xpander-z Slider Gen5 card, Thunderbolt 5 card, and the Wi-Fi antenna, are snug in foam padding. The rest—cables, stickers, guides, and whatnot—are packed in a separate box.

• M.2 Xpander-z Slider Gen5 AIC
• Thunderbolt 5 AIC
• EZ Control Hub
• USB drive (drivers)
• EZ Wi-Fi Antenna
• EZ M.2 Clips II Remover
• 1 to 3 ARGB G2 extension cable
• 1 to RGB extension cable
• 1 to 3 EZ Conn-Cable (V2)
• EZ Link Cable
• EZ Control Hub SATA power cable
• M.2 Xpander card fan control cable
• Thunderbolt 5 header cable
• EZ Front Panel cable
• (4) SATA cables
• (2) Thermistor cables
• M.2 Xpander card thermal pads
• M.2 screw and standoff set
• EZ Control Hb screw seets
• Cable stickers
• Flyer/installation guide/EU regulatory notice

Design of the Z890 Godlike

The Z890 Godlike certainly looks the part of a flagship motherboard. Nearly all of the matte-black 10-layer PCB is covered by shrouds and heatsinks or the EZ Bridge. On top of the VRM heatsink is one of three RGB features, this one sporting the MSI Dragon. On top of the EZ Bridge is a 3.99-inch full-color TFT offering several features, including real-time hardware monitoring, troubleshooting, BIOS update status, and clock display. You can also personalize it with your own image.

The bottom is covered in plate-style heatsinks with a powder-coat finish, along with grooves for increased thermal dissipation. The top M.2 heatsink and chipset also feature RGB lighting. Overall, the board exudes a premium vibe, and the all-black look is sure to blend in with most builds, while the Dynamic Dashboard III feature sets it apart from others.

Focusing on the top half of the board, we gain a clearer view of the RGB feature and the VRM heatsink, which features a wavy fin design and a direct-touch cross-heatpipe. This design, in part, comprises the Frozr design, which also includes the backplate, thermal pads, and other components. There’s plenty of mass and surface area to keep the high-end power delivery components running cool, no matter which processor and what overclock you’re using.

To the right of the CPU socket are four DRAM slots with a locking mechanism on the bottom side only. MSI lists support for up to 256GB of DDR5 RAM with speeds up to 9200+ MT/s. We had no issues running our DDR5-8200 kit during testing, and I’m sure the board will reach advertised speeds with the proper kit and chip.

MSI’s EZ Bridge consolidates multiple headers and connectors into a single area. With all headers on one side, it creates a more organized cluster of cables, as opposed to having them scattered across the board. At the top are the two front-panel headers, the hub connection, dual 2-pin thermistor headers, tuning controller header, EZConn header, and a USB 2.0 header. At the bottom, there are a total of three fan headers, an ATX and PD Power Connector (up to 60W for the motherboard), a front-panel 20 Gbps USB Type-C header, and an additional Type-C connector for front-panel Thunderbolt 4 support.

For most users, the three available 4-pin fan headers won’t be sufficient. But don’t fret, as MSI includes the EZ Control Hub with seven additional fan headers, a 3-pin water flow header, two 3-pin ARGB headers, and a single 4-pin RGB header. The Sys fan headers output 1A/12W, the CPU Fan header 2A/24W, and the two pump headers each output 3A/36W. If you find yourself short on power or headers, you may want to re-evaluate your needs. MSI's EZ Bridge streamlines cable management by grouping numerous headers and connectors on one side of the motherboard. This creates a concentrated (albeit busy) area for connections, which can be more organized than a scattered set of headers.

This consolidated section includes: two front panel headers, the hub connection, two 2-pin thermistor headers, a tuning controller header, an EZConn header, and a USB 2.0 header. Located along the bottom edge are three fan headers, an ATX and PD Power Connector (supporting up to 60W for the board), a front panel 20 Gbps USB Type-C header, and another Type-C connector enabling front panel Thunderbolt 4.

Power delivery on this board is one of the ‘Godlike’ features. Totaling a whopping 30 phases, MSI splits off 26 dedicated to Vcore. Power heads from the two 8-pin EPS connectors at the top of the board onto the Renesas RAA229131 PWM controller. Power then goes to the 26 Renesas R2209004 110A SPS MOSFETs in a Deut Rail configuration. Some napkin math yields 2,860A available to Vcore, which is enough to support any overclocking adventure, even sub-ambient sessions. You’ll get everything the chip will give you with this board.

On the bottom half of the board, on the left and hidden under metal shrouds, is the flagship Realtek ALC4082 audio codec. Supporting the codec is the ESS9219Q DAC/HBA Combo. This is one of the best native audio solutions available; anything better will come from a discrete sound card.

In the middle of the board are two full-length PCIe slots and a single x4 slot. The top two slots are reinforced and PCIe 5.0-capable. The top slot supports up to PCIe 5.0 x16, while the bottom slot supports up to x8. Anytime the bottom slot is in use, the top slot drops to x8. When using the Xpander-x Slider Gen5 card in that slot, it splits to 5.0 x4/x4. If you have an M.2 drive in M.2_6, the second slot runs at 5.0 x4. You can configure the lanes in the BIOS (Advanced> PCIe Subsystem settings). Lastly, the bottom x4 slot operates at PCIe 4.0 speeds and connects directly through the chipset.

Hiding underneath plate heatsinks and one slightly larger heatsink are six M.2 sockets. M.2_1 hides under the RGB-topped heatsink and runs at PCIe 5.0 x4 (218 Gbps). M.2_2-5 are all PCIe 4.0 x4 (64 Gbps). M.2_5 also supports SATA- based M.2 modules. M.2_6 is unique, as it connects through the chipset and CPU, running up to PCIe 5.0 x4. If, somehow, these six M.2 sockets aren' t enough, the included M.2 Xpander-z Slider Gen5 card adds two more PCIe 5. 0 x 4- capable sockets. I like how the board separates the generally hot-running PCIe 5.0 M.2s, with the second on the bottom and not tucked directly under a video card.

Along the right edge are several more headers, including two 19-pin USB 3.2 Gen 1 (5Gbps) for USB, and another PCIe 6+2 pin for additional board/PCIe power. Next to that are the four SATA ports, a USB 2.0 header, and finally, the front-panel audio. If you're wondering where some connectivity is, it's probably along the right edge.

While an overwhelming majority of headers and connectivity run down the entire right edge, there are still some other headers and switches along the bottom. You'll find things like the TB 5 header, LED and BIOS switches, and even a couple for extreme overclocking. Below is a complete list, from left to right.

• Reset and Power jumpers
• Thunderbolt 5 header
• Dashboard/LED/BIOS switches
• TPM header
• Power LED 1
• Slow Mode/LN2
• Battery
• OC retry button connector
• Safe boot jumper

Looking at the rear IO, the first thing that jumped out to me is the copious number of USB ports. You can connect just about anything back here, including the fast Thunderbolt 4 (40 Gbps) Type-C. In the middle are three convenience buttons (Flash BIOS, Clear CMOS, Smart Button), Marvell 10 GbE and Killer 5 GbE ports, Killer-based Wi-Fi antenna connections, and last, the 2-plug plus SPDIF audio stack.

Although it’s not part of the rear IO the included Thunderbolt 5 accessory card adds even faster ports to the back of the PC. Boasting 160 Gbps transmission speeds (120/40) it’s the fastest port around. The two DisplayPorts support multiple 8K external displays and 27W charging power. What’s missing?

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Firmware

MSI has updated its BIOS for Z890 (and X870), renaming it “Click X” and improving the aesthetic and functionality. The company changed the format from its previous UEFI, with the headings on the left side, details in the middle, and system status information on the right. Like most other board partners, it has an informational EZ Mode, allowing limited functionality, such as Game Boost, PBO, XMP, and more. The black background that fades to gold with light-colored text is easy to read, and most of what you need is at your fingertips. Overall, I like the new layout, and maneuvering around is intuitive, though it does take some getting used to.

Software

MSI Center is a single utility that offers a wide range of functionality. From hardware monitoring to RGB control with Mystic Light, the software features numerous applets and serves as a one-stop shop for downloading additional utilities, including overclocking and fan control. Those looking to use the Gamebar feature, Super Charger, or any other utilities MSI offers will find them all in MSI Center.

Test System / Comparison Products

We’ve updated our test system to Windows 11 (24H2) 64-bit OS, with all updates applied as of late September 2024, including the Branch Prediction Optimizations for AMD. Hardware-wise, we’ve updated the RAM kits (matching our Intel test system), cooling, storage, and video card. Unless otherwise noted, we use the latest non-beta motherboard BIOS that is publicly available. Thanks to Asus for providing the RTX 4080 TUF graphics card and Crucial for the 2TB T705 SSDs. The hardware we used is as follows:

Test System Components

Benchmark Settings

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

Our standard benchmarks and power tests are performed using the CPU’s stock frequencies (including any default boost/turbo) with all power-saving features enabled. We set optimized defaults in the BIOS and the memory by enabling the XMP profile. For this baseline testing, the Windows power scheme is set to Balanced (default) so the PC idles appropriately.

Synthetic Benchmarks

Synthetics offer a valuable method for evaluating a board's performance, as identical settings are expected to yield similar results. Turbo boost wattage and advanced memory timings are areas where motherboard manufacturers can still optimize for stability or performance, and these settings can impact specific testing scenarios.

The Z890 Godlike performed well across the synthetic benchmarks, leading the pack in some and landing around average in others. There’s nothing to worry about here.

Timed Applications

The timed applications told the same story, with the board posting the lowest times in both Handbrake tests and LAME. Everything is in line with the timed applications.

3D Games and 3DMark

Starting with the launch of Zen 5, we’ve updated our game tests. We’re keeping the F1 racing game but have upgraded to F1 24. We also dropped Far Cry 6 in favor of an even more popular and good-looking game in Cyberpunk 2077. We run both games at 1920x1080 resolution using the Ultra preset (details listed above). Cyberpunk 2077 uses DLSS, while we left F1 24 to native resolution scaling. The goal of these settings is to determine if there are differences in performance at the most commonly used (and CPU/system-bound) resolution with settings most people use or strive for (Ultra). We expect the difference between boards in these tests to be minor, with most falling within the margin of error. We’ve also added a minimum FPS value, which can affect your gameplay and immersion experience.

In the most crucial tests for some people, the board performed on the faster side of average in the actual games, and was merely average in the UL 3DMark benchmarks. In other words, the Godlike is a solid performer across a wide variety of applications.

