Intel on Thursday published its financial results for the second quarter of 2026, posting revenue of $16.1 billion — a 25% rise year-over-year — amid high demand for client and data center products. The company also said that due to extraordinary demand for its own CPUs, it will initiate mass production using its 14A (1.4nm-class) fabrication technology in 2028, which is in line with TSMC's plans for its A14 process technology.

"With encouraging external customer progress and increased demand for our internal products, we remain on track for 14A risk production for our internal products in the second half of 2027, and we made the decision in Q2 to fully commit to high volume ramp in 2028," said Lip-Bu Tan, chief executive of Intel, during the company's earnings call with financial analysts and investors.

Typically, companies initiate high-volume manufacturing (HVM) using a new process technology about a year after initiating risk production. Assuming that Intel intends to start risk production using 14A in the second half of 2027, it is reasonable to expect the company to start 14A HVM in the second half of 2028. It remains to be seen whether by the 'second half' Intel means July or December. If Intel initiates high volume ramp in November or December 2028, actual products made using the technology will emerge in 2029. In any case, Intel typically begins manufacturing using its leading-edge nodes in its development fab in Oregon and while it formally calls it HVM, actual volumes produced at such fabs are relatively low.

Another thing to note about Intel's 14A is that in 2028 it will start making its own products using this process, not products from external customers. Apparently, Intel still does not have any external clients that have committed to use the technology to make their products.

TSMC typically initiates production using its latest nodes in December and usually calls it the 'second half of the year.' Assuming that it follows the same pattern with A14 (though the progress of the node can enable the company to pull mass production in provided that customers' designs are ready), then the volume ramp will occur in 2029. TSMC claims that multiple customers have already taped out their ICs on A14.

Intel reported a strong second quarter as its revenue reached $16.1 billion, up 25% year-over-year and $1.8 billion above the midpoint of its guidance. Formally, Intel's GAAP losses reached a whopping $11 billion. However, this was driven by the $13.619 billion of mark-to-market losses on Escrowed Shares related to Intel's CHIPS Act Secure Enclave agreement with the U.S. government. Meanwhile, the company's non-GAAP net income was $2.2 billion, which reflects profitable underlying operations. The company's GAAP gross margin increased to 40.1%, up from 27.5% in Q2 2025. Intel's Operating cash flow totaled $7.0 billion, prompting the company to raise its capital spending outlook for both 2026 and 2027 as AI-driven demand continues to exceed available supply.

Intel's Client Computing and Physical AI Group (CCPG) generated $8.9 billion in revenue, up 13% year-over-year. The company confirmed that the CCPG result was not driven by increased unit sales, but was a result of higher average selling prices (ASP) due to supply constraints.

"Client obviously exceeded expectations. I would say it was largely ASP, of which some of that was mix related, some of that was our own like-for-like changes in ASPs where we thought we had seen some inflation on our cost and needed to pass that on to the end customer," said David Zinsner, chief financial officer of Intel, during the call.

The Data Center and AI (DCAI) business delivered the strongest growth as its sales climbed 59% year-over-year to $6.3 billion amid surging demand for Xeon processors, expanding AI infrastructure deployments, and rapidly growing purpose-built silicon sales.

"Q2 year-over-year server growth was the strongest on record, Xeon 6 continue to be one of the fastest ramping products in Intel history, reflecting improving execution and strong customer demand," Tan said.

"We also continue to see strong momentum in our purpose-built silicon product line, with revenue up roughly 20% sequentially and nearly tripling year-over-year," Zinsner said. "Purpose-built silicon revenue nearly tripled year-over-year."

Intel Foundry posted $5.8 billion in revenue, an increase of 31% year-over-year, as Intel 18A production ramped. At the same time the production units losses dropped to $2.1 billion — down from $2.4 billion in the previous quarter and $3.2 billion in the same quarter a year ago. External foundry revenue reached $293 million.

"Intel Foundry operating loss in Q2 was $2.1 billion and $348 million better quarter-over-quarter as higher yields improved cycle times and increased factory scale across Intel 4, Intel 3, and 18A drove improved wafer costs," Zinsner said.

Intel guided its third-quarter revenue to $15.8 billion – $16.8 billion and a projected non-GAAP gross margin of 42% and an EPS of $0.38.

A new 3D V-Cache-equipped CPU is reportedly in the works and could arrive as a mobile counterpart to the Ryzen 7 9800X3D. According to prominent tech industry insider Golden Pig Upgrade Pack on Weibo, the Ryzen 9 9800HX3D is expected to enter mass production in Q4 2026, with a potential launch at CES 2027. The chip is rumored to feature an 8-core, 16-thread configuration with boost clock speeds of up to 5.1 GHz, and 96MB of L3 cache.

It is also claimed that the processor was previously rumored to launch as the Ryzen 7 9755HX3D; however, it seems that the company might settle on the 9800HX3D. This essentially brings the branding in line with the desktop Ryzen 7 9800X3D. The chip could follow the same cache configuration with an eight-core CCD with 32MB of native L3 cache and a 64MB 3D V-cache stack.

Currently, AMD's most powerful 3D V-Cache-equipped mobile processor is the Ryzen 9 9955HX3D, featuring 16 cores, 32 threads, and a maximum boost clock of 5.4 GHz. It also offers a substantially higher L3 cache at 128MB and a configurable TDP of 55-75W.

While AMD hasn’t confirmed any details, the Ryzen 7 9800HX3D could likely be featured on premium gaming laptops that don't require the additional cores of the flagship 9955HX3D. An 8-core Zen 5 processor paired with 3D V-Cache could strike a better balance between gaming performance, power consumption, and overall system cost.

The rumored processor also aligns with AMD's recent approach of refreshing its X3D lineup instead of introducing new gaming CPUs. Similar to the recently announced Ryzen 9 9950X3D2 and the reintroduction of the 5800X3D, the 9800HX3D appears to build on an existing design while bringing AMD's latest branding and gaming-focused cache technology to another product tier.

The introduction of the 9800HX3D would also give laptop manufacturers a new gaming-focused option that is positioned below the flagship 9955HX3D. The processor would likely attract consumers seeking desktop-class gaming performance without paying the premium typically associated with 16-core mobile CPUs. As always, the rumored specifications and launch timeline should be treated with caution until AMD makes an official announcement.

Although the industry has largely learned to move out of the way when the big green giant that is Nvidia steps into the room, one AMD executive said he was “very happy” to see Nvidia publish SPEC CPU 2026 benchmarks for its Vera CPU ahead of the Advancing AI 2026 event. AMD used the configuration in Nvidia’s white paper as a basis to run SPEC on its new Zen 6 ‘Venice’ CPUs, offering what it calls an “apples-to-apples” comparison between the two chips.

“We are very happy that Nvidia published their Vera performance [numbers],” said AMD’s Ravi Kuppuswany, corporate VP of compute and enterprise solutions. “We were actually being a little conservative. I actually thought we were beating them by smaller numbers than what I have here… we thought we’d have at least a 10% advantage. What we’re finding is… we have 20% advantage, and we have not even finished completely tuning.”

Kuppuswany’s comments overlaid the slide you can see above, claiming 2.2x higher throughput with Venice compared to Vera, and 1.2x faster per-core performance. AMD has certainly stacked the deck in its favor here (as did Nvidia when it first published its Vera results), so let’s go through what this chart actually says.

As a quick aside, you’ll see “estimated” in the images above and below. These numbers aren’t guesses (they’re based on real runs), but SPEC maintains strict reporting guidelines for “official” runs. And because these runs aren’t reported to SPEC and therefore haven’t gained official status, they must come with the “estimated” disclaimer.

The throughput number is the easiest one to clarify. AMD compared its 256-core, 600W Epyc 9996 against the 88-core Vera, both in a 2P configuration. Yes, the CPU with more than twice the number of threads and an extra 150W stacked on top of its TDP has significantly higher throughput, as it should.

AMD would argue that it’s a fair comparison given that Nvidia is only offering Vera as a single, 88-core SKU. But the reality remains that the throughput comparison is one Vera could never win, and Nvidia would (and has) argued that it’s not trying to win in a race against the 256-core Venice. It’s building a single CPU for a single purpose.

The more interesting and consequential number here is the per-core performance. SPECrate_int is a throughput benchmark. A typical run loads all threads with a copy of an application and measures how much work gets done within a set amount of time. SPECspeed is the inverse of that, looking at a single application and how quickly it can run. AMD arrived at the numbers above by taking the overall SPECrate_int score and dividing it by the number of cores.

Nvidia’s results show Vera achieving a score of 925 overall. AMD says a 96-core High Frequency Epyc CPU achieved a score of 1,210. This processor, however, doesn’t seem to exist. AMD has the 96-core Epyc 9686F that boosts up to 5 GHz (much higher than the standard max frequency across the stack), but it has a rated TDP of 500W, not 600W.

Regardless, AMD divided that score by the number of cores; about 6.3 for AMD, and about 5.3 for Nvidia (remember these are 2P configurations). AMD says 1.2x, which actually translates to about an 18.8% lead. That’s not far off enough to say AMD was maliciously juicing its own numbers, but it’s important to note.

Given that AMD is using two different SKUs here (or maybe just one, considering the 9686F discrepancy), we can do the same per-core napkin math on the Epyc 9996 against Vera. Once again, Nvidia shared an overall score of 925, while the Epyc 9996 achieved a score of 2,070. That gives AMD a per-core score of 4.04.

That number isn’t important as a comparison point to Nvidia – again, we’re comparing a 256-core CPU to an 88-core one – but rather as a metric to see how Venice scales when normalized for per-core performance on a fully-loaded chip. That’s about 35% lower than the per-core score from the souped-up 9686F.

Although dissecting these numbers and sidelining the back-and-forth of two of the most powerful companies in the world is interesting, it’s not all that informative. SPEC maintains its strict reporting requirements for a reason. We won’t be able to say, with certainty, how these chips match up until we have official, reported runs. And even then, there’s an additional layer of compiler optimization (AMD and Nvidia both used GCC 15.2) and the broader context of the servers and workloads that these chips will serve.

The battle lines have been drawn, though only when looking at integer-based workloads. Vectorized performance is important, as well, and that’s an area where AMD holds a strong position in the current server CPU market. It doesn’t look like that will change with Vera, though we don’t have floating-point results to draw any conclusions from yet.

AMD has announced its long-term CPU roadmap, including multiple generations of its Zen microarchitecture at its Advancing AI event. Much of the event was focused on the new 256-core EPYC 9996 sporting the Zen 6 architecture, as well as Venice-X, which is slated to arrive next year. But AMD also teased a Zen CPU roadmap going out to Zen 8.

"In 2028, we are going to introduce Florence," Lisa Su, chief executive of AMD, said. "Florence brings the next-gen Zen 7 cores and its leading-edge process technology. It's a new set of AI compute extensions to really ensure that we have all of the AI capability, and it supports the latest memory technologies."

Su did not disclose specific details about the process technology, core counts, memory subsystem, or AI extensions planned for (Zen 7) Florence. However, she indicated that Florence will not be a standalone processor family, instead joined by the Ferrara AI host node and Faenza Agentic Sandbox for different AI workloads. AMD says this approach will enable it to offer purpose-built CPUs for a wider range of applications rather than relying on a single architecture configuration across the entire server market.

"And we are not stopping there," Su said. "We are already deep in development of Ravenna, our 8th Generation EPYC family built on Zen 8, and that family is already well under development for 2030."

The CPU roadmap is part of AMD's general strategy to offer a predictable cadence for its data center platforms.

On the accelerator side, the company is developing its MI500-series Instinct products with next-generation HBM, larger scale-up domains, and new copper and optical interconnect technologies, similar to Nvidia's outlook toward Rubin Ultra and Feynman.

"MI500 will deliver the largest generational leap in the history of Instinct, putting us on track to deliver more than 2,000 times higher inference throughput in just four years," Su said.

AMD also confirmed that its CDNA Next-based Instinct MI600 family is already deep in development for 2028. At the event, AMD launched its MI455X GPU, as well as teased more about its upcoming MI500X range, which we expect to launch next year.

AMD is bringing its Strix Halo APUs into the world of physical AI. The new X100 series of processors come with similar specs as the various Ryzen AI Max models floating around in client devices, but they’re tailored for 24/7 operation, with a 10-year lifecycle in embedded applications like robotics.

There are three SKUs that align with the three original Strix Halo models (not the updated versions with 40 CUs). The top-end X199 comes with 16 Zen 5 cores and 40 RDNA 3.5 CUs. The X188 steps down to 12 cores and 32 CUs, while the X168 comes with eight cores and the same 32 CUs. AMD hasn’t shared detailed specifications for each model, but the company says the range goes up to a 5.1 GHz boost clock and 128 GB of unified memory. They also include an XDNA 2 NPU with up to 50 TOPS, a configurable TDP between 45W and 120W, and operating temperature between -40 degrees Celsius up to 105 degrees.