Overclocking

The Godlike ran our DDR5-8200 memory without issue, but that was to be expected considering the 9200 MT/s peak listed. With the right memory kit (read: one on the memory QVL) and a good IMC, we have no doubt it can reach those speeds, and likely even further with exotic cooling.

Power Consumption / VRM Temperatures

We used AIDA64’s System Stability Test with Stress CPU, FPU, Cache, and Memory enabled for power testing, using the peak power consumption value from the processor. The wattage reading obtained is from the wall via a Kill-A-Watt meter, capturing the entire PC, excluding the monitor. The only variable that changes is the motherboard; all other parts remain the same. Please note that we have moved to using only the stock power use/VRM temperature charts, as this section aims to ensure that the power delivery can handle flagship-class processors.

Power usage on this motherboard peaked at 346W, with the CPU reaching a maximum of around 255W, averaging 225W. Idle use was on the high side at 75W, but still slightly lower than the Taichi OCF and its sibling, the MEG Z890 Ace. Power usage should be marginally higher, as the board features numerous components that require power.

VRM temperatures from our probes peaked at just over 48 degrees Celsius, with the sensor reporting a slightly higher value of 52 degrees Celsius. The Frozr Design heatsink and wavy fin design effectively maintain the high-power components within specification. The only factor holding your CPU back is the cooling method, as the Godlike is capable of taking your scores to the promised land.

Bottom Line

The MSI Godlike boasts the fastest and most comprehensive features available on an E-ATX-sized motherboard. 10 GbE and Wi-Fi 7? Check. Eight M.2 sockets (4x 5.0 with the expansion card)? Check. Thunderbolt 5? Yes. EZ Features and AI Features? They’re in there, too. In terms of aesthetics, the clean design, tastefully implemented RGB lighting, and the Dynamic Dashboard III exude high-end vibes while still prioritizing functionality. It sounds perfect, and hardware-wise, it does have it all.

The competition, including the Gigabyte Z890 Aorus Xtreme AI TOP and the Asus ROG Maximus Z890 Extreme, are priced at $999.99. However, as of this writing, the Asus model is available for $799.99. I won't explore all the differences, but there aren’t many, as both offer 10 GbE, Thunderbolt 5, and overbuilt power delivery, and possess the characteristics of a flagship board.

In other words, the Godlike’s biggest (and maybe only) drawback is the price point. $1,329 is a substantial investment. And although it boasts a ‘Godlike’ feature list and specifications, it is overshadowed by competing companies who offer their competing wares for about $300 less. Still, many desire the best of the best, and the Godlike shares its throne with no one as one of the most capable motherboards on any consumer platform.

MORE: Best Motherboards

MORE: How To Choose A Motherboard

MORE: All Motherboard Content

With limited global coverage and availability, Lenovo's Yoga Pad Pro series is likely reserved for the Chinese market, serving as 2-in-1 hybrid devices that can be configured as either a laptop or a tablet. The previous iteration, which launched last December, was powered by Qualcomm's Snapdragon 8 Gen 3 chipset. Since Lenovo is predominantly a PC/Laptop-first company, their decision to invest in a custom chip for a relatively smaller market is somewhat intriguing.

The SoC in question features a decacore (2+2+3+3) layout based on the Arm64-v8a ISA, though the precise CPU core architectures and configuration details have not been specified. The fastest cluster is specified to hit 3.29 GHz, complemented by a trio of cores at 2.83 GHz, another pair at 1.9 GHz, and the remaining at 1.71 GHz.

Digital Chat Station alleges the SoC uses an Arm Immortalis G720 GPU, also found in MediaTek's Dimensity 9300 and 8400 SoCs. Fabrication-wise, another leaker, Olrak, alludes to a 5nm process from TSMC instead of SMIC or Samsung.

Chinese manufacturers are actively pushing out indigenous chip solutions, aligning with the nation's objective to reduce reliance on Western technology. A significant impediment to China's progress lies in semiconductor manufacturing, where SMIC has been limited to 7nm-technology for several years. Surprisingly, the US ban hammer hasn't struck down Lenovo or Xiaomi, for that matter, so these entities can actively procure leading-edge chips from TSMC, at least for the time being.

Since Lenovo has not detailed the specifics of this SoC, we will have to wait until independent reviews of the tablet or the official datasheet, whichever comes first.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Among the changes of the new AI GPU is reduced memory capacity, which will affect the ability of the processor to train large language models with tens and hundreds of billions of parameters, one Reuters source noted. In addition, the final performance characteristics may be adjustable depending on how end customers configure the HGX GPU modules, potentially enabling minor flexibility despite the overall downgrade.

In April, U.S. authorities restricted sales of AMD's Instinct MI308, Nvidia's HGX 20, and some other AI processors to China (including Hong Kong and Macau) due to unpublished export regulations. As a result, AMD and Nvidia had to write down inventory worth billions of dollars ($5.5 billion for Nvidia, $800 million for AMD) as export licenses for goods bound for China are reviewed with a presumption of denial.

However, to sustain access to the critical market, Nvidia is now incorporating a set of downgrades to ensure it stays within the permissible thresholds set by the U.S. government. Nvidia has already communicated these plans to major customers in China's cloud computing sector.

If the information about the downgraded HGX H20 processor is correct, it will be Nvidia's third Hopper-based GPU designed specifically for China that relies on cut-down H100 silicon. It is noteworthy that there is no word on Blackwell-based AI products for China.

After the U.S. government imposed its first set of restrictions on AI and HPC GPUs exported to China in 2022, Nvidia built its H800 processor with reduced networking capabilities. The H800 only represented a minor performance downgrade compared to the original H100. After the U.S. administration imposed new export rules that restricted the performance of processors shipped to China, Nvidia developed its HGX H20 based on the H100 for AI training as well as L20 PCIe and L2 PCIe cards based on the AD104 GPUs for AI inference.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

The AI was trained on a dataset with more than 47,000 LEGO structures that build over 28,000 unique 3D objects, including bookshelves, tables, chairs, cars, ships, guitars, and more. This was then used to train the AI model, allowing it to create unique and original designs solely from text inputs.

The tool is available for free on GitHub, and you can pair this with a computer vision model or image processing AI. For example, you can take a photo of your available LEGO bricks and let the AI give you a multitude of unique options for building with what you already have.

The team added a validity check and physics-aware rollback during autoregressive inference, ensuring that the final output will always be valid (i.e., no overlapping bricks) and stable (i.e., no floating bricks). Furthermore, LegoGPT’s final output can be built by both humans and robots.

This is how the team created the dataset — StableText2Lego — used to train LegoGPT: a text prompt input is first converted into a ShapeNetCore mesh. This is then plugged into a 20 x 20 x 20 voxel grid from which the initial LEGO brick layout is determined.

This layout is then varied while still keeping the overall shape, and then unstable designs are filtered out from the final output. Those left are then rendered in 24 different viewpoints, and then GPT-4o is used to generate descriptions for the final output.

This is how it creates a new design through text: LegoGPT converts the text into a LEGO design, which is then converted into text tokens ordered from bottom to top. Instructions are then created to pair the structured LEGO bricks with annotations explaining the design, so that the AI will understand the relationships between the text prompt and the physical bricks.

From there, LegoGPT predicts the next brick needed to build the design using an autoregressive model. That means it will verify a brick’s validity at each step, checking if it is well-formatted, exists in the library, and does not overlap with existing bricks. This will continue until the design is completed, after which its stability is tested.

If the AI determines that the output is unstable, it will roll back to the last stable state and continue generating from that point. Once it gets a stable final output, then the design is complete.

If you want to play with the AI yourself, the team released its dataset, code, and models, making it easier for anyone to fork the team’s work. One development we can see is if someone converts this into a downloadable AI app with a customizable brick library.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Following sanctions imposed in 2020, the US significantly limited Huawei in its ability to work with international chip designers and manufacturers. Within a short span, the firm was forced to rely on indigenous alternatives, sourcing chips from its HiSilicon division, with manufacturing primarily handled by China's SMIC. Unlike Loongson's chips, which employ its custom LoongArch architecture, Huawei's cores utilize off-the-shelf Arm Cortex designs, and now its own Arm-based Taishan V-series architectures.

Huawei's laptops have mostly been powered by hardware from AMD/Intel, with a few exceptions. While Huawei showcased its upcoming HarmonyOS-powered Matebook Pro laptops, enthusiasts found evidence that the Kirin X90 SoC powers them through the system information. Further details by leaker Digital Chat Station suggest the Kirin X90, codenamed Charlotte Pro, wields a 10-core (4+4+2) / 20-thread layout.

The design is said to mirror the Kirin 9010, found on Huawei's Pura 70 series mobile phones. Therefore, the four prime cores will likely utilize the Taishan V121 architecture. In contrast, the performance cores step down to Taishan V120, and the efficiency cores probably use a standard Cortex design, likely the Cortex-A510.

There aren't many performance indicators yet, but server-oriented designs, likely the Kunpeng 930, of the Taishan V120 architecture, match AMD's Zen 3 in single-core performance. The Kirin X90 will still probably be built using SMIC's aging 7nm process node, which is the likely bottleneck. Having their own OS and silicon makes Huawei less dependent on Western tech. It ensures tighter integration within its ecosystem, leading to better optimization, reliability, faster development, and performance.

So, how many developers will port their existing apps to HarmonyOS? That depends on how well the platform takes off. Compatibility layers are a solution, but in the end, if Huawei wants a bite of Apple's market share, it needs a strong ecosystem, including essential daily apps. That being said, building homegrown software alternatives is also an option that fits China's ambitions of tech autarky. No matter what, we can expect to see some actual performance tests of the chip once the Matebook Pro comes out on May 19.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Many reviewers already have the GPU in their labs and test benches, but they still haven’t received the pre-release drivers typically included in the review kit. Several publications that usually receive these pre-release drivers have taken to social media and their respective pages to say they still haven’t gotten the required software to run what would be Nvidia’s most affordable RTX 50-series (codenamed Blackwell) GPU.

Team Green finally confirmed this to Tweakers.net, who said that it will only make the RTX 5060 drivers available on the day of release. This means that the tech media won’t be able to test the GPU until May 19 and publish their results after it goes on sale.

Another complication is that the GPU will arrive exactly one day before Computex 2025. This will make it harder for publications with smaller teams to run tests upon the driver’s release, as most of their people will likely travel to Taiwan to attend the tech expo.

Because of this, budget gamers who are keen on getting a new “affordable” GPU will have no way of gauging how the RTX 5060 will perform on the day of launch. If you’re one of those raring for a new graphics card, we recommend holding off until you see independent reviews for the mainstream card.

After all, a new GPU release does not automatically mean that you’re getting a good, powerful card for the money you’re spending. For example, the RTX 5060 Ti 8 GB, which has a $379 MSRP, falls behind the $429 16 GB version, especially when attached to a PCIe 4.0 slot.

Nvidia did not give a reason for the delay, but it could be that it’s still making some last-minute tweaks to the GPU driver. That way, it can ensure that the RTX 5060 will deliver some performance measure despite only having 8GB of VRAM. Or the company is avoiding getting its lowest-tier RTX 50-series GPU compared to the Intel Arc B580, which Tom’s Hardware has dubbed the $249 GPU champion, even before it starts selling.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

🟦🟥 Stock Build Plates 🟥 🟦

Before we dive into aftermarket print surfaces, a word on stock plates. Every modern 3D printer we’ve reviewed since 2024 has come with a very usable build plate. Textured PEI flex plates have become somewhat standard, and that’s a good thing. If you’re buying a printer today, you may be quite happy with your stock plate. You may only need this list if your build surface has been scratched, worn down, or otherwise damaged.

3D printing has been constantly evolving, with new technologies emerging every year. My first printer had a rudimentary glass build plate held down with binder clips, but now I have a plate with my own logo embedded right in the center.

There’s a buffet of build plate options out there: flexible, coated, textured, dual-sided, and optimized for specific filaments and print styles. But not all build surfaces are created equal. Here are the best third-party build surfaces you can buy for your 3D printer to get that first layer buttery smooth.

Best Build Plates for 3D Printers at a glance

Best Build Plates For 3D Printers You Can Buy Today

Tyson Flex Plate System

Tyson Build Plates are a revolution in 3D printing build surfaces. The T-95 plate is made from a thick Tempered 1095 spring steel with a proprietary coating that is truly universal. The plate can handle everything from PLA to PEEK, and the only limits are from the machine you put it on.

We tested this plate on the Qidi Tech Plus 4 and ran PLA, PETG, TPU, ASA, and Nylon without glue stick or any adhesion aid. Prints pop off easily and release on their own when allowed to cool. The bottom of the print is smooth and glass-like.

For an extra charge, the plate can bear your logo or an etched pattern that will transfer to the bottom of the print.

BIQU CryoGrip Pro “Glacier”

Glacier is one of BIQU’s two new cold plates, which rely on texture and Arctic Fairy Dust to hold down prints with minium heat. Ok, we’re kidding about the fairy dust, but there’s definitely some magic to the “porous oleophobic polymer coating” used on the top surface. Glacier plates have superior holding powers and defy dirt and grease from ordinary handling, and are scratch resistant. They work for every material without a need for gluestick, even TPU.

Because the Glacier is a cold plate, it will continue to hold at room temperature, which is nice if you’re worried about power outages. Most prints come off with a flex after it’s cooled a bit, but smaller things – like flexi animals – may need light persuasion. If you still have problems, 10 minutes in the freezer will help.

BIQU CryoGrip Pro “Frostbite”

Frostbite is the other new cold plate from BIQU, and could have been called “Rottweiler” considering how well it holds on to prints. It has a courser texture than Glacier, and holds prints at low temperatures. The plate is almost TOO sticky, but if you’ve ever had just one segment of a 3-day print pop off, you’ll appreciate the holding power of this plate. Frostbite plates defy dirt and grease from ordinary handling, and are also scratch resistant.

Because the Frostbite is a cold plate, it will continue to hold at room temperature, which is nice if you’re worried about power outages. Prints need more than a little flex to pop off this plate, and definitely need to cool down before you even try. If you still have problems, 10 minutes in the freezer will help.

SliceWorx Dipped PEI

Like Oreos dipped in chocolate, SliceWorx had a brilliant idea: what if we dip a PEI Flex plate in…rubber? Ok, it's not rubber, and they aren’t saying what the “high quality polymer” actually is, but rubber is pretty close. The coated surface of SliceWorx’s dipped plates is pure genius for those of us hooked on tiny articulated prints with a high failure rate.

Like the FrostBite plate from BQ, these dipped wonders are strictly for cool filaments like PLA and PETG. They need to be cooled a bit before flexing to remove your prints.

WhamBam PEI

Wham Bam PEI flex plates are no longer “revolutionary” after manufactures adopted PEI sheets for themselves. But their commitment to quality ensures that when you need a better build plate, Wham Bam has you covered with every size imaginable.

It’s line of PEI plates are known for consistent quality and come in both a smooth sheet that gives a shiny surface, or a baked on textured plated.

Holographic Build Plates

A holographic build plate has a specialized surface that leaves a shimmering “holographic” pattern on the bottom of the print. The plate has an ultra fine textured surfaces that diffracts light and can be transferred to the PLA print.

The holographic sticker is easily scratched, so great care should be taken when removing prints. The effect is also spoiled by finger oils and glue stick, so you’ll need to keep the plates super clean and additive free, which is why PLA is the go-to filament for these special effects plates.

The plates do not have any special gripping powers, so you’ll need to have a well tuned printer, and slow your speed for the best chances.

Custom Textured PEI Plate

Ember Prototypes knocked our socks off when it introduced custom PEI plates that bore whatever artwork you wanted to show off on your prints or the printer itself. The plates are not laser cut, but rather uses a proprietary coating that leaves a raised image on the plate that can transfer to the bottom of your print.

While they have some generic artwork available, the best plates are the ones where you supply the art. I had one made up with a hexagon pattern and my logo in the center, which is a little hit or miss as to what will end up on the finished print. Because these are bespoke plates, theres’s no refunds, so double-check your artwork before sending it over.

Best Build Plates for 3D Printers: How We Test

How we test the best build plates for 3D printers

We test all the build plates we recommend.

Our favorite test model for build plate adhesions is the humble “Mini Cute Octopus” by McGybeer – one in each corner – run at default speeds for the printer. We used a Bambu Lab P1P, A1, A1 Mini, QIDI Plus 4 and MK4s, all capable of running at an average speed of 200mm/s. It’s eight tiny tentacles each have 5 sections, meaning there’s 40 individual sections, some as small as 6x11mm.

I tested each printer’s stock plate – many of which failed without the help of glue stick - before switching to the aftermarket plate. Each plate had a few weeks of regular printing in addition to the test prints. The Octopus models will be given to children (for free) at 3D printing festivals when they visit my table, so there is no plastic wasted.

Best Build Plates for 3D Printers: FAQs

FAQs for the best build plates for 3D printers

🟢 What’s the best build plate surface to print on?

A clean one! Many first layer issues can be fixed by washing your build plate with hot soapy water, then drying the plate with a lint free cloth. Preheating the bed and using a little “purple” washable glue stick can also help give some extra stick for your models.

🟢 What’s the best build plate surface for PLA?

Textured PEI is one of the best all around surfaces, very affordable and generally works well for easy going PLA.

🟢 What’s the best build plate for PETG?

PETG is known to bond with just about anything, but a good clean, smooth PEI works best. And it never hurts to add a little glue stick for insurance.

🟢 What’s the best build plate for ABS?

We highly recommend the Tyson Build plate, a specialized high temperature plate that works for all filaments and absolutely excels at high temperatures.

MORE: Best 3D Printers

MORE: Best Resin 3D Printers

MORE: Best Budget 3D Printers

MORE: Best Filaments for 3D Printing

MORE: How to Buy the Right 3D Printer

Hive Systems is in its 5th year of benchmarking password cracking times on GPUs; the last update they made was in 2024, featuring the RTX 4090. Rather than cracking an actual password itself, the outlet cracks hashes, which are scrambled versions of reproducible text generated by software.

Hashes are a popular method for securing customer passwords against hackers by storing customer passwords as hashes on servers. When hackers go to steal a password database from a company's server that is secured with hashing, all they will see are hashes of the passwords they represent rather than the actual passwords themselves.

Cracking passwords with this method involves making a list of all possible combinations of characters to make a password, then matching them to a stolen database featuring password hashes. Graphics cards are particularly good at this methodology due to highly parallelized computing capabilities.

With 12 RTX 5090s, Hive Systems found Nvidia's latest flagship can crack the hash equivalent of eight-character passwords featuring numbers only in just 15 minutes. Passwords comprised of eight lower-case letters, the dozen RTX 5090s were able to crack in 3 weeks.

On the other end of the spectrum, passwords taking advantage of numbers, upper and lowercase letters would take 12 RTX 5090s 62 years to crack, and 164 years to crack with symbols added into the mix.

For a generation-on-generation perspective, a single RTX 5090 can crack a numerical-only password hash with eight characters in three hours. A single RTX 4090 can accomplish the same task in four hours, representing a 33% performance improvement for the RTX 5090 over the RTX 4090 in this specific scenario.

However, as the password complexity increases, the RTX 5090 becomes faster and faster in comparison. Moving up to a hash representing an eight-character password made up of numbers, upper and lowercase letters combined with symbols, the RTX 5090 is allegedly twice as fast as the RTX 4090. But cracking a password that complex would take the RTX 5090 a millennium to accomplish.

Hive Systems' research demonstrates the dangers of low complexity passwords and how adding just a bit of complexity in the form of symbols, capitalized letters, and numbers can make a huge difference for security. The good news is that cracking hashes in the way Hive Systems demonstrates requires hackers to have a stolen database of password hashes in the first place. Without this, hackers can't brute-force hack password hashes.

Baptized as the C86-5G, the upcoming server processor will feature up to 128 cores with simultaneous multithreading (SMT). However, there's a slight twist. Instead of utilizing the typical two-way SMT (SMT2) that we've grown accustomed to on modern mainstream and server chips, the C86-5G leverages four-way SMT (SMT4), which means four threads per core. That's why the C86-5G has 128 cores and 512 threads.

Manufacturers have long utilized various SMT layouts; Hygon is not introducing any innovative concepts. For instance, Intel's defunct Xeon Phi, particularly the Knights Landing variant, was equipped with an SMT4 architecture. Moreover, IBM's Power8 processors offer SMT4 capabilities and extend to SMT8.

Hygon C86-5G Specifications

The C86-5G, in contrast to its predecessor, the C86-4G, features up to twice as many cores and four times as many threads. Unfortunately, Hygon has not disclosed the specific microarchitecture utilized in the C86-5G, only mentioning that it has been "upgraded again." Consequently, the chipmaker claims that the C86-5G will significantly improve instructions per cycle (IPC) by more than 17%.

Hygon has come a long way since its first-generation server processor, which was built on the Dhyana microarchitecture derived from AMD's Zen IP. However, it seems to be a thing of the past, as Hygon's roadmap indicates that since the C86-4G model, the company has adopted a "new self-developed microarchitecture." Consequently, the C86-5G is said to employ an enhanced version of this mysterious microarchitecture.