AMD’s range bites back at Intel, which launched a range of Panther Lake SoCs for physical AI at the beginning of the year. Both make an argument for SoCs in robotics, reducing latency when the CPU, AI accelerator, and memory are fragmented across separate chips. The X100 range is just physically larger than Panther Lake, packing much more silicon on the SoC for more powerful deployments.

The company shared a range of benchmarks comparing the flagship X199 against Intel’s Core Ultra X7 358H, a 16-core chip with Intel’s Arc B390 iGPU that has 12 Xe3 cores. AMD claims a lead of 1.2X and 1.3X, respectively, in GeekBench 6.1 and PassMark, as well as 1.5X in an unofficial SPECrate 2017 run looking at integer workloads. In graphics, AMD unsurprisingly takes the edge with 1.4X faster Vulkan and 1.7X faster OpenGL performance (both measured with GFXBench 5 on Ubuntu), as well as a 1.6X lead in Unigine Heaven Extreme.

On the physical AI front, AMD claims a 1.4X improvement in Time to First Token (TTFT) and 3.5X faster tokens per second in Llama-bench, with a Vulkan backend running at a 45W TDP. These results need a massive dash of salt, however.

AMD tested the Ryzen AI Max 395+ “configured to reflect Ryzen AI Embedded X199 specifications.” It tested on the Maple reference board with a 5.1 GHz CPU clock, 2.9 GHz GPU clock, and sustained 45W TDP. The X7 358H, meanwhile, was tested in an MSI Prestige 16 Flip AI+ with an enforced TDP limit of 30W. AMD then “projected” 45W performance on the Intel chip “using scaling factors derived from public benchmark data.”

It’s not exactly an apples-to-apples comparison, in other words. There’s some sort of proxy stand-in or extrapolation of data across all of the benchmarks here, so keep that in mind as we work through the rest of AMD’s X100 announcements.

AMD X100 Kria SOM and robotics developer platform

utside of the chips themselves, AMD is offering X100 models as part of a Kria System on Module (SOM) or an integrated robotics developer platform. The Kria X100 board measures 120mm x 120mm and conforms to the standardized COM-HPC form factor. If you’re a developer that wants to develop for the board, AMD is offering its Kria AI robotics developer platform.

It’s a fully-integrated box, leveraging the X100 Kria SOM alongside AMD’s Spartan UltraScale+ FPGA baseboard. AMD says it’s a “turnkey” solution for robotics development, including specialized connectivity for cameras and industrial networking, along with robotic sensors. The platform is available in early access now, and AMD says it’ll be in full production in Q4 of this year. .

AMD shared some benchmarks for the X100 Kria, as well, comparing it to Nvidia’s Thor T5000. These benchmarks weren’t run internally at AMD. They were commissioned by AMD and ran by Open Navigation and Mimix. Critically, the benchmarks didn’t test an X100 Kria board, or at least, not exactly in the form it will take once it’s inside a robot or AMD’s developer box.

Instead, AMD is comparing Nvidia’s Jetson AGX Thor developer kit to a GMKtech EVO-X2 AI mini PC with a Ryzen AI Max+ 395 “configured to reflect Ryzen AI embedded x199 specifications.” Naturally, the thermal and power environment of these chips will heavily influence performance.

AMD is continuing its attempt to siphon developers away from Nvidia’s CUDA platform for development, as well. It’s HIPIFY tool converts CUDA code to AMD’s HIP C++ portable code, and the company claims it can now handle 70-80% of the “effort” of porting on its own. AMD tested on a Ryzen AI Max+ 395, once again configured to match the X199, and it ported 15 CUDA applications, comprising 1,199 lines of code, to arrive at that 70% to 80% range.

X100 Kria lives at the “brain” of the robotics platform, but AMD envisions an end-to-end solution for humanoid-style robots with its Spartan UltraScale+. Zynq UltraScale+, and Versal AI Edge Gen 2 FPGAs and SoCs

AMD is finally providing some details on its first Zen 6 CPU, which it has been teasing for over a year. The Epyc 9996 is a 256-core / 512-thread chip, packing AMD’s new Zen 6 architecture, and it’s the first to launch in what AMD describes as a “broad portfolio” for Venice. In addition to claiming significant performance advantages over the impending Nvidia Vera and Intel’s Xeon 6, AMD says it will continue to build out the Venice range with bespoke designs over the next year.

“It’s not just a single processor,” said AMD’s Ravi Kuppuswany, corporate VP of compute and enterprise solutions.q “It’s a portfolio.” AMD says it has purpose-built solutions, splitting its offerings depending on the application, not dissimilar to how Intel has split its Xeon ranges over the past few generations (nor how AMD has softly segmented its Epyc offerings). The roadmap starts with the main Venice lineup on the SP7 socket, which is what AMD has been teasing for so long. It scales up to 256 cores and 512 threads, 1.6 TB/s of memory bandwidth with fast MRDIMMs, and 128 PCIe 6 lanes in 1P configuration (160 lanes in 2P).

Note: When scaling up to 256 cores, AMD uses its Zen 6c “dense” design. With a standard Zen 6 design, AMD says Venice scales up to 128 cores and 256 threads, while high-frequency variations top out at 96 cores.

In the first half of next year, AMD plans to launch Venice on its SP8 socket, offering as few as eight cores and up to 128, focused on smaller deployments. These chips support eight-channel memory with two DIMMs per channel, and the same 128 PCIe 6 lanes.

Venice-X is expected in the second half of 2027, on the SP7 socket. We didn’t see Turin-X, but the last, last-gen Genoa-X came with 96 cores and up to 1152 MB of stacked L3 cache. Those specs haven’t changed (short of the Zen 6 microarchitecture), but AMD says it's able to clock Venice-X up to 5.15 GHz.

Finally, Verano should arrive in the second half of next year on the SP8 socket, and it looks like the most direct competitor to Vera (AMD’s Kuppuswamy had some fun with calling it “Vera-No”). It’s optimized to be an AI host node, says AMD, packing up to 72 cores and 5 GHz peak clocks. Critically, it comes with a 24-channel LPDDR5X memory system, leveraging SOCAMM2 modules.

One of the advantages Nvidia claims with its Vera chip is lots of memory bandwidth through the LPDDR5X system. AMD’s approach is different with Venice SP7. It’s scaling up to 16-channel memory with Venice SP7, with support for MRDIMMs running at 12,800 MT/s, or standard DDR5 RDIMMs running at 8000 MT/s.

It’s a significant jump over Turin, which uses 12-channel memory, with support for RDIMMs running at 6400 MT/s. AMD claims per-socket bandwidth of 1.6 TB/s, significantly higher than the 1.2 TB/s available on Vera, and nearly triple the 576 GB/s per-socket bandwidth of Turin. Intel recently enabled 8000 MT/s RDIMMs on select Granite Rapids and Clearwater Forest SKUs, and it says support for MRDIMMs with speeds up to 8800 MT/s is coming in Q1 2027.

Zen 6 is built on TSMC’s N2 (this has been previously confirmed). AMD confirmed that there are 32 cores on a CCD, along with two IODs. Keep in mind that the 32-core CCD is using Zen 6c, not full Zen 6. There has been plenty of speculation about 32-core CCDs in consumer Zen 6 CPUs, but that seems unlikely.

The 256-core configuration comes with a massive 1,024 MB of L3, nearly triple the amount of the Epyc 9965. This isn’t stacked cache, either; that will come with Venice-X. Each CCD has access to 128 MB or L3, or 4 MB per core, double what was available on Turin.

Although AMD has focused a lot of its teases on the 256-core Venice, the initial SP7 offerings will also hold a 96-core, high-frequency model that can clock up to 5 GHz.

AMD shares first 256-core Epyc ‘Venice’ benchmarks

Unlike the extrapolated performance AMD shared a few weeks back, we have some concrete benchmarks for the Epyc 9996 now. AMD has, unsurprisingly, focused the workloads around agentic AI. However, many of the workloads applicable for agentic AI are applicable elsewhere, as well, including high-concurrency networking tasks, code compilation, and media processing.

Note that AMD includes just the Epyc 9965 as a gen-on-gen comparison point in the charts above. This is a “dense” Zen 5 design with 192 cores. Results for the 128-core 9755 are included in the tables below.

Starting with front-end operations, AMD claims a 1.2x gen-on-gen improvement and a 2.8x improvement compared to Intel Xeon 6980P, with an NGINX web server using the WRK load generator. Unlike most of these competitive performance figures, AMD included the actual numbers for the benchmarks it ran in the footnotes, which you can see in the table below.

In data-heavy workloads that are common among AI agents, AMD claims a 1.7x improvement over Turn, and a massive 3.4x over the Xeon 6980P. AMD used the TPCx-AI benchmark to gather these results. The primary metric for this test is AI use cases per minute (AIUCpm), for which AMD shared the median result2. If you’re interested in more about the reporting of this benchmark, Dell has published an extensive breakdown.

In vectorized workloads, AMD claims a 1.6x gen-on-gen improvement and 2.3x improvement compared to the Xeon 6980P. For this test, AMD used Meta’s open-source FAISS (Facebook AI Similarity Search) library to search for similar vectors in the siftm1 dataset. The metric here is QPS, or queries processed per second, looking at overall query throughput.

For its “enterprise tools” benchmarks, AMD ran several tests, including TPC-H, TPC-C, and Redis, and it reports the results as “geomean throughput.” We have actual numbers here, but they’re a geomean representing several different tests rather than a single benchmark. Broadly, however, AMD claims a 1.6x gen-on-gen improvement in these workloads, and a 2.6x improvement compared to Intel.

A lot of agentic workloads are applicable outside of agents, but AMD also tested a few agents directly. It replayed five different agent personas across the chips and, once again, gathered a throughput geomean. We don’t have the metrics here, nor for the previous benchmark, so it’s possible there’s an angle of performance that we’re not seeing with the data provided by AMD.

Regardless, the company claims a 1.5x gen-on-gen improvement in this test, and a 2.5x improvement compared to the 6980P.

AMD ran these tests earlier in the month. But just a few days ago, Nvidia published its first SPEC CPU 2026 results for Vera. AMD ran some tests of its own using the same compiler for a comparison between Vera and Venice. AMD’s Kuppuswamy says, “everything is apples-to-apples comparison, same compiler.” That’s GNU 15.2, if you’re curious.

In throughput, AMD claims a 2.2x improvement in the SPECrate integer suite, compared to Vera using the dense Venice design with 256 Zen 6c cores. More importantly, AMD claims a 1.2x improvement in per-core performance when comparing Vera to a 96-core “High Frequency” Venice chip. AMD says it used Nvidia’s results as the basis for comparison. With both Venice designs, AMD used a 600W TDP.

In SPEC CPU 2017 (again using SPECrate with integer workloads), AMD has data comparing Venice to Intel’s 6980P and Arm’s new AGI, showing 2x throughput compared to Intel, and 1.3x per-core performance. Note the core counts here for AMD. SPECrate is a throughput test, and AMD stepping down to a 128-core model suggests that performance will likely drop off as the core count increases.

Although AMD wants to focus Venice performance on agentic workloads, it shared a range of what are now being called “legacy” workloads across the cloud and HPC. Some of the results are repeated from the earlier slides, such as Redis and NGINX, but there are some additional data points, including NAMD and SQL. The performance improvements here are large, though not surprising. You can see that across these tests, even Turin beats the competition from Intel and AWS.

AMD’s Venice-X CPU is launching in the second half of 2027, the company revealed Thursday at its Advancing AI event. The CPU is built on the same Zen 6 microarchitecture as Venice SP7, which AMD says is now in production, but it comes with smaller core counts and leverages AMD’s 3D V-Cache to stack an unbelievable 1152 MB of L3 cache on the chip, all while offering 96 cores and clock speeds up to 5.15 GHz.

We’ve previously heard teases of Venice-X, and AMD has released variants of its Epyc chips like this in the past, such as Milan-X and Genoa-X. But we never saw Turin-X, and we haven’t gotten any concrete details about Venice-X up to this point. Now, we have a peek into what Venice-X will offer, which looks similar to Genoa-X in some regards.

AMD is sticking with 96 cores with Venice-X, as well as 1152 MB of stacked L3 cache. There are a few major differences compared to Genoa-X, however. First, AMD says clocks are much higher here, claiming Venice-X tops out at 5.15 GHz (the 96-core Epyc 9684X clocks up to 3.7 GHz). Venice-X also comes with the memory improvements available in most of the chips in the range.

It supports 16-channel memory, either with standard DDR5 RDIMMs at up to 8,000 MT/s or with new MRDIMMs 12,800 MT/s, offering up to 1.6 TB/s of aggregate memory bandwidth. The memory system in Venice broadly is a massive increase over Turin, nearly tripling the aggregate bandwidth, mainly on the back of speeds enabled by MRDIMMs.

Venice-X is one of four variations of AMD’s Zen 6 lineup for data centers, mainly targeting HPC workloads, with its massive L3 cache and relatively high clock speeds. Venice-X uses the same SP7 socket as the main Venice range. Other variations of Zen 6 offerings, including Verano, will use the SP8 socket instead.

AMD’s standard Venice chips support up to 1024 MB of L3 cache on their own, nearly triple that of its previous flagship offering. That cache is split among 256 cores, however. Venice-X doesn’t come with much more cache overall, but that cache is serving far fewer, more powerful cores, which should be useful in HPC workloads.

AMD didn’t disclose any other details about Venice-X, nor any performance projections. Given that it’s set to launch in the second half of 2027, we’re still a ways out from that.

Intel and AMD are signing longer-term purchase commitments with Chinese server customers for data center processors, according to a new Reuters report that cites two people familiar with the talks. Prices for some server CPU products in China have climbed more than 40% since the start of the year and are still rising by over 10% month-on-month in some cases, one of the people said. Most of the agreements guarantee purchase volumes for about a year without fixing prices, and neither company responded to the publication's requests for comment.

Some customers have discussed commitments running two years or longer, one of the people told Reuters, with the deals covering unit volumes only, which leaves Chinese cloud providers and internet companies fully exposed to a market that has already added 40% to some CPU prices in under seven months. Memory makers struck similar long-term agreements with hyperscalers over the past year as AI demand consumed DRAM and NAND output, but those contracts typically trade a volume commitment for some pricing visibility. Buyers here get neither price protection nor, based on current lead times, fast delivery, but they do get a place in the queue.

Intel said in March that it was pursuing long-term agreements with potential customers as server CPU demand spiked, with CEO Lip-Bu Tan telling analysts on the company's April earnings call that demand "continues to run ahead of supply," singling out Xeon server parts. Tan also cited a multi-year supply deal with Google among several long-term contracts signed in the first quarter, so the model now spans U.S. hyperscalers and Chinese channel buyers alike.

Reuters first reported in February that Intel had warned Chinese customers of lead times stretching to six months on some server CPUs, with AMD quoting eight to 10 weeks. By late March, server makers told Nikkei Asia that average CPU lead times had blown out from one to two weeks to eight to 12 weeks and beyond. Intel CFO David Zinsner put a value on that in April, telling investors that unmet Xeon demand "starts with a B," and the company later confirmed price hikes on select consumer and server CPUs.

Agentic AI workloads are driving much of the demand, as inference and orchestration tasks pull server CPU-to-GPU ratios back toward parity. AMD, which is due to report earnings in early August, has raised its server CPU market forecast to more than $120 billion by 2030, based on the same trend.

Intel will report its second-quarter results later today, where the shortage and the durability of these commitments are likely to come up. Chinese buyers remain cut off from the most advanced AI accelerators under U.S. export controls, but Xeon and EPYC server CPUs carry no such restrictions, making them one of the few U.S.-made AI infrastructure components China can still purchase freely, if it can get them.

Nvidia’s Vera CPU is its first bid to become a key player in the data center CPU market. Although Grace has seen some success (most notably with Grace standalone deployments at Meta), Vera is Nvidia’s first CPU with a custom core design. It’s arriving at an ideal time, as well, with the server CPU market exploding in the last few months on the back of agentic AI demand.

Vera isn’t a chip built to chip away at the market share of AMD and Intel in the cloud. It’s built to grab market share in an expanding market, as hyperscalers look to widen AI infrastructure beyond legacy clouds. As such, it’s designed in a much different way than Nvidia’s x86 competitors, and it even holds some unique architectural design points compared to the swath of Arm-based designs.

Nvidia has slowly revealed more details about Vera as it ramps into general availability, which is on track for the back half of this year. Now, we have a full picture of the chip. Nvidia shared its Vera white paper, along with unofficial SPEC CPU 2026 results comparing Vera to AMD’s Turin-based Epyc 9755.

We’re going to break down the white paper here, including all of the details about the Olympus core and a look at the benchmarks Nvidia ran. At the end of this piece, we’ll also take a brief look at the larger context of Vera and how it fits into Nvidia’s wider AI ecosystem compared to standalone deployments.

But plenty of ink has been spilled about Vera’s technical capabilities and Nvidia’s next-gen AI infrastructure vision. Let’s start with the important thing: the benchmarks.

Nvidia Vera CPU benchmarks

We’ve seen Vera in action before, though only through a series of selected benchmarks ran at Nvidia HQ by Phoronix. In the Vera white paper, Nvidia shared benchmarks for SPEC CPU 2026, specifically the integer suite from SPECrate, against AMD’s Epyc 9755, with both chips running in a dual-socket configuration. Before getting into the results, there are some important notes about how SPEC runs work, and the reporting criteria for them.

Nvidia’s run here isn’t official, as Vera was tested in a reference system due to the fact that it’s not broadly available yet. It’s ramping for general availability in the second half of the year. Due to that, Nvidia is unable to report its results. That’s why you see “estimated” in some of the charts below. Nvidia ran SPEC CPU 2026; it’s not extrapolating expected performance like we’ve seen from AMD so far with its upcoming Venice chips.

SPEC CPU 2026 is split into four suites, but Nvidia tested the SPECrate integer suite, which is focused on system throughput with integer-based workloads. The “rate” result is looking at how much work is completed within a certain amount of time. Here, each thread in the system has a copy of the workload. The score is how much time it takes for those workloads to complete, regardless of thread count, naturally giving chips with more cores an advantage.

If you want more detail on the benchmarks included in the suite, make sure to read our original coverage of SPEC CPU 2026. Here are the overall results:

Nvidia didn’t share the exact results for the 9755 it tested, short of the overall score of 898. Taking that overall score into account, Vera is 3% ahead of the 9755. It’s worth noting that Vera is ahead here despite a large thread disadvantage. An overall score of 898 for a dual-socket Epyc 9755 system isn’t unreasonable compared to publicly-submitted SPEC CPU 2026 runs, though higher results have been published. SPEC CPU ships as source code, which the tester must compile with their compiler of choice, and that can heavily influence results (particularly with vendor-specific compilers). Nvidia used GNU 15.2 with both systems.

Above, you can see Vera’s results stacked up against the 9755, but these aren’t comparing the numbers directly. Nvidia has normalized the per-core performance, which isn’t how SPECrate results are normally shared. According to the overall numbers, Vera is still completing more work within the same amount of time, despite a thread disadvantage, but the margins aren’t in the range of a 70% or 80% advantage as the above chart suggests.

We asked Nvidia about the results given that they're obfuscated by comparison; we could not reverse-engineer the Epyc 9755's scores with the information Nvidia has provided. Here's the response it gave: "Per-core performance under a fully loaded socket is important because agentic AI and RL run many sandboxes concurrently, while each agent step remains sequential and latency-sensitive. It measures how much performance each core sustains amid contention for shared power, memory, cache, and fabric. We therefore normalize by physical core, with SMT enabled on both systems."

The “agentic” workloads Nvidia has highlighted here are code compilation and interpretation workloads, which is something an agent is often doing, querying repos for dependencies and building source code. Below are data science workloads (or Exploratory Data Analysis), and below that are data processing workloads like SQLite database management. The results here align with Nvidia’s overall messaging of Vera, that it’s highly competent at data-rich, backend operations.

Although Nvidia is sharing per-thread results, it argues that SPECrate is still the correct benchmark to run. The per-thread results here are in the context of a fully-loaded socket. Here’s the justification from the white paper: “This metric is non-trivial for agentic AI and RL systems, where many sandboxes, tools, and environments run concurrently rather than as isolated single-thread tests. Fully loaded per-core performance captures how well each core sustains throughput while sharing socket-level power, memory bandwidth, cache, and fabric resources.”

In addition to running the workloads, Nvidia analyzed the code execution for architectural benchmarks, which you can see in the gallery above. Nvidia claims an overall IPC gain of up to 1.9x compared to Turin, up to 2.3x more branch predictions and 3.5x taken branches per cycle, and up to 2.4x higher instruction fetch operations per cycle.

Outside of SPEC, Nvidia shared a few benchmarks highlighting the capabilities of the Olympus core. First up is PageRank, an algorithm developed by Google to originally rank web pages, which highlights Olympus’ prefetch engine. Nvidia scaled this workload to higher core counts, showing Vera maintaining much of its single-core performance up to 32 cores, while the Turin chip hits a wall around 20 cores.

In addition to the above results, Nvidia shared some tests of the Vera memory system compared to Turin. These microbenchmarks are good for validating Nvidia’s specifications, but they’re looking at architectural performance, not application performance. An architectural advantage translates into a performance advantage, but not always in a linear, expected fashion.

Nvidia used internally-developed tools for the memory tests, though they're available on GitHub for anyone to run.

First is loaded memory latency, stressing the memory subsystem as bandwidth usage increases. Vera has much higher bandwidth overall, but you can see the Turin chip hit a latency wall below its maximum, which Nvidia attributes to Non-Uniform Memory Access (NUMA) domain traversal and CCD-to-CCD latency.

Looking at per-core bandwidth, Nvidia claims Vera provides more than four times the bandwidth of AMD’s 9755. The suggestion here is that “real-world” per-core bandwidth is even better than Nvidia’s specs lead on (or perhaps worse than AMD’s).

Maybe the most consequential of these tests is the one you can see above, looking at core-to-core latency. It’s no secret that crossing the CCD on AMD’s chiplet-based architecture incurs a big latency penalty. You can see that in action even in our Ryzen 9 9950X3D2 review, and the penalties compound as you scale up the number of CCDs.

In fairness to AMD here, chiplet-based designs aren’t built for this type of cross-CCD traversal, preferring to keep workloads localized and optimizing for core density. Vera’s design goal is clearly to keep latencies consistent across the entire die and sacrificing core density in the process. Nvidia’s Ian Buck told us that this design trade-off “will come at the cost of the legacy workload,” when we recently visited Nvidia HQ.

That’s important context. Nvidia isn’t gunning to steal existing market share from AMD and Intel as much as it’s trying to grab market share in an expanding market before AMD and Intel can. Some financial institutions (including Morgan Stanley and Bank of America) suggest the server CPU market could double in size (or grow even larger) by 2030. That context is important because there will be a continuing demand for CPUs that can handle workloads Vera is not optimized for, and it’ll be interesting to see how AMD and Intel tackle that dynamic with future products, trying to keep a legacy base of customers while pushing ahead into the expanded market.

Nvidia clearly has a vision of how that expanded market looks, and to that end, hasn’t shared SPEC CPU floating point results. Presumably, this is due to the fact that SPEC’s vectorized suite is focused primarily on HPC workloads, whereas Nvidia focused on what it believes are critical agentic workloads that are integer-based. Vera has a vector engine complete with SVE, but that doesn’t seem like Nvidia’s focus.

In an end-to-end Nvidia system, those vectorized workloads would be offloaded to a Rubin GPU. Still, we don’t have any vector results for Vera yet. Up to this point, we’ve only seen integer results, which is strange given the memory system at play in Vera.

Vera is Nvidia’s first CPU with a core design created in-house, which is the Olympus core. It’s built on Arm v9.2-A, but the design was created by Nvidia, unlike Grace, which leveraged a stock Arm design. Each Vera CPU has 88 Olympus cores on a monolithic die, breaking from the chiplet-based designs available from Nvidia’s x86 competitors.

Nvidia says Vera comes with a 1.5x increase in instructions per cycle (IPC) throughput compared to Grace, and 50% higher performance compared to x86 competitors (it seems that number is per-thread performance with a fully-loaded socket). Nvidia has a single 88-core design with Vera that supports spatial multithreading for 176 threads.

The CPU has a configurable TDP range of 250W to 450W. It uses a SOCAMM2 LPDDR5X memory system with capacity of up to 1.5 TB and speeds up to 9600 MT/s, and comes with 164 MB of L3 cache and 2 MB of L2 per core. Vera includes significantly less L3 than Intel’s highest-specced Xeon 6 and AMD’s Zen 5 chips. It actually has more L2 than L3 overall. This, presumably, is due to Nvidia’s fabric, which distributes the L3 in a mesh across the monolithic die.

Below, you can see a layout of the Olympus microarchitecture. Nvidia has disclosed some of the highlights of the architecture previously, such as the 10-wide instruction decode and neural branch predictor, but we now have a full view of the architecture courtesy of Nvidia’s Vera white paper.

The front end starts with Nvidia’s neural branch predictor that can run two branches per cycle “with zero penalties,” according to Nvidia. Research on neural branch prediction dates back to the late 90s, but Nvidia says it has a “novel” neural branch predictor, perhaps building on previous research such as BranchNet.

The BPU feeds into the Instruction Fetch Unit, which holds 64 KB of L1 instruction cache, and loads into a decode queue that supports 48 instructions (we’ll go into the memory/cache layout later). At the last stage of the front end is that 10-wide decode, feeding more instructions into the execution engine per cycle than the 8-wide decode in AMD’s Zen 5 microarchitecture.