The C86-5G has a rich feature set, including AVX-512 instructions and 16-channel DDR5-5600 support, up from the 12-channel DDR5-4800 configuration on the C86-4G. The former can substantially help the C86-5G power through server and enterprise workloads. Meanwhile, supporting 16 memory channels will allow Hygon's upcoming platform to house an enormous amount of memory. The manufacturer didn't specify which DIMM formats the C86-5G can handle, but being a server chip, it should support RDIMMs and such. But even with conventional DDR5 memory modules that have scaled up to 64GB per module, the C86-5G can have up to 1TB of DDR5-5600 at its disposal.

The earlier C86-4G model offers 128 PCIe 5.0 lanes for connectivity, matching the capacity of AMD's fourth-generation EPYC 7004 (codenamed Genoa) chips. Although the number of PCIe 5.0 lanes for the C86-5G is yet to be disclosed, Hygon has confirmed that the new flagship will support Compute Express Link 2.0 (CXL2.0). This positions the C86-5G competitively alongside the latest AMD EPYC 9005 (codenamed Turin) and Intel's 5th Gen Xeon (codenamed Emerald Rapids) processors in terms of interconnect standards.

The C86-4G has presumably been available since last year, suggesting that Hygon probably began development on the C86-5G already. However, since the roadmap is quite ambitious and lacks specific dates, we cannot determine the current status of the C86-5G.

Before the tariff conflict with the U.S., China was already pursuing technological self-sufficiency. The tariffs merely accelerated the nation's efforts to reach this goal, with companies like Hygon vital to the process. It's unrealistic to believe Hygon can launch a chip that rivals AMD's and Intel's performance levels. Nonetheless, the C86-5G seems to compare well with competitors in terms of features, at least. Considering the current landscape, we expect to see C86-5G benchmarks soon.

The company initially specialized in EDA, allowing companies to build chips more efficiently. Some of its customers include Intel, AMD, Apple, and Nvidia, and they use Cadence’s tools to help develop next-generation chips. However, it has also expanded in other industries; for example, GE and NASA use its systems for both computational fluid dynamics (CFD) and SDA. It has also entered the life sciences field, with Pfizer and AstraZeneca using its products for drug discovery and molecular modeling.

Cadence’s use of Nvidia systems, plus its deployment of AI tech within the M2000, allowed it to optimize the M2000 for efficiency. The company says that it can deliver this increased capability while simultaneously reducing the power demand by 20 times across various applications.

“The Millennium M2000 Supercomputer will drive the next leap in AI-accelerated engineering by leveraging our massively scalable solvers, dedicated Nvidia Blackwell-accelerated computing, and AI to help designers continue to push the limits of what is possible,” said Cadence CEO and President Anirudh Devgan. “Purpose-built for the most advanced AI models of today and tomorrow, the Millennium M2000 Supercomputer delivers unprecedented designer productivity to propel the next generation of AI infrastructure, physical AI systems, and drug discovery.”

Specialized AI has helped supercharge many industries, and we’ve seen many companies offer systems designed to run them at different price points. For example, the TinyBox AI accelerator, which uses six RX 7900XTX or RTX 4090 GPUs, starts at $15,000, making it useful for SMEs.

On the other end of the spectrum, we have massive AI data centers like Musk’s Memphis Supercluster, which is estimated to cost between $3 to $4 billion just for the hardware. The Millennium M2000 sits nicely in the middle of this price range, with each unit estimated to cost $2 million. Aside from no longer needing to contract data centers or hyperscalers for computing time, companies that acquire an M2000 can also ask Cadence to customize it to specifically fit their needs.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

While working at Huawei, the suspect allegedly took IP related to CMOS image sensors (CIS) and hybrid bonding chip packaging technology that is used for 3D NAND, HBM3/HBM3E/HBM4 memory, as well as advanced stacked multi-chiplet packages (e.g., TSMC's SoIC and Intel's Foveros 3D). It is unclear whether the suspect's targets were hybrid bonding and image sensors, or 'just' hybrid bonding as modern CMOS sensors also use hybrid bonding. 

SK Hynix reportedly clarified that the leaked bonding-related information concerned general wafer-to-wafer processes and not the specific hybrid bonding technique currently being developed or used for commercial products. 

For Chinese companies like Huawei (which does not have access to advanced process technologies and packaging methods used by TSMC and SK hynix) as well as SMIC (which does not have access to the latest lithography equipment made by ASML), getting wafer bonding know-how from SK hynix, which had licensed such technologies from Xperi, may be considered as a major breakthrough as wafer bonding is becoming increasingly important for multiple semiconductor applications.

The individual in question, a South Korean citizen named only in the complaint by his surname, Kim, previously worked for the Chinese unit of SK hynix. Authorities allege that in 2022 he secured a position with HiSilicon and, during the hiring process, improperly accessed and misused internal data belonging to SK hynix.

Kim reportedly printed out and captured photographs of internal files against policy. Investigators discovered that over 11,000 photographs of confidential content were taken, and steps were allegedly made to mask the origin by removing company identifiers such as brand logos and secrecy labels from the images. 

Evidence indicates that Kim submitted these materials as part of job applications to two separate companies in China, using them to enhance his qualifications. Authorities now claim this was a calculated effort to secure employment using unlawfully obtained information from his former employer. 

The report confirms that many technologies developed in China in general and by Huawei specifically originate from multi-national corporations. But there is an interesting twist: Huawei did not explicitly ask the suspect to bring it SK hynix's trade secrets (even though the company is known for poaching highly-skilled employees from world-class companies by offering them huge salaries), but accepted them as part of the hiring process. 

The case has drawn attention in South Korea due to rising concerns about the outflow of specialized semiconductor knowledge, specifically as China is trying hard to achieve semiconductor self-sufficiency. It highlights the increasing risks around workforce migration and the global effort to enforce tighter control over the transmission of high-value industrial IP. 

This is not the first time when an SK hynix employee left for Huawei and takes the company's IP with them. Last year the Suwon District Court sentenced one of SK hynix's employees who departed to Huawei with 4,000 secret documents to 18 months behind bars and a fine.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

In this how-to we’ll look at various ways of searching and re-using our command history. Whilst you become accustomed to these commands it’s important to double check you don’t unintentionally reissue a command that could cause problems. Take your time using these new techniques and double check the details before pressing enter!

All the commands in this how-to will work on most Linux machines. We’ve used an Ubuntu install but you could run this how-to on a Raspberry Pi. All of the how-to is performed via the Terminal. You can open a terminal window on most Linux machines by pressing ctrl, alt and t or by searching for the terminal app in your Applications menu.

Scrolling Through Your Previous Commands in Linux

The simplest way to look through your recent commands is to use the up and down arrow keys on your keyboard to scroll through the previous commands. If you want to reissue a found command simply press the enter key.

Viewing Your Command History in Linux

The history command, in its most basic use case, lists and annotates the last 1000 commands issued in the terminal emulator. Each command has a number associated with it.

1. Run the history command to see a list of the last 1000 commands. You’ll see that all the listed historical commands are given a unique reference number.

history

2. Reissue the history command but constrain the amount of results to a specific number. This is useful if you know roughly when a command you are looking for was issued. You should see only the last 20 results listed.

history 20

Reissue a Previous Command

Now we can use history to view our previous commands, we can choose and reissue a command using the number assigned to the history results.

1. Run history 20 to create a list of commands, choose a command to reissue making sure that the chosen command is safe to run. Choosing a simple command like cd Music (1660) is a good safe example. Note there is no space between the exclamation mark and the command number.

history 20
!1660

Enhanced Linux History Search using Grep

By piping the output of history into grep we can perform a search of our command history returning results for a specified term or string. This is an excellent approach for finding a partially remembered command.

1. Search for a specific term using history and grep. We used the example search term “silhouette” as we recalled we had issued some commands to rectify a problem with a silhouette vinyl cutter. Replace that search term with something suited to your machine.

history | grep silhouette

Using a Reverse Search of Linux Command History

Another handy approach to retrieve previous Linux commands is to use the reverse search function built into the terminal. To enter this mode you simply press ctrl and r. You can then enter a search term and use repeat presses of ctrl and r to step back through the list of previous commands containing that term. When you find a command you want to reissue press enter.

1. Press ctrl and r enters the reverse search mode, you should see the prompt now reads (reverse I search)`':

2. Type a search term and you should see the last command issued that contained this term. For example we added the search term sudo to show the previous commands issued with sudo privileges.

3. Repeat pressing ctrl and r to step through other results.

4. Run a previous command by pressing enter or quit the reverse search by pressing esc .

Quickly Reissuing the Previous Linux Command

Often we will want to simply rerun the last command we issued. We can achieve this simply using the !! command.

1. Run the ls command to set this as the example to test.

ls

2. Reissue the last command using !!. Note that the previous command is listed and performed.

!!

Sometimes we may try to reuse a command that requires elevated privileges, for example editing a file outside of our home directory. To do this we can preface the previous command with sudo. In the following example we append the first ls command to be reissued with sudo.

sudo !!

Hiding Your Commands from Linux History

There may come a time where you need to keep a command out of your history, and if that scenario ever occurs, all you need to do is preface the command with a single press of the spacebar.

For example here are two ls commands, the second has a single space, hiding it from the history file.

ls
 ls

With a little practice all the above approaches become quite instinctive to use and can make your terminal session more powerful and efficient. The ability to locate and reissue commands is extremely useful, especially when recovering a rarely used command or a command that was hard to create in the first instance.

Related Tutorials

• How To Check Disk Usage in Linux
• How To Kill a Process in Linux
• How To Find Files in Linux

According to Fastino, this was accomplished by developing architecture for their AI model that is "task-specific" and designed to function on a smaller scale. This change has made for a faster model that has a noticeable degree of accuracy. The best part, however, is the cost effective benefits of training their system on cheap gaming hardware.

Fastino may have less than $100K invested in GPUs but it has already started raking in money from investors that see the potential. Most recently, Fastino acquired $17.5 million from Khosla Ventures following $7 million raised late last year by both M12, Microsoft's venture capital fund, as well as Insight Partners.

Fastino isn't the only company to innovate on the AI hardware front with relatively affordable development options. Tiny Corp released what it calls TinyBox last year with six repurposed AMD Radeon 7900 XTX graphics cards specifically for AI processing.

So far, Fastino has yet to share any performance data, so we can't confirm exactly how well their new architecture stacks up against other AI systems. The team has released statements suggesting noteworthy performance, touting its ability to provide a complex answer within just milliseconds. Without more information, it's hard to tell how well it will perform against other big names in the AI field.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Check out Newegg's PC Gaming Week sale and pick up this CPU/SSD combo of AMD's Ryzen 5 9600X processor and a Patriot P400 Lite 500GB SSD for just $208. The list price of the 9600X shows as $279, so you're saving $71 off of that price and getting a free SSD.