Past the front end, the mid-core rename / allocation engine is built to keep instructions moving while waiting on dependencies. In addition to renaming and allocation, instructions work through value prediction, which can speculatively execute the instruction, and memory renaming, where the instruction can move forward while a load is happening if the data relationship can be determined.

Inside the execution engine, Nvidia includes eight simple Arithmetic Logic Units (ALUs), two complex ALUs, and four branch units for resolution. For SIMD instructions, the execution engine includes a vector cluster for Arm’s Scalable Vector Extension (SVE), including six vector units that support 128-bit SVE instructions at FP8 precision, along with two crypto-enabled vector units that can handle AES, SHA, and SM3, among other prominent algorithms. Keeping data moving through the engine are four load units and two store units.

The cores support spatial multithreading, giving a Vera CPU with 88 cores access to 176 threads. Traditional SMT time-slices execution, giving both threads access to all of the core resources and sharing them as instructions execute in parallel. With spatial multithreading, each thread of an Olympus core has access to dedicated resources, allowing one of the threads to execute with high-throughput while the other thread handles simple tasks, or to operate as two independent execution environments.

The execution resources are partitioned, explaining the wide decode front end. It’s not clear, however, if the SMT implementation can also opportunistically grab resources, particularly in the scenario Nvidia describes where one of the threads is maximizing throughput while the other handles smaller tasks.

There’s a lot going on in Vera between the 10-wide decode, neural branch predictor, and spatial multithreading, but perhaps the most significant architectural design point is Nvidia’s second-generation Scalable Coherency Fabric (SCF). It underpins Nvidia’s approach of using a monolithic die as opposed to a chiplet-based design, distributing last level cache in a mesh across the die and avoiding the cross-CCD latency penalty with localized L3.

The mesh distributes data through a series of Coherency Switch Nodes (CSNs) that serve as routing points between cores and the 164 MB of distributed L3. These routing points further connect the cores and L3 to the memory system, I/O, and NVLink C2C for cache-coherent communication between chips. Nvidia’s benchmarks comparing Vera to AMD’s Epyc 9755 show that AMD can achieve slightly lower core-to-core latencies within a cluster, but Vera maintains significantly better core-to-core latency across the die, as expected.

Nvidia says SCF inside Vera has 3.4 TB/s of bandwidth, enabling faster core-to-core communication, especially when traversing the die. However, Vera also supports Memory System Resource Partitioning and Monitoring (MPAM), allowing portions of L3 to be partitioned in multi-tenant environments.

Vera uses SOCAMM2 LPDDR5X, which is a relatively new advancement that Nvidia’s competitors haven’t had the chance to benefit from. With the use of SOCAMM2, LPDDR5X provides similar modularity and capacity as traditional RDIMMs, but at significantly lower power draw.

The memory can run at up to 9600 MT/s, with aggregate bandwidth of 1.2 TB/s and per-core bandwidth of 14 GB/s, doubling the bandwidth of Grace. The Vera board supports eight SOCAMM2, offering capacity ranging from 256 GB to 1.5 TB. Nvidia claims a “fully populated” memory subsystem consumes between 30W and 40W depending on capacity.

For I/O, Vera supports PCIe 6.4 with 88 lanes per CPU and bifurcation support down to x2. It also supports CXL 3.1.

Unlike Grace, Vera includes Arm’s Confidential Computing Architecture (CCA) and Realm Management Extension (RME), including Device Assignment and Coherent Device Assignment, offering a boon to multi-tenant environments where VM isolation is key. Nvidia also implements TDISP for coherent devices, allowing for encrypted communication between GPUs and PCIe devices.

Nvidia is already sampling Vera as a standalone chip to partners, and it says it will continue to do so, but the vision is an end-to-end solution built on Nvidia’s CPUs, GPUs, switches, NICs, and even rack specifications. Nvidia doesn’t make all of these individually, at least not at scale — just like with desktop graphics cards, Nvidia provides its MGX reference design, which customers can purchase, but partners also offer their own racks, some built solely to Nvidia’s specifications and others with more speciality.

Each tray comes with two Vera Rubin superchips, each of which contain a single Vera CPU to two Rubin GPUs, giving you two CPUs and four GPUs per tray. At the front, Nvidia partitions off three spaces, with the MGX design carrying two NVIDIA ConnectX-9 SuperNIC on either side and a Bluefield 4 DPU in the middle. Critically, this design doesn’t include any hoses or fans. It’s entirely liquid cooled, and it contains just two cables throughout the entire tray.

Nvidia demonstrated this cable-less design, showing the Vera Rubin Superchip sliding in and out of the track with a retention mechanism in a matter of seconds. The company says assembling the rack takes less than a few minutes and is handled entirely by robots, which is a far cry from GB200 and GB300 trays.

GB200 and GB300 trays are dense designs, but they’re also cluttered with cables and hoses. Nvidia says this massively slowed down production time, eventually leading to production issues that delayed Nvidia’s rollout. The company says that won’t happen with Vera Rubin and its largely cable-less design. Whereas a Grace Blackwell tray took around two and a half hours to assemble by a human, the company says a Vera Rubin tray is assembled within five minutes and entirely automated by robots.

Each tray needs to dissipate several kilowatts of heat, which Nvidia says it’s able to do using dry cooling. Liquid cooling is nothing new in the data center, either through an external chiller (essentially an A/C unit) or evaporation, where a fan evaporates water over a mesh and cools without the need for a compressor. With Vera Rubin, Nvidia uses “dry cooling,” with a maximum inlet temperature of 45 degrees Celsius.

Nvidia says it’s able to get the full performance out of a tray given an inlet temperature of 45 °C, allowing trays to operate without an additional water consumption in environments up to 100 degrees Fahrenheit. The tray essentially uses a large closed-loop similar to what you find from a consumer AIO, just scaled up. Water moves out of the tray and outside the data center, and it passes through a radiator where fans dissipate the heat. There’s some extra power consumption from water pumps and fans, but not nearly on the scale of evaporation methods or chillers.

The result is a tray completely free of fans, essentially noise-less in operation, and doesn’t strain local water infrastructure. That’s what Nvidia says, at least. In many locations around the U.S. where data centers are located (Texas and Virginia chief among them), temperatures easily climb above 100 °F during the Summer, prompting some sort of backup method of cooling. Nvidia says the external temperature ceiling can go higher depending on different factors — running at lower power, for example, and using more efficient heat exchangers — but under normal conditions, 100 °F is the ceiling.

It’s worth noting that nothing about a Vera Rubin tray explicitly requires this method of dry cooling; the hardware is just capable of offering full performance with an inlet temperature of 45 °C.

In addition to an NVL72 design, Nvidia has a Vera standalone deployment that compacts each tray into a series of SOCAMM2 slots and Vera chips. In Nvidia’s 48U MGX design, a standalone Vera deployment can include up to 256 CPUs in a rack.

Underpinning scale-up communication is Nvidia’s sixth-generation NVLink, which is deployed as switches in the rack and connected to compute trays using Nvidia’s NVLink spine. You can see the spine on its own in the image above, which features over two miles of thin copper wire to allow every tray in the rack to communicate with each other.

Localizing storage, networking, security, and telemetry operations is Nvidia’s BlueField-4 DPU. A Vera Rubin NVL72 compute tray includes a single DPU and two ConnectX-9 NICs to maximize CPU/GPU utilization. You can read more about Bluefield 4 in our original coverage from GTC.

Nvidia’s Ian Buck, vice president of hyperscale and high-performance computing and the inventor of CUDA, says the company has “shipped... let's put it in the hundreds of thousands of Grace standalone servers.” In May, Nvidia disclosed that it had shipped over 2.5 million Grace CPUs in total, and the company announced a partnership with Meta to deploy standalone Grace servers in February. Buck’s comments suggest the scale of deployment may be even larger, however, as Nvidia tries to compete in a market dominated by other players.

It’s an interesting comment, though not a surprising one. Nvidia has become the dominating force of Silicon Valley as demand for its GPUs skyrocketed during an unprecedented data center buildout for AI inference. Since peaking earlier this year, however, around $1 trillion in Nvidia’s market cap has been wiped away as investors rally behind CPU makers like Intel. Evolving agentic AI workloads have changed the hardware balance, shifting away from as many as eight GPUs per CPU, toward a one-to-one ratio in some cases.

Nvidia wants to ride that train with its new Vera CPU, which was architected specifically for those types of workloads. Even before the recent rise of agents, however, Nvidia says it has seen demand for its CPUs for data-hungry workloads. “They weren’t running a web server [with Grace]… or they aren’t being used for, what the cloud uses, of cheap, dollar-per-core,” Buck said. “They were being deployed for the backend, data-rich operations, like the data processing.”

Grace represents an on-ramp for Nvidia into data center CPUs. It uses 72 stock Arm Neoverse V2 cores, but it’s differentiated by Nvidia’s Scalable Coherency Fabric (SCF). Vera uses an updated SCF, but it also features Nvidia’s first custom core design, called Olympus. Grace cracked the door, and Vera represents Nvidia's big entrance into the market against AMD and Intel.

Regardless of where Vera ends up in the battle of next-gen data center CPUs — which is heating up now, as AMD is expected to launch its Zen 6 Venice CPUs this week — the design is vastly different from what we’ve seen out of Intel and AMD. Most notably, Vera is monolithic, placing all of its 88 cores on a single piece of silicon. AMD and Intel, years ago at this point, pivoted away from monolithic dies in favor of chiplets, allowing an extremely high density of cores at the cost of latency and coherency issues. Vera is radically different in that regard, not only being built on a single die, but also dedicating significant die space to the fabric.

“One of the reasons we don’t have 128 cores is because we’ve dedicated so much of the die area toward the fabric,” Buck said. “It’s 3.4 TB/s of bandwidth inside of that CPU that is dedicated toward allowing every core to talk to every cache, every memory [controller] at full speed without any collisions.”

For clarification’s sake, Buck is referencing 3.4 TB/s of core-to-core bandwidth in Vera. There’s up to 1.2 TB/s of aggregate memory bandwidth (14 GB/s per core) through the LPDDR5X interface.

But just as chiplet-based designs made trade-offs in per-thread performance, Vera will likely make trade-offs for its unique architecture. The majority of data center workloads are still “legacy” tasks that hyperscalers have built for, and even with seemingly insatiable demand for AI infrastructure, that is unlikely to change for several years.

Buck recognizes this trade-off, asking: “Can Intel and others build rich fabrics? Do they have the IP and the ecosystem to do it and connect it all the way through to LP memory? They need to tell you when they’re going to do it… but that trade-off will come at the cost of the legacy workload.” Earlier this year, at GTC in March, Buck was even more clear. “The world is not going to be served by one SKU of CPU, and that is not our intention,” the executive said in a news conference at the time.

Still, it’s clear Nvidia has ambitions with data center CPUs beyond what headlines are floating around on the New York Stock Exchange. Nvidia says CPUs represent a $200 billion TAM (Total Addressable Market) opportunity for the company, a rather rosy forecast compared to the rest of the industry, which sees a TAM of around $120 billion by 2030 (though recent estimates have climbed as high as $170 billion). And agentic AI is expanding that market, with Morgan Stanley in April estimating that agents could add as much as $60 billion to the data center CPU market.

Vera is in full production alongside Nvidia’s next-gen AI infrastructure, including Rubin GPUs, ConnectX-9 NICs, SpectrumX Ethernet switches, and the various components that go into building a Vera Rubin NVL72 rack. The company says there are around 1.3 million components that go into a rack, and it has a list of over 300 partners globally to build them. As part of our visit to Nvidia HQ last week, we saw a Vera Rubin NVL72 rack in action, running workloads for OpenAI.

AMD is expected to refresh both its desktop and mobile lineups with Zen 6 next year. Ryzen AI 500 series, codenamed Medusa Point, will be the company's next-gen mobile family and we've already seen a 10-core part leak from it a couple of times. Now, that same SKU is back with another Geekbench listing, this time posting its best results that put it ahead of any mobile chip currently made by Intel or AMD (in one way).

AMD's upcoming chip ended up scoring 3,329 points in the single-core test and 16,555 points in the multi-core test. The Geekbench page once again shows the platform name as "AMD Plum-MDS1," which we've known to be associated with Medusa Point for a while now. The SKU is listed as "AMD Eng Sample 100-000001713-33_N," which is identical to how we saw in the last leak, basically confirming we're looking at the same chip.

The single-core result is almost 5% higher while the multi-core score is about 9.7% higher compared to the numbers we saw previously. In fact, the single-core result is better than any mobile part on Geekbench apart from the new Snapdragon X2 Elite lineup. Compared to desktop equivalents, it's very close to the 9900X and the 9800X3D — two absolute desktop powerhouses carrying 12 cores and 8 cores, respectively.