AMD's Ryzen 5 9600X is built around the Zen 5 architecture, and has six cores and 12 threads with a base clock of 3.9 GHz. When boosted, it can reach speeds as high as 5.4 GHz. The 9600X also features integrated graphics, so if you're building a non-gaming rig or just want to do exceptionally light gaming, this processor does away with the need for a separate graphics card.

The free Patriot P400 Lite 500GB SSD is a PCIe 4x4 M.2 SSD with sequential read and write speeds of 3,500MB/s and 2,700MB/s, respectively. This SSD drive alone retails for $41 at Newegg, and although it has a low capacity for today's uses, it's still large enough to help top up a pre-existing system's total capacity.

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

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Nvidia typically optimized pre-release drivers as part of the review kit. Nvidia typically releases the public-facing enablement drivers for a new GPU on the day of its retail launch. If proper drivers aren't installed, the system may either fail to detect the GPU or the results obtained from benchmarking the card will be unreliable at best.

Nvidia didn't send RTX 5060 Ti 8GB models to reviewers, leading them to acquire retail units at their own expense. As expected, that GPU struggles at 4K, at 1440p, and even at 1080p in a handful of titles compared to its 16GB sibling. The RTX 5060 and its 8GB may face similar woes, but that's at least somewhat counteracted by its cheaper price point. It's conceivable that internal software issues may be causing driver delays, if you want to give Nvidia the benefit of the doubt.

The RTX 5060 is set to hit shelves on May 19th, exactly one day before Computex kicks off. This leaves reviewers with virtually no time for independent testing, so day-one impressions will largely be based on Nvidia's first-party metrics and the few reviewers who aren't traveling.

Ironically, potential RTX 5060 supply constraints at launch might be a blessing in disguise this time around, giving buyers just enough time to see independent reviews before jumping the gun.

On paper, the RTX 5060 offers 25% more CUDA cores (3,840) than the RTX 4060, suggesting a noticeable performance uplift. At its suggested $299 MSRP, the RTX 5060 might be a decent pick as long as you're aware of its limitations. Further down the stack, Nvidia is reportedly preparing an RTX 5050, but based on early specifications, don't expect it to perform any miracles. At the same time, rumors allege AMD is also scheduling its budget RX 9060 XT series for Computex, reigniting the company's rivalry on the low end of the hardware stack.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

The operating system has been in development since 2012, and it first hit Huawei devices in 2019 in the Honor Vision smart TV. By 2021, Huawei started deploying it in its smartphones, with older models getting the homegrown operating system, too. Most Huawei laptops come with Windows 11, although the company has also offered some models that use Linux, which is open source.

But relying on its own operating system allows it to create an ecosystem of devices. This is especially true as it already has several other devices running HarmonyOS, from smart TVs and laptops to tablets and smartphones, running the same software. This will make it easier for the company to market its devices as one seamless platform, similar to how Apple does with iOS, iPadOS, and macOS.

HarmonyOS-powered Huawei smartphones have already overtaken Apple in China, with the former achieving 19% market share over the latter’s 17% last quarter. This will likely make it more enticing for Chinese customers to switch to HarmonyOS for their desktop or laptop computing needs.

Apps for the HarmonyOS laptop include WPS Office, a free substitute to the Microsoft Office suite, and DingTalk, an enterprise collaboration system, which reportedly works on the laptop. The PC is also expected to natively run the thousands of mobile apps that are already available for HarmonyOS smartphones.

We don't yet know the specs for this HarmonyOS-powered Huawei computer. But if it still uses an Intel or AMD chip like other laptops in its lineup, there’s a chance that users who need Windows can install it after purchase. They might even be able to dual-boot it, allowing them to stick to temporarily use Windows for the apps they need but aren’t available yet on HarmonyOS.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

We are going to look at a number of different ways to get RAM and CPU data in the terminal, before we finally look at two applications which can provide a basic level of assurance, at a glance.

These commands will work on most Linux machines. Our test PC is running Kubuntu, the KDE flavor of Ubuntu, but we also certain that you can also run through this how-to on a Raspberry Pi or any other Linux device. All of the how-to is performed via the Terminal.

How to Check Your CPU in Linux

1. Open a terminal.

2. Use the cat command to display the data held in /proc/cpuinfo.

cat /proc/cpuinfo

This command will produce a lot of text, typically it will repeat the same information for the number of cores present in your CPU.

A more concise means to get most of this information is via lscpu, a command that lists the CPU details.

1. Open a terminal.

2. Use lscpu to display the CPU details. The command is quite verbose and we can easily see the number of CPU cores, minimum and maximum CPU speed and the CPU architecture.

lscpu

Using a little grep magic we can pull out just the information that we need.

1. Open a terminal.

2. Using a vertical pipe, send the output of the lscpu command to grep and search for “max”. This will give us the maximum possible CPU speed. Pipes are a way to send the output of one command as the input for another. Classic examples are piping the output of a command to a text file for later review.

lscpu | grep max

Another means to get CPU information is using dmidecode, a command that dumps the DMI (SMBIOS) contents into something we can understand.

1. Open a terminal.

2. Type in the dmidecode command using sudo, and the argument -t 4. There are many DMI types, with a numerical reference used to pull information for that component. In this case -t 4 is for the CPU. We can replace the 4 with processor for the same effect.

sudo dmidecode -t 4

How to Check Your RAM in Linux

1. Open a terminal window.

2. Use the free command. This will show the available memory, and how the memory has been allocated, in Kilobytes.

free

3. Use the -m switch to show the available memory in Megabytes, or -g for Gigabytes.

free -m

Alternatively use the -h switch to show the memory information scaled to the three shortest digits. Giving us at a glance information.

free -h

There is a similar way to /proc/cpuinfo for memory information.

1. Open a terminal window.

2. Use the cat command to print the contents of /proc/meminfo. The output is incredibly verbose and can prove useful for debug. In general use we would stick with free -m.

cat /proc/meminfo

If we need to know the timings and breakdown of memory modules in a machine then we can use lshw.

1. Open a terminal window.

2. Using sudo, issue the lshw command with the -short switch (the device tree) and -C with the class memory.

sudo lshw -short -C memory

We can also use dmidecode for a more detailed query.

sudo dmidecode -t memory

Information at a Glance

Should we need a more general level of information, our current CPU speed, RAM usage, network bandwidth then we have two commands which can be installed.

How to Install and Use Htop

The standard top command is great, but htop is a much better alternative. Htop provides us with an interactive list of running processes.

To install htop.

1. Open a terminal window.

2. Update your list of repositories.

sudo apt update

3. Install htop using apt.

sudo apt install htop

4. Run htop.

htop

At the top of the window we can see the utilization of our CPU cores, under that is our RAM, and finally is the swap.

We can scroll through the list with our cursor keys or search using F3 followed by the name of a process or application. Here we have searched for the GNU Image Manipulation Program (GIMP). We can kill the app by pressing F9.

To close htop, press F10.

How to Install and Use Bpytop

Our personal preference is bpytop, a Python implementation of htop but with much more to offer.

To install bpytop.

1. Open a terminal window.

2. Install bpytop using the Python package manager pip.

sudo pip3 install bpytop

3. Run bpytop from the terminal.

bpytop

Bpytop’s interface is split into a series of areas.

1. CPU Status
CPU Speed
Core utilization
Overall utilization

2. RAM Usage
Total RAM
Used RAM
Available RAM
Cache

3. Disk / Storage Usage
Used and free space on all mounted devices

4. Network Interface Usage
Up and down speeds

5. Processes (just like htop)
Process management

We can show or hide an area by pressing the corresponding number (1..4). If we just wanted CPU, RAM, storage and Network information then we press 4. To bring it back, press 4 again.

To filter the processes for a particular application.

1. Press f to filter. It has to be lower case f.

2. Type in the application / process name / Pid. The search will adapt to show the process, in our case GIMP.

3. Press T to terminate the process, or K to kill. These have to be upper case T and K.

To exit bpytop press q at any time.

"The Biden AI rule is overly complex, overly bureaucratic, and would stymie American innovation," a statement by the Commerce Department’s Bureau of Industry and Security published by Bloomberg reads. "We will be replacing it with a much simpler rule that unleashes American innovation and ensures American AI dominance."

The regulation in question, known as the AI diffusion rule, was set to take effect on May 15 and would have categorized foreign nations into three groups for licensing purposes. Under this system, only the U.S. and 18 allied 'Tier 1' countries have full access to processors like Nvidia's H100. Over 100 'Tier 2' nations would face a cap of about 50,000 H100-class GPUs annually over the next few years, unless they obtain verified end user (VEU) status. Note that entities in Tier 2 nations may still import up to 1,700 advanced AI processors per year without a license, which would not count toward the cap. 'Tier 3' countries — such as China, Russia, and Macau — would be barred from getting advanced processors entirely due to arms embargoes.

The new bilateral approach

Trump officials, however, will not implement the above framework. Instead, they are crafting a new approach focused on bilateral arrangements with countries such as Saudi Arabia and the United Arab Emirates. To get advanced GPUs, such countries may have to invest in the U.S., or broaden diplomatic cooperation, according to the report. For example, nations like the UAE have advocated for looser rules and have committed to long-term U.S. technology investment, pledging as much as $1.4 trillion over the next 10 years. Until recently, the UAE had to obtain an export license from the U.S. government to obtain advanced AI processors, such as Nvidia's H100 and H200.

While the Trump team seeks to revise global export rules for AI processors, it has reinforced restrictions against China's access to advanced AI GPUs designed in the U.S., including AMD's Instinct MI308 and Nvidia's H20. As a result, the two companies had to write down billions worth of AI processors in the most recent quarter. It is unlikely that the U.S. government lifts the ban on these processors, but Nvidia is already celebrating the fact that the AI Diffusion Rule is gone. As a result, it can continue shipping its AI and HPC GPUs to Europe without any restrictions.

"We welcome the Administration's leadership and new direction on AI policy," a statement by Nvidia reads. "With the AI Diffusion Rule revoked, America will have a once-in-a-generation opportunity to lead the next industrial revolution and create high-paying U.S. jobs, build new U.S.-supplied infrastructure, and alleviate the trade deficit."

Nvidia has argued that controlling sales to non-Chinese nations could alienate allies and drive them to China's Huawei AI ecosystem, thus threatening U.S. leadership in the AI sector in the long term. The change may also benefit U.S. firms with global infrastructure ambitions. For example, Oracle's expansion plans in Malaysia faced potential obstacles under the prior rule due to shipment caps. However, keeping in mind the fact that the Trump administration has a strict stance against China, it is possible that it may curb sales of advanced GPUs to countries used to smuggle Nvidia's H100 processors and servers to China, which includes Malaysia, Thailand, and Singapore.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

From our understanding of the details shared regarding Baidu's patent, it will rely on a multitude of inputs to deliver a human language rendering of an animal's current emotion. So, you aren't going to be using this AI tool for anything much more than recognizing whether your pet is hungry, or has some other pressing need. You weren't expecting to chat with your Burmese feline about geopolitics, were you?

It will be interesting to see any eventual applications based upon Baidu's work. We'd expect to be able to enjoy a wider gamut of 'conversation' with one of the intelligent dog breeds than a feline, caged bird, or fish friend. Of course, the machine learning will have to be rejigged for each species.

Previous and current smart device apps with touted 'animal translation' ability are said to be quite a popular phenomenon. The SCMP notes that a Japanese dog translation device (not app) was crowned one of the 'best inventions of 2002' by Time Magazine. However, some more recent and more sophisticated attempts at animal translators have fallen by the wayside.

A quick look on my Android mobile for animal translation apps reveals that there are a multitude of these 'quack' apps (pun intended) currently available. Several of these apps, usually targeting either dogs or cats, have enjoyed tens of millions of downloads. However, they are almost universally badly rated (no surprise). If someone got this right, even with a little accuracy, we are sure the app(s) would be very lucrative.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

Asus looks to fill that desire with its ROG Swift PG27UCDM. This QD-OLED panel will come up in the filter set I just mentioned and deliver a full list of features, premium build quality, terrific performance, and addictive gaming. Let’s take a look.

Asus ROG Swift PG27UCDM Specs

The PG27UCDM is every bit a flagship display, and its price of $1,200 reflects that. You can spend that same money on a 32-inch 240 Hz OLED monitor from Dough, Acer, and others. But those screens won’t give you the highest pixel density available in the mainstream, 163ppi. That’s a level of sharpness you can clearly see. The PG27UCDM makes the most of that with a Quantum Dot layer that covers over 103% of DCI-P3. Accuracy is excellent right out of the box with no need for calibration and a pro-level sRGB mode.

The panel is capable of 1,000-nit peaks in HDR mode when measuring a 3% window pattern. I got 465 nits from a 25% window, which tracks with the competition. The PG27UCDM is plenty bright, and you can see that same output in SDR mode if you turn off the Constant Brightness option. That adds another option for picture enhancement that isn’t universally available. One thing you won’t find in more than a handful of monitors is Dolby Vision. While that codec is widely supported by televisions, gaming screens rarely include it. Its dynamic tone mapping delivers a superior HDR image to HDR10’s static metadata.

Video processing starts with a 240 Hz refresh rate that can be enjoyed over any of the four video inputs. The DisplayPort is version 2.1 with 80 Gbps bandwidth, and the two HDMI 2.1 ports can move 48 Gbps. The USB-C port covers the same DisplayPort functions and adds 90 watts of power. You also get KVM with one upstream and three downstream USB 3.2 ports. And a 3.5mm headphone jack delivers audio. There are no internal speakers. ELMB is included, which operates a black frame insertion routine up to 120 Hz. And Asus has included its anti-flicker tech that maintains consistent brightness during frame rate transitions.

Peripheral features include Asus’ full array of LED lighting with Aura RGB and Light In Motion displayed by a large ROG logo in the back. The show can be synced with on-screen action if you like. Panel longevity is supported by multiple OLED care options like logo protection and pixel refresh. Extensive cooling via heatsink is also provided. Gaming aids run the full array with aiming points, sniper mode, timer, stopwatch, frame counter, and multi-display alignment marks. You can control all the fun through an extensive OSD, or with Asus’ DisplayWidget Center app

Assembly and Accessories

Asus is transitioning to fully recyclable packaging, and the PG27UCDM is from the new generation with molded pulp forms protecting the contents. The base and upright mate with a captive bolt, and the ROG log projector lens attaches with magnets. The panel snaps on to create a solid package. An included adapter provides a 100mm VESA mount if you’d rather use an arm. The accessory bundle comes in a nice zippered pouch and includes USB cables, DisplayPort, HDMI, and IEC for the internal power supply.

Product 360

The PG27UCDM’s front view is minimalist with thin flush borders around the panel and no graphics except a backlit ROG logo on a small protrusion at the bottom center. It also contains a proximity sensor to dim the screen when no user is present as part of the OLED care option package. The front layer is shiny like most OLEDs, with an optical coating to cut glare. It’s reasonably effective, but some thought should be given to optimal placement by avoiding sunny windows and bright light sources.

The stand features classic ROG styling with a cast metal base and a stout upright with a cable hole in the middle. A logo projector shines onto the desktop at the bottom. It’s part of the lighting system and can show in multiple colors and effects. More lighting adorns the base and back, where you’ll find a large, pixelated ROG logo that reminds me a bit of the Lite Brite toy from the 60s and 70s where you’d stick colored pegs into a light box to make pictures. Ah, the nostalgia! Ergonomics include 5/20 degrees tilt, 30 degrees swivel, 4.4 inches of height range, and a 90-degree portrait mode. Movements are smooth and solid as I’d expect from a high-end display.

Underneath you’ll find a well-stocked input panel with a DisplayPort 2.1 (80Gbps), two HDMI 2.1 (48Gbps), a USB-C (DP functions with 90 watts of power), a 3.5mm headphone jack and USB 3.2 ports, one upstream and three down. You can also see the OSD joystick and two control keys, which toggle power and exit the menu.

OSD Features

The PG27UCDM’s OSD is extensive and includes everything imaginable to control gaming, image parameters, KVM and more. You can also download Asus’ DisplayWidget Center app, which mimics the OSD and adds firmware updates and app sync.

GamePlus delivers a full set of gaming aids with multiple aiming points, sniper mode, timers, stopwatch, FPS counter and display alignment marks. GameVisual is Asus’ term for picture modes, and you get nine of them aimed at specific tasks or game types. The default is Racing, and it is spot-on accurate for color and grayscale with a slightly light gamma. It can be calibrated to an even higher standard with fixed color temps, RGB sliders, gamma presets and gamut options.

The only video processing options are an Adaptive-Sync toggle and ELMB. Extra Low Motion Blur is the same as backlight strobing for an LCD, but in OLED form, it’s black frame insertion. At speeds above 150fps, it’s unnecessary because, like all fast OLEDs, there is no motion blur to remove. It’s useful at lower frame rates though and that’s a good thing because it’s limited to a maximum of 120 Hz.

The PG27UCDM includes both uniform and variable brightness options for both SDR and HDR. Using the latter punches up contrast noticeably, but can be harsh for workday tasks. The uniform setting keeps the light level constant regardless of content. A super cool touch is that the brightness slider setting stays where you set it when switching Constant Brightness on and off. That way, you can balance output without readjusting every time. This is something all OLEDs should have. Asus is the only company I know of that includes Anti-Flicker. It’s a subtle enhancement that appears during rapid changes in frame rate by keeping brightness consistent.

HDR10 and Dolby Vision are fully supported here, and the latter is a rare thing indeed. There aren’t many games that use it, but it is common in video, especially from Netflix. This makes the PG27UCDM great for content creators. HDR content enables four new picture modes with a toggle for brightness and contrast adjustment. You can also tweak grayscale if you like.

OLED care options are extensive and include screensavers, pixel cleaning, orbiter, logo and edge dimming and a proximity sensor. When activated, the screen dims if you aren’t sitting in front of the PG27UCDM. At the bottom, you can track panel hours and automatically run a refresh cycle every eight hours.

If you want to view two video sources at once, there are options for PBP and PIP with independent color settings. Lighting is controlled by Aura RGB and Light In Motion options where you can pick from different colors and effects, and sync the light show to what’s happening on the screen. The four joystick directionals can be programmed to a variety of quick access functions like picture modes and gaming aids. In the System menu is the KVM feature which gives you many options for configuring the USB hub and video/peripheral bindings. Finally, if you need backward compatibility for the DisplayPort, it can be set to version 1.4 or 1.2.

If you prefer to control the PG27UCDM from the Windows desktop, Asus offers DisplayWidget Center. It mimics all OSD functions and adds some extras like firmware updates and app sync. This lets you bind games or other applications to monitor settings for automatic mode switching when the appropriate program is run. This is something you won’t find anywhere else.

Asus ROG Swift PG27UCDM Calibration Settings

The PG27UCDM has many calibration options and picture modes but the only one you really need is Racing. It can be used sans calibration out of the box, but I noticed that gamma was a tad light. I selected the User color temp and adjusted a very precise set of RGB sliders. This not only reduced color errors, but it also tightened up gamma tracking to perfection. The choice between variable and constant brightness is made easier by the fact that brightness settings made in each mode are independent. That allows one to equalize picture brightness without readjusting the slider each time, very cool. My settings below reflect both configurations.

You only need to pick between Dolby Vision and HDR10 in HDR mode. For HDR10, there are four picture modes, of which Gaming is the default and best choice. A True Black mode is included, but it measured almost identically to Gaming. HDR color was spot-on from my sample, but you can tweak RGB and gamma if you like. You can also engage a toggle to allow brightness adjustments.

Gaming and Hands-on

While it’s easy to be seduced by large and colorful monitors, I’m talking about 32 inches or more, there is something about high pixel density that is equally enticing. In the mainstream, there isn’t anything with more density than a 27-inch 4K panel and its sharpness and clarity is something you can see in a side-by-side comparison.

Though Doom Eternal is a frenetic fragfest, there is a lot of fine detail and texture in its virtual world that should be paused and enjoyed when playing on a monitor like the PG27UCDM. Compared to other OLEDs, it’s equally colorful and boasts equal contrast. But the 163ppi is something you won’t see on a 32-inch screen. I stopped many times, when I wasn’t being attacked, to appreciate what the game’s creators had done, especially on surfaces. Reflective metal looked particularly realistic with different levels of shine that depended on color and texture. Little chips and scratches were rendered clearly and the word “tactile” came to mind on many occasions. This is a case where size does not matter. Spend five minutes playing your favorite game on the PG27UCDM and you won’t wish for a larger screen.

Video processing is OLED-excellent with perfect motion resolution at all frame rates above 150fps. My test system kept the action above 220 and I never saw the slightest hesitation or blur. There are a few QHD monitors out there with less input lag but the PG27UCDM never failed to deliver on my wishes. I can’t imagine anyone not thinking it’s fast.