The multi-core score puts this 10-core SKU right alongside the Ryzen AI Max+ 390, which is just a tier below the top-end AI Max+ 395 from Strix Halo. Since this is a 10-core part, we can infer it's the Ryzen AI 9 565 because the Ryzen AI 9 365 (from Strix Point) and Ryzen AI 9 465 (from Gorgon Point) share the same configs. The Ryzen AI 9 365 has an average score of 12,410, so Medusa Point's offering is 33% faster in this regard.

The Geekbench listing also reports the correct clock speeds this time — 2.0 GHz is still lasted as the base frequency, but the boost frequency is marked at 5.37 GHz, which makes a lot more sense. Moreover, the listing shows 10MB of L2 cache and 32MB of L3 cache while the Ryzen AI 9 465 had a combined cache tally of 34MB.

Overall, this benchmark is an incredibly promising showing for Medusa Point. The Red Team's next-gen mobile lineup seems to benefit heavily from an architectural jump thanks to the Zen 6 silicon. Keep in mind that this is the company's mainstream mobile offering that's supposed to compete with Intel's Panther Lake, while Nova Lake mobile and AMD's own Gator Range (Ryzen 10000) will serve only as high-end CPUs.

The Zilog Z80 has just turned 50 years old. This iconic 8-bit processor first went on sale in July 1976 and stayed in production for 48 years until Zilog, now a Littelfuse subsidiary, stopped accepting orders in June 2024. However, there’s an open-source replacement closer than ever to shipping in the chip’s original 40-pin DIP package thanks to community-funded fabrication.

The original Z80 packed 8,500 transistors on a 4μm process and typically ran at 2.5 MHz, with later CMOS variants reaching 20 MHz. Binary compatibility with the Intel 8080 let it absorb the existing CP/M software base, with an on-die DRAM refresh counter that cut the number of support chips a system needed. Development of working prototypes cost roughly $400,000 against $500,000 in funding from Exxon, per the Computer History Museum.

The chip powered the ZX Spectrum, TRS-80, MSX machines, Nintendo's Game Boy, Sega's Master System, the Pac-Man arcade cabinet, and Texas Instruments' graphing calculators, then shipped in industrial controllers for decades after home computing moved on to more powerful successors. Zilog's end-of-life notice, dated April 15, 2024, told customers its wafer foundry was discontinuing support for the Z84C00 family, and last-time-buy orders closed that June.

However, Renaldas Zioma's FOSS Z80 project, launched shortly after the end-of-life notice, now has working silicon. The first version, fabbed on SkyWater's 130nm node through Tiny Tapeout 7 on a die of just 0.064mm², has been confirmed as functional via the project’s GitHub repository. A QFN64 version with all 40 pins exposed followed on the Efabless CI2406 shuttle, two further runs then went through IHP's 130nm process, and the current run targets the classic DIP40 form factor using chip-on-board assembly on GlobalFoundries' 180nm GF180MCU node via Wafer.Space. The end goal here is to fab a drop-in replacement for machines like the ZX Spectrum and RC2014 kits.

The design is built around Guy Hutchison's TV80 Verilog core, and the project's Tiny Tapeout page says the 130nm CMOS implementation should support clocks up to 50 MHz, against 4 MHz for the original NMOS part.

Zilog is trimming the Z80's official successor line as well. A product change notification from last October put the eZ80L92, along with several Z8F-series microcontrollers, on end-of-life, citing "little to no demand." Last-time-buy orders closed on January 20 this year, with shipments scheduled through April 20, on non-cancelable, non-returnable terms. The eZ80L92 is the only eZ80 part named in the notice; the pipelined eZ80 architecture, introduced in 2001 and still inside TI's current TI-84 Plus CE calculators, otherwise remains in Zilog's catalog.

Hobbyists keep finding work for the original chip regardless. For example, earlier this year, a developer ran a tiny conversational AI on a Z80 with 64KB of RAM.

AMD launched the Ryzen 7 7700X3D a couple of days ago to universal acclaim, with reviews praising its consistently excellent gaming performance while critiquing the price point. It's officially priced at $329, which is just shy of what the Ryzen 7 7800X3D goes for these days, rendering the newest X3D chip a bit pointless. However, thanks to a new promo code on Newegg, you can purchase it for as low as $279 right now.

• Get the Ryzen 7 7700X3D on Newegg

Just add the CPU to your cart, and at checkout, the promo code "PKC337" will be automatically applied to give you a $49 discount. That should bring the price down to just $279 before tax. At that price, the Ryzen 7 7700X3D suddenly becomes much more feasible thanks to its improved value proposition. Now, there's at least $120 separating the Ryzen 7 7700X3D from the Ryzen 7 7800X3D, freeing up money you can put toward other upgrades.

For context, both chips are essentially the same; you can look at the Ryzen 7 7700X3D as the binned-down version of the Ryzen 7 7800X3D silicon that couldn't be used for the more expensive SKU. Both are 8-core, 16-thread parts with 104MB of combined cache and 120W TDPs. The Ryzen 7 7700X3D can boost up to 4.5 GHz, while the Ryzen 7 7800X3D has a 5 GHz boost clock. That's enough to just barely edge the Ryzen 7 7700X3D in gaming performance.

On the other hand, if you're looking for an all-rounder that also excels in professional workloads, then Intel's latest Arrow Lake refresh chips are better. Both the Core Ultra 5 250K Plus and the Core Ultra 7 270K Plus are faster at productivity tasks than the Red Team's offerings in the same class. But now that there's a clear price distinction among them, the Ryzen 7 7700X3D makes sense if you're just gaming.

The Ryzen 7 7700X3D is also a Newegg exclusive in North American markets till Q4 2026 so it makes sense that the retailer can discount it this quickly. The Ryzen 7 7800X3D is also on sale, but it has not received any promo codes on the site as of now. So, if you're looking to finally jump on the X3D train and on the AM5 socket, but don't want to spend over $300, this is your chance to grab the Ryzen 7 7700X3D for as low as $279.

Intel was expected to unveil its next-generation desktop processors later this year. The upcoming Nova Lake lineup has been making the rounds online, and a new report from VideoCardz suggests that Intel could introduce it under the Core Ultra Series 400 branding. For context, the current Arrow Lake and Arrow Lake Refresh desktop CPUs follow the Core Ultra Series 200 naming scheme, while Intel's latest Panther Lake mobile processors carry the Core Ultra Series 300 branding.

The report also claims to reveal the review embargo windows and launch timeline for several Nova Lake models. According to the report, Intel will initially introduce a 28-core DS package, which is expected to launch between January and March 2027. The new DS suffix is said to be an internal package designation for processors featuring dual-compute tiles. This will reportedly be followed by 28-core K-series (unlocked) models between March and April 2027, while 16-core and 8-core variants are expected to arrive between late March and May 2027. The flagship 52-core DS model is reportedly scheduled for a much later launch, potentially between late May and September 2027.

While Intel is yet to officially confirm a launch date for Nova Lake, various leaks have suggested that the lineup could be one of the company's biggest generational leaps in recent years. The flagship desktop SKU, featuring a 52-core configuration, is expected to combine 16 Coyote Cove Performance (P) cores, 32 Arctic Wolf Efficiency (E) cores, and four Low Power Efficiency (LPE) cores. This would be a notable jump over the current Core Ultra 9 285K, which features a total of 24-cores. The introduction of Coyote Cove and Arctic Wolf also points to an entirely new CPU architecture, replacing the Lion Cove and Skymont cores found in Arrow Lake.

Nova Lake is also rumored to bring new platform upgrades including support for DDR5-8000 memory, up to 24 PCIe 5.0 lanes for expansion, Thunderbolt 5, and Intel's next-generation Xe3 Celestial integrated graphics. The processors are also expected to feature an upgraded NPU5 for AI workloads along with a 150W Processor Base Power (PBP) and 253W Maximum Turbo Power (MTP) on the flagship model, despite the substantial increase in core count. Earlier reports have also indicated that Nova Lake will transition to a new LGA1954 socket, meaning users will likely need a new motherboard to upgrade from the existing Arrow Lake platform.

Last month, at Computex 2026, AMD unveiled the Ryzen 7 7700X3D — its brand-new 3D V-Cache chip meant to slot in between the 7600X3D and the 7800X3D. As such, it comes in at a suggested retail price of $329 and, surprisingly, is available exclusively at Newegg in North American markets. It's a capable CPU in a vacuum, almost matching the 7800X3D in gaming while being more efficient in some titles, but it still doesn't provide the best value overall given the existence of the 7600X3D.

• Buy the Ryzen 7 7700X3D at Newegg

The 7700X3D and 7800X3D chips share pretty much identical specs. Both are 8-core, 16-thread CPUs based on the Zen 4 architecture, carrying a combined 104MB of cache. They have the same 120W TDP as well, with a 162W max power limit. The difference lies in the clocks, where the 7800X3D can boost up to 5 GHz; the new 7700X3D is limited to just 4.5 GHz.

When it comes to performance, at 1080p, the 7700X3D is actually the third fastest chip we've ever tested, right behind the 7800X3D and the 9800X3D. The stepped-down 7600X3D is very close to the 7700X3D as well, despite being almost $100 cheaper — this is the 7700X3D's biggest downfall: it's only 2% faster than the much cheaper 7600X3D. In our efficiency geomean, the 7700X3D posted the best FPS per watt results we've ever seen, just edging out the aforementioned 7600X3D.

Productivity is where the 7700X3D takes a hit, as it ranks toward the bottom of our charts when it comes to professional workloads. Intel takes the lead in this department, but it doesn't really matter since no one is buying an X3D chip outside of gaming priorities. If you're looking for an all-rounder, we suggest taking a look at the Core Ultra 250K Plus for around $200 or the Core Ultra 270K Plus for over $300.

The Ryzen 7 7700X3D is available across the world, but we can confirm that Newegg is the only place you can buy it if you're in Canada or the United States. This exclusivity runs through till Q3 2026, so the CPU isn't available even on Amazon or Micro Center — the latter of which has exclusivity for the Ryzen 5 7600X3D, Ryzen 5 7500X3D and the older Ryzen 5 5600X3D in America.

As per our review, the 7700X3D comes within 5% of the 7800X3D's gaming performance, but is only 2% faster than the 7600X3D, which you can buy on Amazon for $239. This makes it a tough sell, but if you're set on grabbing one, Newegg is the place to be.

The 7700X3D always made sense. Ever since the 5800X3D released and showed itself as the best CPU for gaming (at the time), AMD has continued to double down on 3D V-Cache, dominating the competition from Intel in games by double-digit margins. Because of the immense success of X3D CPUs, we’ve seen several variations with lower bins. Originally we had the 7800X3D, and now we have the 7700X3D. It’s cheaper and has lower boost clocks, but it gives you the same eight Zen 4 cores and 104 MB of combined L2 and L3 cache. It’s a 7800X3D for less money.

That, at least, is the assumption. Reality is a bit different.

• AMD Ryzen 7 7700X3D available exclusively at Newegg

A little less than two years after the 5800X3D released, AMD introduced the 5700X3D. Like the 7700X3D, it came with a cut to maximum boost clocks (400 MHz with the 5700X3D instead of 500 MHz here, but we’re splitting hairs), but still largely offered the gaming performance of the 5800X3D for less money. The problem here is that, although the 7700X3D could be a worthy successor to the 5700X3D, it’s too expensive.

The 5800X3D released in April 2022 for a suggested price of $450. Flash forward to January 2024, and the 5700X3D rolls out at $250. The 7800X3D launched in April 2023 for $450. The 7700X3D is arriving more than three years later in July 2026 for a suggested retail price of $330. It’s safe to call the 7700X3D a day late and a buck short, even ignoring the external pricing circumstances of the DIY market now.

That’s just a high-level analysis of launch pricing, too. Looking at prices now, the comparison is even more rough. Intel’s Core Ultra 7 270K Plus is the same price, within 5% of average gaming performance, 2X multi-threaded performance, and around 40% faster in single-core performance. AMD’s own Ryzen 7 7800X3D is, at the time of writing, available for $349, just $20 more than the 7700X3D (though I suspect that price will change). Buy a secondhand 7800X3D from Amazon, and it’s cheaper than the 7700X3D.

And, if you’re just focused on gaming performance and getting the best bang for your buck, the Ryzen 5 7600X3D is around $100 cheaper than the 7700X3D and within 2% of the average gaming performance.

The 7700X3D performs exactly how I expected it to. It’s not as fast as the 7800X3D, but if you squint hard enough, it’s close enough. It’s just too expensive. At $330, you’re almost forced to step up or down to AMD’s other Zen 4 X3D chips to get into a value sweet spot, and if you’re not solely focused on gaming, Intel offers much more powerful CPUs around the same price.

If the 7700X3D followed in the 5700X3D’s footsteps and released at $250 (even after three years of the 7800X3D on the market), it’d be a slam dunk. That’s not where we are for release, so let’s hope a price cut is waiting in the wings.

AMD Ryzen 7 7700X3D specifications and pricing

There’s a good chance the Ryzen 7 7700X3D is the last X3D processor we’ll see sporting AMD’s Zen 4 architecture, short of a potential Ryzen 5 7500X3D in the future. It further segments AMD’s last-gen lineup, and although there’s good pricing separation between each of the Zen 4 options, the spec differences are small.

For the Ryzen 7 7700X3D, the only spec difference it carries is a cut to clock speed compared to the Ryzen 7 7800X3D. You lose 200 MHz on the base clock and 500 MHz on the maximum boost clock. It’s a similar setup to what we saw with the Ryzen 7 5800X3D and 5700X3D, just a bit more aggressive. The 5700X3D shaved 400 MHz off the base and boost clock of the 5800X3D.

Otherwise, you’re getting the same eight Zen 4 cores available in the 7800X3D, along with 104 MB of combined L2 and L3 cache (64 MB of that L3 is stacked on the CCD). It also comes with the same rated TDP of 120W, though as we’ll see throughout our benchmark results, the 7700X3D never crossed into triple-digit wattages during our testing.

The 7700X3D slots into existing 600- and 800-series motherboards, and AMD says that it will boot on existing AM5 BIOS images (though the company recommends flashing the latest firmware). We didn’t need to install a specific BIOS image or chipset drivers to boot with the 7700X3D, so if you already have an AM5 motherboard, you should be set. In addition, the 7700X3D does not come with a stock cooler, despite arriving in AMD’s larger box design that we saw in the previous generation.

Outside of the step up to the 7800X3D, there’s a step down to the 7600X3D, which trades two cores (and consequently 2 MB of L2 cache) for a bump to a 4.7 GHz boost clock. As we’ll get to in our gaming benchmarks, there’s something about the range of 4.5 to 4.8 GHz where these Zen 4 X3D chips hit their stride, and the 7700X3D just barely hits that range without PBO assistance.

The 7700X3D arrives at a recommended retail price of $330, putting it in hotly contested waters. Intel has its newer 270K Plus around that same price, while the last-gen Core i7-14700K lands closer to the $380 mark. Down a step, the 250K Plus is more than $100 cheaper at $220, while the Core i5-14600K is available for around $250.

For AMD, you can step up to the 7800X3D for a $50 premium (at current prices) or down $90-$100 and get the 7600X3D. Of this tight grouping of Zen 4 X3D chips, none of them are bad options purely for gaming. Deciding between them is tricky. We’ve seen the 7800X3D on sale for as low as $348, which is a negligible price difference compared to the 7700X3D. And even at list price, the Ryzen 5 7600X3D is significantly cheaper, yet comes within just two points of the average gaming performance of the 7700X3D.

It’s impossible to give a concrete conclusion about which is the best because even a minor sale of $20 or $30 off tips the scales. The 7700X3D isn’t a bad processor, but there are a lot of situations where it’s not the optimal choice, mainly due to its proximity in price to the 7800X3D. Given the 7700X3D’s performance, it would ideally be priced around $260 to $280.

You’ll spend much more if you want to get one of AMD’s latest X3D chips with the Zen 5 architecture, which represent somewhere around a 15% to 20% improvement in average gaming performance. The 9800X3D is the cheapest Zen 5 X3D processor right now, and you can expect to spend about $450 to $480 on one. AMD suggested to Tom’s Hardware that it’s looking into a cheaper Ryzen 5 9600X3D for a future release, but that’s not available at the moment.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

The Ryzen 7 7700X3D puts to bed any notion that clock speed is irrelevant in a modern gaming PC. The chip is sandwiched between two other X3D chips: the coveted Ryzen 7 7800X3D and the less-considered Ryzen 5 7600X3D. The 7700X3D shares its DNA with the 7800X3D, both sporting eight Zen 4 cores and 104 MB of combined L2/L3 cache. The 7700X3D just shaves 200 MHz off the base clock and 500 MHz off the boost clock.

As usual, we tested a suite of modern games at 1080p with a mixture of High and Ultra settings, and without upscaling or frame generation enabled. We use the RTX 5090 on our CPU test bench to remove any GPU bottlenecks and isolate CPU performance as much as possible.

Given that we have eight cores to play with, I expected the 7700X3D’s gaming performance to land closer to the 7800X3D, not the 7600X3D. That’s not the case. The 7800X3D is 4% faster than the 7700X3D on average in games, while the 7700X3D is just 2% faster than the 7600X3D. You don’t give up much performance with the 7700X3D compared to the 7800X3D, but on the other hand, you don’t gain much performance compared to the 7600X3D, despite a $90 to $100 difference in price.

The Ryzen 7 9800X3D remains the fastest gaming CPU on the market, outside of the better-binned Ryzen 7 9850X3D, and it’s 19% ahead of the 7700X3D. Looking backwards a generation, the re-released Ryzen 7 5800X3D at $350 is in a tough spot. The 7700X3D is nearly 20% faster despite arriving on the market for $20 less. The impact of memory shortages has created some bizarre value comparisons, especially for CPUs.

Breaking out of the X3D bubble, Intel’s latest Core Ultra 7 270K Plus remains potent competition. The 7700X3D is ahead by 5%, and slightly less compared to the Core i7-14700K, which isn’t as wide of a margin as we’re used to seeing with new X3D chips. Particularly when bringing application performance into the mix, which we’ll get to next, the Core Ultra 7 270K Plus justifies a single-digit drop in gaming performance for how much it gains elsewhere; that’s assuming, of course, that you aren’t going for a pure gaming rig.

With Raptor Lake Refresh and Alder Lake chips, we tested with DDR5 memory (you can read more about our testing procedure later in this review). We tested DDR4 memory recently on these chips in our Ryzen 7 5800X3D re-review. That data is excluded here to keep the charts readable, but you shouldn’t expect miracles with DDR4. Based on our results, Raptor Lake Refresh chips with DDR4 are in the low single digits behind the Ryzen 7 5800X3D.

Outside of frame rate, you can see that the Ryzen 7 7700X3D landed squarely on 4.5 GHz in our gaming tests. Interestingly, the Ryzen 7 7800X3D locked in the same average as the 7600X3D at 4.7 GHz, despite the former boosting up to 5 GHz. The Ryzen 7 7700X3D also arrived at the lowest average power consumption, despite carrying a 120W TDP, clocking in just 0.1 W behind the 7600X3D.

The power consumption figures are interesting. The 7700X3D carries the same TDP as 7800X3D, but real-world power consumption is closer to the 7600X3D, which has a 65W TDP. And that lowered power consumption helps temperatures, with the 7700X3D clocking a cozy 55-degree Celsius average.

There’s headroom here in temperatures and power consumption for AMD’s Precision Boost Overdrive (PBO) to close the gap with the 7800X3D (but, then again, that chip has access to PBO, as well). We manually disable PBO in our testing, as it can push the chip outside of AMD’s default specifications, and in turn, void your warranty. If the warranty isn’t a concern for you, however, the 7700X3D should take nicely to a PBO bump.

007 First Light Benchmarks

007 First Light is the newest game in our test suite, and despite IO Interactive’s Glacier engine taking particularly well to X3D CPUs in Hitman 3, we see a shift toward Intel here. The 270K Plus is 7.6% faster than the 7700X3D, and that’s without optimization for Intel’s new iBOT feature. We can also see slightly weaker 1% lows on the 7700X3D compared to the other Zen 4 X3D chips, though nothing that completely changes the perceived smoothness.

Baldur’s Gate 3 Benchmarks

Baldur’s Gate 3, on the other hand, sees a nice boost with 3D V-Cache, as evidenced by the fact that the 5800X3D nearly matches the much newer 9700X (though with worse 1% low performance). The 7700X3D is in lockstep with the 7600X3D, and just shy of 6% behind the 7800X3D. Compared to the 270K Plus, the 7700X3D handily earns a victory with a 16.7% lead in average performance.

Looking at clock speeds, you can see Baldur’s Gate 3 gets much more out of these chips than First Light, all without an increase in power consumption. That leads to some exceptional efficiency results, with the 7700X3D offering two and a half frames for every watt consumed.

Crimson Desert Benchmarks AMD Ryzen 7 7700X3D

Crimson Desert is another recent addition to our test suite, and it shows remarkable CPU scaling in our benchmark of the dense city streets in the Hernand town center. Breaking from our other results, the 7700X3D is 6.9% ahead of the 7600X3D here, and just 1.3% behind the 7800X3D. The efficiency here is second to none at nearly three frames per watt. The 7700X3D is offering 7800X3D-like performance at 7600X3D-like power, which is the best-case scenario for this CPU.

Counter-Strike 2 Benchmarks

Counter-Strike 2 offers a closer look at the importance of 1% low performance, with our averages exceeding 600 FPS across most of the test pool. The 7700X3D is just 1.1% faster than the Core i7-14700K on average, but we can see a 24.7% jump in 1% low performance. That’s a common theme among all the X3D chips we tested in this game. That smoothness is important here. With such high average frame rates, swinging 100 FPS (or more) in either direction isn’t uncommon.

Cyberpunk 2077 Benchmarks

Cyberpunk 2077 shows a much tighter contest in average performance. Here, the 7700X3D is in lockstep with the Core Ultra 7 270K Plus and 14700K. Interestingly, the 7600X3D is a little faster here, though by less than a frame compared to the 7700X3D. Flipping over to our clock speed geomean, we can see what’s going on. Most of the chips in our test pool pushed up toward their maximum boost clock in this game, leaving the tame boost clock of the 7700X3D behind.

We’re looking at very tame power usage here, with an average of just 67 watts, and plenty of thermal headroom. PBO would help the 7700X3D push out a small lead over the Intel competition.

Doom: The Dark Ages Benchmarks

Doom: The Dark Ages is the only game in our suite using the Vulkan API, and it features always-on ray tracing. However, we still see CPU scaling even with such a GPU-focused pipeline. The 270K Plus squeezes out a marginal lead over the 7700X3D, but AMD’s chip is a bit more consistent in 1% low performance. The 250K Plus is especially impressive here, offering average performance on the level of $300+ CPUs for just $220.

As we can see from our clock speed results, the CPUs we tested ran comfortably below maximum boost clocks, suggesting the game is heavily threaded. As a result, the 7700X3D claims a clean 5% lead over the 7600X3D, nearly matching the 7800X3D.

F1 2024 Benchmarks

F1 2024 is an interesting benchmark because it provides yet another example of the 7600X3D outclassing the 7700X3D. The performance is close enough to call it identical, but I ran five passes on this game on both the 7600X3D and 7700X3D, and came away with the same results. Regardless, F1 2024 shows a massive advantage toward X3D chips, with even the lowly Ryzen 7 5800X3D beating every non-X3D chip in our test pool.

Far Cry 6 Benchmarks

Far Cry 6 mirrors what we saw in Doom, with the 7700X3D, 7600X3D, and 270K Plus all arriving around the same average frame rate. We don’t see an upper ceiling like in Doom, however, allowing the 7800X3D to claim a lead of 6.9%, and the 9800X3D shooting ahead to a 30% lead.

The power use for the 7700X3D here is especially low, falling behind the 7600X3D by about 6%. There’s probably some small, untapped optimization here that would help the 7700X3D close the gap with the 7800X3D.

Final Fantasy XIV Benchmarks

Final Fantasy XIV is another title that loves 3D V-Cache, and even among X3D CPUs, the game shows big jumps for newer architectures. The 7700X3D is 21% ahead of the 270K Plus here, but all of the Zen 4 X3D chips are tightly grouped between 175 and 180 FPS. In a blind shootout, it’d be impossible to tell between them.

Once again, the efficiency of the 7700X3D is remarkable at over three frames per watt consumed. The chip drew just 55.7 watts on average during our test, 32% lower than the 270K Plus and in line with the 7600X3D.

Flight Simulator 2024 Benchmarks

The Ryzen 7 7700X3D slips in the rankings in Flight Simulator 24, clocking in a frame behind the 250K Plus and 4.6% behind the 270K Plus. The Ryzen 7 9800X3D enjoys a comfortable lead, but there’s very little difference between AMD’s Zen 4 X3D offerings. They remain the most efficient of the bunch, however, with average power consumption below 60W.

Hogwarts Legacy Benchmarks

The 7700X3D once again trails Intel in Hogwarts Legacy, with the 270K Plus offering a 9% bump in average performance. The 7700X3D is a marginal 1.5% faster than the 7600X3D in this game, and it trails the 7800X3D by 3.8%. The 9800X3D, meanwhile, is a staggering 29% faster than the 7700X3D.

Marvel Rivals Benchmarks

The Unreal Engine 5-based Marvel Rivals remains the most popular hero shooter on Steam, but despite its technical backbone, the game shows clear CPU scaling at 1080p. As expected, the 7700X3D is marginally faster than the 7600X3D, and 5.1% behind the 7800X3D. Similar to Hogwarts Legacy, Intel holds an advantage with Arrow Lake refresh, though at significantly higher power draw and worse efficiency as a result.