To test its Dolby Vision capability, I hooked up an Apple TV 4K. To enable the codec, I had to select the format in the OSD, then test it from the source. It worked without any issues when playing Netflix shows like Formula 1 Drive To Survive and Car Masters Rust To Riches. I also noted that it correctly matched frame rates with content mastered in 24p and 50p. The PG27UCDM is a great personal TV although since there are no internal speakers, you’ll need a good pair of headphones to enjoy the show.

I had installed Asus’ DisplayWidget Center during a previous review and noted that when I ran it, the PG27UCDM was instantly detected along with the settings I had already made in the OSD. It’s a super convenient way to make adjustments and I appreciated the added function of app binding where you can change picture modes automatically when running specific games or programs, very cool.

For productivity, there is nothing better than 4K when viewing fine photo detail or writing prose using ornate fonts. Pixel density is a clear asset for workday tasks and though my vision isn’t fighter pilot worthy, I had no trouble making everything out. That’s another instance where I didn’t notice the 27-inch size even though I’m accustomed to a 32-inch screen.

Takeaway: During my time with the PG27UCDM, I could find nothing to complain about. It serves incredibly well for gaming, entertainment and work and makes those activities a pleasure. Gameplay has the addictive quality I’ve come to expect from any fast OLED and that high pixel density is something you won’t regret paying extra for.

MORE: Best Gaming Monitors

MORE: How We Test PC Monitors

MORE: How to Buy a PC Monitor

The PG27UCDM is my first look at a 27-inch 4K OLED. The other screens in the group are 32-inches but all run up to 240 Hz. They are Asus’ PG32UCDP, Aorus’ FO32U2P, Acer’s X32 X3, Dough’s Spectrum Black 32, and HP’s Omen 32 OLED.

Pixel Response and Input Lag

Click here to read up on our pixel response and input lag testing procedures.

240 Hz OLEDs, regardless of pixel count or density, refresh the screen in 4ms. They are on equal footing regarding motion blur, of which there is none. The PG27UCDM includes ELMB (black frame insertion) for lower speeds up to 120 Hz. But no extra processing is required for blur-free operation over 150fps.

In the lag test, the PG27UCDM proved to be the fastest 4K OLED I’ve yet tested, tied with the HP Omen 32 OLED at 18ms. This is very quick but there are a few QHD screens that are faster. The overall record still belongs to the 360 Hz Philips 27M2N8500 at 10ms. The PG27UCDM is more than fast enough for any gamer though, even at the pro level. Its smoothness is a definite asset in competition.

Test Takeaway: The PG27UCDM delivers the same blur-free operation as any OLED running over 150 Hz. It has less input lag than all other 4K OLEDs except the HP Omen 32 OLED, which it tied in my test. It adds flexibility with ELMB up to 120 Hz which makes it better suited for slower PCs. If you have the processing power, it is on equal footing with the very best and provides an addictive gaming experience.

Viewing Angles

The PG27UCDM has perfect viewing angles, meaning there is no visible color or brightness shift at 45 degrees to the sides. Some QD-OLEDS show a slight warmth, but Asus is using the latest panel part here, which does not have that artifact. The top view is a bit red with lighter gamma. This is excellent performance.

Screen Uniformity

To learn how we measure screen uniformity, click here.

OLEDs are consistently solid in my uniformity test, and the PG27UCDM is one of the better panels I’ve seen. That is as it should be for a $1,200 monitor. With just 5.01% measured deviation, it has no visible hot spots or other anomalies. All field patterns are perfectly uniform from edge to edge.

MORE: Best Gaming Monitors

MORE: How We Test PC Monitors

MORE: How to Buy a PC Monitor

To read about our monitor tests in-depth, please check out Display Testing Explained: How We Test PC Monitors. We cover brightness and contrast testing on page two.

Uncalibrated – Maximum Backlight Level

Not all OLEDs have variable brightness available for SDR content, but the PG27UCDM does. I measured a 25% window and got 445 nits for the win. This isn’t a big deal because in practice, 250 nits is plenty for an indoor setting. And for productivity, constant brightness is easier on the eyes. However, variable brightness is great for gaming, video, and HDR. We’ll talk more about that on page five.

After Calibration to 200 nits

Calibration doesn’t change either black levels or contrast; they remain unmeasurable. For SDR content viewing and testing, I left the PG27UCDM on constant brightness for consistent measurements and for more comfortable productivity. I appreciated the ability to set separate brightness values for each mode, meaning I didn’t have to readjust when switching back and forth. Variable brightness is great for movies and video content in SDR or HDR modes. The ANSI test is also unmeasurable whether constant brightness is on or off.

Test Takeaway: The PG27UCDM has the same peak brightness, infinite black levels, and high contrast as its OLED competitors. It adds a variable brightness option for SDR, which sets it apart from some screens like the HP Omen 32 OLED and Aorus FO32U2P. Ultimately, any level over 250 nits is enough for indoor environments.

MORE: Best Gaming Monitors

MORE: How We Test PC Monitors

MORE: How to Buy a PC Monitor

The PG27UCDM includes factory calibration data, and my tests confirm its out-of-box accuracy. In the default Racing mode, it does not need calibration. It also has an accurate sRGB mode and plenty of image controls for those who want or need to tweak.

Grayscale and Gamma Tracking

Our grayscale and gamma tests use Calman calibration software from Portrait Displays. We describe our grayscale and gamma tests in detail here.

The only flaw in the PG27UCDM’s default chart is light gamma tracking. Grayscale is visually perfect with all values below 2dE. But gamma is a bit under the reference line with a 2.11 average. This is hard to spot in content, but with so much contrast available, a darker value will add texture to the image.

I calibrated the user color temp with its precise RGB sliders and achieved better grayscale tracking and improved gamma. It is now exactly on the line with a 2.2 average. It can’t get better than that.

The PG27UCDM has two ways to select sRGB. You can choose the sRGB Cal mode from GameVisual or retain calibration ability by simply sticking with Racing mode and picking the sRGB option from the Display Color Space sub-menu. The result is visually perfect grayscale and decent gamma that’s marred only by light values from 70-90%. This is still well within tolerance which makes the PG27UCDM qualified for color critical applications.

Comparisons

The PG27UCDM’s 1.23dE default grayscale is certainly impressive, but so is that of the other screens. None of them require calibration. These monitors are expensive. but their performance and accuracy justify the price. Calibration makes slight improvements across the board but visually, the differences are invisible.

The PG27UCDM manages better gamma results than all but the Dough, which has a slightly tighter value range. Errors are well below the visible point. All the monitors are super close except the Aorus, which runs a bit lighter than the rest.

Color Gamut Accuracy

Our color gamut and volume testing use Portrait Displays’ Calman software. For details on our color gamut testing and volume calculations, click here.

The PG27UCDM measures like other QD OLEDs with full coverage of the DCI-P3 gamut, plus a little bonus red. There is slight oversaturation, which is easy to spot in SDR content but nearly impossible to see in HDR mode. Hue and saturation points are all on or close to their targets. With a 1.60dE score out of the box, the PG27UCDM is visually free of errors. Calibration tightens up hue tracking a bit, but the visual change is minuscule.

The sRGB gamut option is also visually perfect with no points outside the target box. The PG27UCDM is fully qualified for color grading to DCI-P3 and sRGB specifications.

Comparisons

It would be easy to say it doesn’t get much better than the PG27UCDM’s 1.28dE color score if there weren’t four other monitors that measured lower. This is impressive performance, and it speaks to the consistent quality of the 4K OLED category, in which I have yet to see an underachiever.

In the volume test, it’s easy to spot the Quantum Dot panels. They’re the ones with over 100% coverage, like the PG27UCDM’s 103.27% score. The Aorus and HP screens have slightly greater volume, but this is a small difference in practice. You also get near-ideal sRGB coverage of 97.89%. If you’re looking for maximum OLED color, there are a few QHD screens that deliver around 110%, like the recently reviewed Samsung OLED G6 at 112%.

Test Takeaway: The PG27UCDM makes good use of its Quantum Dot panel with some bonus red, but its accuracy is what makes it a compelling choice. It doesn’t need calibration, and it is spot on for both sRGB and DCI-P3 mastered content. With extensive adjustment options, it’s well suited for tweakers. Color-wise, it is essentially flawless.

MORE: Best Gaming Monitors

MORE: How We Test PC Monitors

MORE: How to Buy a PC Monitor

Our HDR benchmarking uses Portrait Displays’ Calman software. To learn about our HDR testing, see our breakdown of how we test PC monitors.

The PG27UCDM is primed for HDR content with HDR10 and Dolby Vision support. There are multiple picture modes and many possible adjustments that few HDR monitors offer. The switch is automatic when HDR content is present.

HDR Brightness and Contrast

Asus claims 1,000 nits peak for a 3% window, and I have no reason to doubt this given the 465-nit measurement I got from a 25% window. This is with variable brightness engaged and brightness maxed. If you find this too harsh, you can switch on the brightness slider and dial things down to your preference. I measured Gaming HDR and True Black and saw no difference between them. The Spectrum Black 32 and PG32UCDP push their panels a bit harder, but peak brightness is still around 1,000 nits for a 3% window, so they aren’t brighter than the PG27UCDM in practice. All the panels here will look about the same when showing HDR10 content. Of course, if it's Dolby Vision you’re after, only the PG27UCDM will show that.

Grayscale, EOTF and Color

Though you can calibrate the PG27UCDM in HDR mode, there’s no need. My test shows visually perfect grayscale and EOTF tracking. I measured both Gaming and True Black modes and got the same results. In the gamut tests, hue points are perfectly in line with a general and slight oversaturation in all colors. This is about as close to perfect as I’ve seen from any HDR display. The BT.2020 test shows similar behavior until the PG27UCDM runs out of color at 90% red, 75% green, and 90% blue. This is excellent performance.

Test Takeaway: The PG27UCDM is a reference-level HDR display with high brightness and superlative accuracy. It’s on par with competitors in performance but stands above with its extra adjustability. The HDR image, like any OLED, is stunning.

MORE: Best Gaming Monitors

MORE: How We Test PC Monitors

MORE: How to Buy a PC Monitor

I have yet to review an OLED gaming monitor I didn’t love. They are incredibly consistent in all the important areas: image fidelity, color accuracy, video processing, build quality, even styling is a cut above the rank and file. I always look for some standout feature, and though it is usually small, most displays have something that sets them apart.

In the case of Asus’ ROG Swift PG27UCDM, there are two things to recommend: high pixel density and Dolby Vision. The first one is easy, put 4K on a 27-inch screen and you get 163ppi. Add in OLED contrast, QD-OLED color and out-of-box accuracy and you get a fantastic image that makes you forget about jumbo monitors.