Minecraft RTX Benchmarks

Minecraft is easily the most difficult benchmark for Intel. We’ve seen consistently low performance out of Arrow Lake CPUs in this test. Note that our Minecraft test uses a render chunk distance of 96, specifically stressing the CPU and memory chain to it.

Spider-Man 2 Benchmarks

Spider-Man 2 shows the 270K Plus once again besting the 7700X3D, this time by 4.9%. The 7800X3D also offers around a 5% jump. We can again see clock speed acting as a big influence in this test, but that extra clock speed comes at the cost of increased power draw, with the 270K Plus nearly doubling the average wattage of the 7700X3D in this test.

Starfield Benchmarks

Starfield is one of the few examples where we can see a clear benefit of the 7700X3D over the 7600X3D, with the former posting a 6% lead. The 270K Plus is slightly ahead of the 7700X3D, with a 3.6% lead, while the 7800X3D pushes further with a 6.7% jump in average performance.

The Last of Us Part One Benchmarks

Rounding out our game suite is The Last of Us Part One, where both the 270K Plus and 7800X3D are around 6% faster than the 7700X3D. AMD’s latest CPU is 5% faster than the base Ryzen 7 7700X, and 4% faster than the 7600X3D. You can see AMD’s non-X3D chips running into a performance wall, with the 7700X and 9700X posting virtually identical results.

This is one of the few games where the 7700X3D consumed more power than the 7600X3D, though only with a 5% bump to average wattage. Still, that’s enough for the 7600X3D to steal the top slot in efficiency away from the 7700X3D, which sits at the top of the efficiency rankings in most other titles.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

AMD’s X3D chips are built first and foremost for gaming, with only niche (and expensive) chips like the recent Ryzen 9 9950X3D2 pulling double duty in gaming and productivity applications. The hindered clock speed of the 7700X3D means it struggles even moreso than the 7800X3D in lightly- and heavily-threaded applications, which is already an area of weakness.

In our multithreaded geomean, the 7800X3D is 6.9% ahead of the 7700X3D while the base Ryzen 7 7700X is 12.8% ahead. AMD’s latest non-X3D eight-core, the Ryzen 7 9700X, is around 13% ahead with its default 65W TDP and a massive 28% ahead in its 105W TDP mode. The Ryzen 7 9800X3D pushes things even further with a 37% lead.

Short of the Ryzen 5 7600X3D, which the 7700X3D leads by a clean 25%, AMD’s latest X3D chip is at the bottom of the pile for multithreaded performance, at least among our DDR5 offerings. The comparison isn’t great among AMD’s offerings, but it's far worse among Intel’s.

The Core Ultra 7 270K Plus is 130% ahead, more than a 2X increase. We’re comparing radically different architectures and core counts, but the 270K Plus is leaps and bounds faster than the 7700X3D in multithreaded applications, and within 5% in games. Even the $220 Core Ultra 5 250K Plus is 70% ahead in our multithreaded geomean. The 7700X3D wins for gaming, even if its lead is small. But if you even just dabble in apps like Handbrake, Blender, and DaVinci Resolve, you’re giving up a lot of performance at this price with the 7700X3D.

The margins are always tighter in our single-threaded geomean, but the limited clock speed of the Ryzen 7 7700X3D means it ends up below every other DDR5 option in our test pool. The 7800X3D is 10% ahead, while the 9800X3D pushes ahead with a 25% lead. The base Ryzen 7 7700X also leads by a clean 20% margin.

In Intel’s camp, the Core Ultra 7 270K Plus is 42% ahead, while the 250K Plus is 33% faster. The Core i7-14700K and Ryzen 7 9700X are in lockstep, both beating out the 7700X3D by around 29%.

Single-threaded performance is especially limited on the 7700X3D, not only due to lower maximum boost clocks, but also the SRAM stacked on top of the CCD. It acts as an insulating layer between the CCD and IHS, giving you minimal headroom for large clock speed boosts on a single core, even with manual overclocking and robust cooling.

Rendering Benchmarks

In demanding rendering workloads like Cinebench, Blender, and POV-Ray, the 7700X3D struggles. The 4.5 GHz boost clock limits single-core performance, while the underlying Zen 4 architecture can’t scale up to AMD’s modern eight-core offerings. And, bringing Intel into the mix, strong single-core performance and Intel’s hybrid architecture with large core arrays allow Team Blue to sit near the top of the chart in most of our testing.

In Cinebench 2024, the 270K Plus was 135% faster than the 7700X3D in a multi-core render. The base 7700X, meanwhile, is 10% faster and the 9700X with its 105W TDP is 21% faster. Single-core performance shows big gaps, as well. The 7600X3D is faster than the 7700X3D with its higher boost clocks, while the 270K Plus offers around a 39% boost in performance.

There are even bigger gaps elsewhere. In POV-Ray, the 270K Plus is nearly three times as fast as the 7700X3D in the multi-core test. In Blender, the 7700X is 11% faster than the 7700X3D, and the 9700X 24% faster, in the Monster scene. And in V-Ray 6, even the $220 Core Ultra 5 250K Plus is 66% faster than the 7700X3D.

In heavily-threaded workloads, the 7700X3D and 7800X3D offer very similar performance, with a slight edge to the latter. We only see a big divergence between them in single-threaded workloads.

Encoding Benchmarks

Alongside rendering, our encoding benchmarks factor heavily into our geomean. We can also see a repeat here of what we saw in our rendering benchmarks, particularly among heavily-threaded video encoders like Handbrake. Across codecs, the 270K Plus is usually twice as fast (and sometimes more), while the non-X3D 9700X offers around a 30% boost depending on the codec. Compared to the 7800X3D, the 7700X3D is 7% slower with x265 and AV1.

In single-threaded audio encoding via LAME, the 7800X3D was nearly 9% faster than the 7700X3D, and that gap grows to nearly 10% in our extended LAME results.

Surprisingly, there’s a decent gap between the 7700X3D and 7800X3D in our image encoding/decoding benchmarks. In our multi-threaded JPEG-XL decode, the 7800X3D is 7% faster, while the 9700X is 26% faster. Similar margins appear in the encode.

Creator App Benchmarks

Creator applications like the Adobe suite and DaVinci Resolve are some of the most important for the 7700X3D; if you’re building a PC for gaming, the most likely non-gaming workload is some sort of video or photo editing application. These applications feature a ton of workloads spanning heavily-threaded and lightly-threaded tasks, so the margins between chips are much tighter.

Starting with Photoshop, AMD holds a strong position in this application across all of its CPUs, 7700X3D included. It matches the 14700K and closes in on the performance of the 270K Plus. The scales turn toward Intel in Premiere Pro, with the 270K Plus outperforming the 7700X3D by 9.5%. However, the 7800X3D is only a meager 1.9% ahead. We can see a similar situation in DaVinci Resolve.

Rounding out our tests is After Effects, where the 7800X3D is around 5% faster than the 7700X3D, and the 270K Plus is more than 30% faster.

Web and Office Benchmarks

General-purpose web and office workloads are usually lightly-threaded, so the 7700X3D struggles in these benchmarks. However, all of the CPUs in our test pool are more than capable of running these workloads, so although the 7700X3D often ends up near the bottom of the pile, the performance difference in real-world use isn't as big as the numbers would suggest.

In web-based workloads measured via WebXPRT, you can see the 7700X3D only manages to outclass AMD’s DDR4 options in our test pool, even falling short of the Core i7-12700K. Those single-core speeds really put a damper on performance here, though the extra two cores on the 7700X3D allowed it to achieve 15% better application start-up time than the 7600X3D in PCMark 10.

Microsoft Word and Outlook show similar performance differences, but the 7700X3D gains back some places in Excel and PowerPoint. A strong performance in Excel is important, but PowerPoint is tricky. As you can see from our results, the 7700X3D actually outperformed the 7800X3D; that’s just a consequence of PowerPoint not being a demanding application to run in most situations.

Chess Engines, Compilation, Compression, AVX, and Other Benchmarks

Rounding out our application tests are a series of highly specific workloads. We have a mixture of practical workloads like LLVM code compilation and compression/decompression with various algorithms, as well as some tests like various chess engines that drill down on IPC and applications like Optcarrot, which is an NES emulator looking at Ruby application performance.

Depending on what you plan to do with your PC, many of these workloads might not be applicable to you. Because of that, they aren’t included in our geomean, as some CPUs scale especially well in these niche workloads in a way that isn’t representative of overall performance.

SPEC Workstation 4 Benchmarks

Similarly, we run SPECWorkstation 4. We already run a subset of the tests included in SPEC in our own benchmark suite, but our full SPEC results give you numbers for the particular benchmark configuration used in the suite, as well as some additional scientific and security results.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

The 7700X3D is one of the most efficient gaming CPUs on the market, even offering more performance per watt than the 7800X3D. It’s highly optimised for that workload. Outside of games, power consumption remains very low, often below 80W even in all-out heavily-threaded tasks. Performance also slips, however, leading to worse efficiency.

Starting with raw power consumption, the 7700X3D averaged just 74W during a multithreaded Cinebench 2024 run, matching the 7600X3D. Again, the 120W rated TDP here is interesting, as the 7700X3D isn’t pushing past what the 7600X3D demands. In Blender and Handbrake, we can see the two chips in lockstep, as well.

The 7700X3D justifies its higher TDP a bit more in Linpack, where it drew 74W to the 7600X3D’s 65W. It’s worth highlighting our single-threaded y-cruncher pass, as well. In this test, the 7700X3D drew less power than the 7600X3D.

Outside of heavy workloads, we also measure idle and active idle (YouTube playback) power, and the 7700X3D does surprisingly well in these tests. As you can see from our two Zen 3 chips, AMD’s idle power consumption increased massively with a switch to a DDR5 platform with Zen 4. The 7700X3D pulls that idle power consumption back significantly, even compared to the 7600X3D and 7700X3D.

Turning to efficiency, you can see that the 7700X3D remains one of the most efficient options out of our test pool, though the margins are much thinner. In games, the X3D stack blows everything else away, but in Handbrake, we can see the 9700X offering similar efficiency. More interesting is the 270K Plus. The 7700X3D is 33% more efficient in our Handbrake x256 encode, but the two CPUs are in completely different performance classes.

When there’s such a wide disparity in application performance, these efficiency numbers can look a bit skewed. Cinebench provides a good example of that. The 7700X3D was nearly 40% more efficient than the 270K Plus, but the 270K Plus is around two and a half times as fast in this test. The 270K Plus and 9800X3D offer identical efficiency metrics, on the other hand.

A clearer way to visualize that is with a scatterplot. You can see in our Blender scatterplot, for example, a tight grouping of AMD’s processors around the bottom left of the chart, noting great efficiency but weaker performance. There are a few CPUs that hit an efficiency sweet spot, most notably the 9800X3D and 250K Plus.

Test Setup

We use a frozen test image and nearly identical test benches across the platforms in our pool. You can see the exact configuration we used for testing below. All of our tests are run on the same stack, including the OS build, chipset drivers, GPU drivers, and application versions. Particularly in game testing, where new updates are released constantly, we retest every chip in our test pool to validate our results, opting for the latest data for each CPU.

Our AMD and Intel test images never mix, so there aren’t remnants of AMD drivers on an Intel platform or vice versa. Across both vendors, we enable EXPO/XMP, turn off Virtualization-Based Security (VBS), and enable ReBAR. We also don’t enable any motherboard-specific optimizations, such as Gigabyte’s X3D Turbo Mode, as that adds an uncontrolled variable to our testing (and can sometimes even hurt performance).

On that thread, we also explicitly disable any performance profiles or optimizations that push the processor outside of warrantied operating specifications. That means running with Intel’s default performance profile (enforced power limits) and disabling PBO. Both push the processor out of the warrantied specifications.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

It’s surprising we didn’t see the 7700X3D sooner. AMD has told us that the 7700X3D (and the re-released 5800X3D) are both targeted releases to ease the burden of high DDR5 prices, but just like with the 5800X3D, it feels like AMD could’ve done more to ease that burden. The 7800X3D has already approached the MSRP of the 7700X3D with sales, and the 7600X3D has clearly come out on top as the value-focused option among AMD’s Zen 4 X3D CPUs.

It’s a tough pricing situation regardless, a consequence of continuing to bin and release various versions of largely similar silicon. Still, the 7700X3D has a pricing window. That window just doesn’t exist at $330. At $300, it would be more competitive with the rest of the market, and at $280, it’d be a difficult CPU to contend with. At $250 like we saw the 5700X3D, it’d be a no-brainer. $330 is the maximum price where you might be able to justify the 7700X3D, and that’d only be if you completely ignore CPUs going on sale and wipe out Arrow Lake Refresh from your memory.