Dolby Vision is the big one here. I used to ding HDR monitors for leaving it out until I noticed that no one was including it. But it’s also something I’ve become spoiled by when watching my LG OLED TV. Dynamic tone mapping sets it above HDR10 enough that you can easily spot the difference. Though it isn’t supported by many games, it is common on 4K Blu-ray and streaming services like Netflix and Disney+. The PG27UCDM then becomes a premium personal television, just add headphones.

For gaming, it is on par with the best OLEDs I’ve reviewed. Its 18ms total lag score ties it with the HP Omen 32 OLED for fastest 4K monitor. It is lightning quick with perfect video processing. And if you happen to pair it with a less capable system, ELMB keeps motion blur at bay below 120 Hz.

Build quality and function are also at a high level. Like any ROG product, the PG27UCDM is a flagship. You get extra cooling, which ensures long life and the backing of a three-year warranty. Asus’ DisplayWidget Center is another value-add. And supreme tweakability means you can make it to your liking in every way.

The Asus ROG Swift PG27UCDM is a premium purchase to be sure but it truly embraces the mantra, “you never regret buying the best.” If you have the budget and want one of the best OLED gaming monitors available, definitely check it out.

MORE: Best Gaming Monitors

MORE: How We Test PC Monitors

MORE: How to Buy a PC Monitor

In time for PAX East, Alienware is launching the Alienware 16 Aurora and Alienware 16X Aurora, systems that blend understated (well, for Alienware) designs with gaming power. It feels like the successor to last year's Alienware m16 R2, which dialed down the gaming looks and added a "stealth mode" for quiet performance and plain lighting – and that feature is coming back with the Auroras.

Both Auroras come in a dark blue "interstellar indigo" chassis that features the alien head logo on the lid, but little else that screams gaming notebook. They have curved edges designed to be comfortable to pick up and carry, and can also be opened with one hand.

Alienware has made these gaming laptops smaller than its classic designs by eliminating the "thermal shelf," a cooling solution that typically juts out the back of the laptop. Instead, there's a cooler on the bottom of the device, where a rear foot would usually be found. This sucks cool air in from the bottom of the laptop and exhausts it from the back and sides.

The 16X Aurora is a more premium version of the system, with higher-end components and parts. It goes up to an Intel Core Ultra 9 275HX, Nvidia GeForce RTX 5070, 64GB of RAM, 2TB display, and a 240 Hz, 300-nit screen. Additionally, it has a higher-resolution webcam than the standard 16, and has a metal lid and bottom (as opposed to just the metal lid on the 16 Aurora).

There are other quality differences: The 16X has single-zone RGB, while the regular 16 just has white backlighting. Additionally, one of the 16X's USB Type-C ports supports Thunderbolt 4, while that's not the case on the 16.

The new systems will start at $1,149 in the US. For the regular Alienware 16 Aurora, that will get you an Intel Core i7 14700H (Raptor Lake Refresh), 16GB of RAM, 1TB SSD, Nvidia RTX 4050, and a QHD, 120 Hz display.

At launch, the Alienware 16X Aurora will start at $1,949, though a cheaper base model will launch "soon after," an Alienware rep tells me. The $1,949 system has a Core Ultra 9 275HX, 32GB of RAM, 2TB of storage, an RTX 5060, and a QHD, 240 Hz screen.

The Aurora laptops follow the launch of the new Area-51 machines announced earlier this year and recently released, and are the first Aurora-branded notebooks from the company in almost 20 years. Alienware's revival of its Aurora and Area-51 laptop sub-brands come as its parent company, Dell, attempts to simplify its naming schemes.

Intel's Battlemage family currently employs one die, BMG-G21, powering both the Arc B570 and Arc B580 GPUs. Rumors have alluded to several other configurations, with the most persistent being a larger BMG-G31 die, which has surfaced numerous times in shipping manifests. The majority of these shipments are destined for Vietnam, which is home to several OSAT companies that assemble, package, and test silicon. Export data alone isn't enough to confirm whether a BMG-G31-based SKU is still being developed, so we're treating this with significant skepticism.

Rumors of Battlemage-based professional cards with 24GB of VRAM have been afloat since last December. Likewise, a recent slip-up from an OEM indicated that Intel had a 24GB B580 model in the pipeline. If we put two and two together, at least one GPU from the Battlemage professional lineup is expected to employ the BMG-G21 die with twelve 16Gb GDDR6 modules arranged in a clamshell configuration. Other specifications should remain largely similar to the Arc B580.

It's important to remember that Intel's BMG-G21 die is still positioned as a lower-end offering, designed to compete in the same segment as Nvidia's RTX 4060 (AD107). With just 20 Xe cores (2,560 shaders), cards based on this design are unlikely to match the RTX PRO 4000 Blackwell, at least in compute, with a similar 24GB configuration but a much more powerful GB203 chip under the hood. But for purely VRAM-capacity bound tasks dealing with high-resolution textures, scientific imaging, and inferencing LLMs (Large Language Models), Intel's Battlemage Pro series has the potential to offer a superior price-to-performance ratio, but that will depend on what Intel's price tags are.

The effects of dropping out of the enthusiast segment are also visible with AMD, with rumors suggesting their upcoming RDNA 4-based Radeon Pro W9000 GPUs will see a decrease in VRAM from 48GB to 32GB. With the rumored 24GB BMG-G21 GPU, Battlemage Pro looks strong on memory capacity. The big question is whether Intel will offer a higher-performance option for compute-heavy tasks, possibly with a model based on the BMG-G31 die? Let's wait for the official reveal to find out.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.

All the commands in this article will work on most Linux machines. We’ve used an Ubuntu LTS install but you could run this how-to on a Raspberry Pi. All of the how-to is performed via the Terminal. If you’re not already at the command line, you can open a terminal window on most Linux machines by pressing ctrl, alt and t or by searching for the terminal app in your applications menu.

Listing Files In Size Order Using the ls Command in Linux

The ls command is used to list the contents of a directory in Linux. By adding the -lS argument we can order the returned results according to the file size. We have copied a collection of files into a test directory to show this command but it can be run in any directory you choose.

To list the directory contents in descending file size order, use the ls command along with the -IS argument. You will see the larger files at the top of the list descending to the smallest files at the bottom.

ls -lS

While this command is useful for seeing, it lacks the actual size of the files so how can we identify the largest files in Linux and display their size?

Identifying Files Larger Than a Specified Size in Linux

In another article, we explained how to find files in Linux using the find command to search based on a filename or part of a filename. We can also use the find command in combination with the -size argument specifying a size threshold where any file larger than specified will be returned.\

1. Use find to search for any file larger than 100MB in the current directory. We are working inside our test directory and the “.” indicates to search the current directory. The -type f argument specifies returning files as results. Finally the +100M argument specifies that the command will only return files larger than 100MB in size. We only have one file in our test folder Baby_Yoda.obj that is larger than 100MB.

find . -type f -size +100M

2. Use the same command, but this time specify a path to search. We can run the same command as in the previous section but replace the “.” with a specified path. This means we can search the test directory from the home directory.

cd
find ./test -type f -size +100M

Searching the Whole Linux Filesystem For Large Files 

It’s sometimes useful to search the whole Linux filesystem for large files. We may have some files hidden away in our home directory that need removing. To search the entire filesystem, we will need to use the command with sudo. We might also want to either limit the search to the current filesystem which can be achieved via the -xdev argument, for example when we suspect the files we seek are in our current main filesystem or we can choose not to add the -xdev argument which will then include results from other mounted filesystems, for example an attached USB drive.

1. Open a terminal.

2. Search the current filesystem for files larger than 100MB. As we are invoking root privileges using sudo we will need to input our password. Note that we are using / to set the command to search the entire filesystem from the root of the filesystem.

sudo find / -xdev -type f -size +100M

3. Search all filesystems for files larger than 100MB. For this example connect a USB drive with a collection of files on it including some that are over 100MB in size. You should be able to scroll through the returned results and see that the larger files on the pen drive have been included in the results.

sudo find / -type f -size +100M

Finding the 10 Largest Linux Files on Your Drive

What are the top ten files or directories on our machine? How large are they and where are they located? Using a little Linux command line magic we can target these files with only one line of commands.

1. Open a terminal.

2. Use the du command to search all files and then use two pipes to format the returned data.

du -aBM will search all files and directories, returning their sizes in megabytes.

/ is the root directory, the starting point for the search.

2>/dev/null will send any errors to /dev/null ensuring that no errors are printed to the screen.

| sort -nr is a pipe that sends the output of du command to be the input of sort which is then listed in reverse order.

| head -n 10 will list the top ten files/directories returned from the search.

sudo du -aBm / 2>/dev/null | sort -nr | head -n 10

3. Press Enter to run the command. It will take a little time to run as it needs to check every directory of the filesystem. Once complete it will return the top ten largest files / directories, their sizes and locations.

With this collection of commands, you have several ways to identify and locate large files in Linux. It's extremely useful to be able to do this when you need to quickly select big files for deletion to free up your precious system resources. As always, take care when poking around your filesystem to ensure you aren’t deleting something critical!

Related Tutorials

• Getting To Know the Linux Filesystem
• How To Check Disk Usage in Linux
• How To Find Files in Linux
• How To Manage Users in Linux

The Frame 4000D offers modularity in how you place your fans, choices on using rear-connecting motherboards, and when they release, options for swapping out parts of the case.

This case deal is available at Amazon, where you can pick up the Corsair Frame 4000D for just £67, down from the original list price of £79. Finding a case like this for just £67 is amazing, and it is well-constructed with plenty of features. But, one of the reasons it is so cheap is that it does not come with any pre-installed fans, or a PSU.

This could easily be on our list of the best PC cases at this price, and with room for large 360mm radiator installations at either the top, side, or front of the case, it can provide cooling for the hottest CPUs. There is space for up to 12 fans in the Frame 4000D, with a unique-ish sliding rail that you can position to fit fans between 120 - 200mm, providing plenty of airflow for your build. Speaking of airflow, the case uses Corsair's funky 3D Y-pattern mesh design for its front panel. This design allows for more rigidity while not obstructing airflow, and it helps that it looks pretty cool.

There will also be modular parts such as an aluminium billet motherboard tray, and front I/O panels with different configurations, such as more USB-C ports, etc. However, these modular parts are currently unavailable, but are expected to be released in the coming months.

There is so much to cover with this case that there just isn't room in this deal's article, but luckily for you, we have a full review of the Corsair Frame 4000D that you can peruse and come to your own conclusions about this great case deal.

Follow Tom's Hardware on Google News to get our up-to-date news, analysis, and reviews in your feeds. Make sure to click the Follow button.