The two CPUs that remain the most potent competition are Intel’s 270K Plus and the 7600X3D, not the 7800X3D. On sale, the 7800X3D is the better buy, full stop, but if the 7700X3D is targeting gamers that want to stretch their dollar the furthest, the 7600X3D offers a lot more value.

You can see that clearly in our gaming geomean. The 7600X3D is just 2% slower than the 7700X3D on average, and even in extreme situations, the performance gap between them never approaches double digits. The 7800X3D and 7700X3D offer a similar value at list price, which is about half a frame per dollar. Meanwhile, the 7600X3D offers three-quarters of a frame per dollar.

The value isn’t bad with the 7700X3D, make no mistake. It’s in line with the 270K Plus and 7800X3D, but it shouldn’t be. It’s a value-focused alternative to the 7800X3D, but it doesn’t provide much additional value. At around $280, it’d shoot up in the value rankings and become much easier to justify.

On the other end of the spectrum is Intel’s 270K Plus. It’s 5% slower in average gaming performance than the 7700X3D, but it makes up for that small gap with more than twice the multithreaded performance of the 7700X3D. Even if you only commonly use one non-gaming application, the performance uplift of the 270K Plus is large enough to justify a small hit to gaming performance.

With most 3D V-Cache CPUs, the drop-in application performance is easy to justify, even if the boost in gaming performance can’t keep pace. You buy an X3D chip primarily for playing games, not compiling code, running Fourier Transforms, or building a web server. Here, however, the margins in gaming are some of the smallest we’ve seen between X3D and non-X3D CPUs, while the margins in applications are some of the largest.

There’s three paths away from the 7700X3D right now. The 270K Plus offers a much more well-rounded CPU around the same price. The 7600X3D delivers the gaming value that the 7700X3D is sorely lacking, and the 7800X3D is only slightly faster in games, but it’s also only slightly more expensive. Hopefully we’ll see the 7700X3D drop to between $250 and $280. At that price, it’s a CPU worth considering. At its current price, it’s hard not to go with another option.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

Intel this week announced that it will invest €5 billion ($5.7 billion) to expand and modernize its manufacturing operations at the company's facility near Leixlip, Ireland. The project is intended to increase production capacity for Intel Xeon 6 processors and next-generation Intel Xeon products built using the Intel 3 fabrication process (3nm-class), as well as advanced research and development (R&D) activities at the site.

The upgrade of the facility will involve installation of new tools presumably at Fab 34 as well as extensive infrastructure improvements designed to increase manufacturing efficiency. One of the key elements of the project is the expansion of the campus' automated material transport network, which will connect separate manufacturing modules into a single high-speed production system. Meanwhile, the modernization will not involve cleanroom expansion. Intel expects the upgrade to enable the Leixlip site to produce larger volumes of Intel 3-based products and make better use of the existing cleanroom capacity.

Intel opened its Fab 34 near Leixlip, Ireland, in 2023 and has been making various chips — including Core Ultra 100-series using Intel 4 and Xeon 6 using Intel 3 production node — using its process technologies that rely on EUV lithography at the site. At present, Intel's Fab 34 is Europe's only high-volume semiconductor production facility that uses EUV tools.

In mid-2024, Intel announced the €10.1 billion sale of a 49% stake in Fab 34 with Apollo Global Management as it badly needed money. This April, the company announced that it would repurchase the 49% stake in Fab 34 for $14.2 billion, which opened doors to the current expansion and investment. Intel claims that it kicked off execution of the project earlier this year, though it did not disclose when the upgrades will be completed.

"By investing in our existing fabs with state-of-the-art technology and installing cutting-edge tools, we are not just increasing output of critical products like Xeon 6 and next gen Intel Xeon processors built on Intel 3, we are ensuring that Ireland remains at the forefront of the world's most advanced manufacturing ecosystems, while strengthening the region’s role in the global technology landscape," said Naga Chandrasekaran, Executive Vice President, Chief Technology and Operations Officer and General Manager of Intel Foundry.

Among other things, Intel says that the investment will strengthen Europe's semiconductor supply chain and support the European Union's technology sovereignty objectives by increasing domestic production of leading-edge CPUs. There is a catch about that claim, though. All the silicon produced in Ireland is transported back to the U.S. for testing and assembly, as well as makes the end products, such as Core Ultra or Xeon 6, 'made in America.'

AMD continues to squeeze more life out of its AM4 platform as a newly listed prebuilt gaming PC has been spotted powered by the Ryzen 7 4700LE. This OEM-exclusive CPU is based on AMD's Zen 2 (Renoir) architecture, all the way back from 2019, and was silently released by the company back in March 2026. The prebuilt gaming PC featuring the CPU is currently listed by Chinese system integrator Qehi on Amazon with a price tag of $799.99. It additionally comes with an Nvidia RTX 3050 8GB graphics card, 16GB of DDR4 memory, and a 512GB M.2 NVMe SSD in a fish-tank style chassis loaded with six RGB fans.

Glancing over the specifications, the Ryzen 7 4700LE comes with eight cores and 16 threads, along with a maximum boost clock speed of 4.2 GHz. The chip also packs 12MB of total cache and a rated TDP of 65W, meaning that it generates less heat and requires a less demanding cooling solution. This should make it suitable for small form factor builds, although one should note that it does not come with onboard graphics, thus relying on a discrete GPU.

In a similar move, AMD had announced the return of the Ryzen 7 5800X3D last month as a special 10th Anniversary Edition, giving its popular AM4 gaming processor a second lease on life. By doing so, the company not only offered gamers a potent yet affordable CPU amid rising component prices, but it also allowed existing AM4 users to upgrade their CPUs without switching to a whole new platform.

While it won't rival AMD's modern processors, the Ryzen 7 4700LE should still be capable of handling everyday workloads and potentially some modern games when paired with the right GPU. OEMs and system integrators additionally gain benefits by making use of existing AM4 motherboards and DDR4 memory inventory, allowing them to build systems at a much lower cost compared to AM5-based systems. The CPU could also help make entry-level gaming PCs more accessible by giving budget-conscious gamers another option at a time when pricing for components like RAM, SSD, and GPUs continues to rise, thanks to the AI boom.

Intel is working on a version of its Nova Lake-S processor platform codenamed Dunlow that will offer up to 28 cores and will target entry-level server and workstation applications, according to shipment manifests located in the NBD database by @x86deadandback.

Formally, Intel's codenamed Dunlow platform will succeed the company's Catlow platform with Xeon 6300P-series CPUs and will support Xeon E-class Nova Lake-S processors (presumably) with up to 28 cores that feature a dual-channel memory subsystem, come in an LGA1954 form-factor, and have a processor base power of 95W, according to shipments manifests at NBD data.

Intel's next-generation Core Ultra 400-series platforms for desktop computers, codenamed Nova Lake-S, allegedly feature up to 52 cores, which include up to 16 high-performance Coyote Cove cores and up to 32 energy-efficient Arctic Wolf cores in the compute tile, as well as four low-power Arctic Wolf cores presumably in the SoC tile. These Nova Lake-S CPUs are aimed at enthusiasts and reportedly pull up to 474W with a single purpose: to offer unbeatable performance and feature set to put Intel back on the map of enthusiast-grade platforms currently dominated by AMD.

By contrast, the Dunlow platform seems to be a completely different kind of animal. The CPU deliberately features 28 cores and up to 95W PBP (TDP). All Xeon processors except Xeon 6700E, Xeon 6+, and some Atom-based solutions for specialty applications released to date have only featured high-performance cores. Even Intel's Xeon 6300P-series 'Raptor Lake-E' based products feature up to 12 P-cores to offer higher sustained all-core frequencies. Therefore, unless Intel plans to offer energy-efficient cores in its next Xeon CPU aimed at entry-level servers and workstations, we may be dealing with a very special processor that features 28 P-cores that is designed to beat all desktop-grade platforms in demanding applications.

While, for now, 28 P-cores inside Nova Lake-S processors for the Dunlow platform is speculation, it should also be noted that 28 cores do not naturally derive from a 16P+32E desktop design and are impossible to derive from a notebook-grade 8P+16E design. Also, Intel typically does not create server/workstation products by fusing off nearly half a desktop die (it does not even matter whether it disables some P-cores and some E-cores, disabling 20 cores in a 48-core tile hardly makes a lot of sense).

A Nova Lake-S CPU for Dunlow featuring a compute tile with 28 P-cores would resemble the abandoned Raptor Lake-32C, which featured an all-P-core design aimed at workstations and entry servers before being canceled. It is also possible that this could be a derivative of a small Xeon die adapted to an LGA1954 packaging and dual-channel memory to reduce platform costs. At the end of the day, many server applications like storage or web hosting do not need extremely high memory bandwidth, so two DDR5 channels could be enough.

Another reason for Intel to release a Nova Lake-S CPU with up to 28 P-cores is to fill the gap between high-end enthusiast-grade desktops that feature up to 16 P-cores and expensive Xeon 6 server and workstation CPUs that may start at 16 cores, but feature an octa-channel memory subsystem that is costly and is an overkill for many applications. Also note that since Xeon 'Diamond Rapids' processors with an octa-channel memory subsystem have been canceled, the gap between desktop and high-end server CPUs just gets way too wide in 2028, making Nova Lake 28 P-core silicon a potentially viable option.

AMD is expected to announce its next-gen mobile CPUs at CES 2027, but leaks have already started to pour in, giving us a decent idea of the performance we can expect. Codenamed "Medusa Point," the Red Team's upcoming lineup will likely be based on the Zen 6 microarchitecture, and one of the SKUs has just popped up on Geekbench. It scored much better than the previous leak, beating most of its contemporaries.

The part showed up as "AMD Eng Sample 100-000001713-33_N" and was marked under "AMD Plum-MDS1," which we know is the platform associated with Medusa Point. It's a 10-core (4+6) chip, with 20 threads, clocked at roughly 2.0 GHz, carrying 10MB of L2 cache and 32MB of L3 cache. The L3 cache and clock speeds might be misreported. Currently, AMD only makes two other 10-core mobile parts — Ryzen AI 9 365 and Ryzen AI 9 465, so we're most likely looking at a purported Ryzen AI 9 565 here.

Coming to the scores, the chip netted 3,174 points in the single-core test and 15,092 points in the multi-core test. Both of those numbers are higher than the Strix Point flagship APU, the Ryzen AI 9 HX 370. On average, that SKU sits around 2,600 single-core points, so the Medusa Point score is 22% higher. In multi-core, the AI 9 HX 370 gets 13,400 points, making our main contender 13% faster on average.

It even beats the Strix Halo flagship, the Ryzen AI 9 Max+ 395, by over 400 points in the single-core benchmark, but loses in the multi-core test. Of course, the onboard graphics is no comparison between the two. Compared to a prior leak also showcasing a 10-core Medusa Point APU, this new listing is significantly better. The previous one came in at only 2,300 single-core and 13,002 multi-core points.

It seems like Zen 6 offers a noticeable leap in performance based on architectural improvements, since the core count between the chips we compared is identical. It's too early to judge anything, though, since Medusa Point is months away at this point, and this is just one SKU from the lineup. The top-end parts almost carry a mandate to be faster than their direct predecessor to be even worth releasing; it's the midrange where the real value proposition lies.

Sipearl has been developing a custom CPU, especially designed for high-performance workloads, named 'Rhea', for over five years. In late May, it was finally announced that the company had received the CPU from the fab, initiating the bring-up process, which is a significant milestone. The HPC CPU sports over 80 cores, in addition to an innovative memory subsystem. We spoke directly with Craig Prunty, vice president of marketing and business development of SiPearl, to learn the fine-grained details.

The Rhea CPU is intended to reach markets by late 2026 or early 2027, and won't be the most performant HPC CPU on the market. Regardless, SiPearl told us at Computex that there is interest both towards Rhea and its successors from rather unexpected parties, so the company is in with a chance to become a successful CPU designer over time.

Rhea's long road toward reality

SiPearl's Rhea (or Rhea1, how the company prefers to call the unit these days) sports 80 Arm Neoverse V1 cores with two 256-bit Scalable Vector Extension (SVE) engines for fast vector computations in FP64, FP32, BF16, and INT8 formats; 1 MB of L2 per core; 80 MB system-level cache (SLC), and 104 PCIe 5.0 lanes. The CPU has a unique memory subsystem comprising four HBM2E interfaces for 64 GB of on-package HBM2E stacks for applications that require massive memory bandwidth (think supercomputer applications like fluid dynamics) and four DDR5 interfaces supporting two 256 GB DIMMs per channel, for up to 2 TB of memory per socket. Rhea comprises 61 billion transistors and is fabbed by TSMC using its N6 process technology.

SiPearl received the first samples of its Rhea processor in mid-May, and the CPU is currently in bring-up mode. So far, it looks like the very first silicon works just fine, so the company will not have to respin it, which means SiPearl has a good chance of shipping it to customers in the coming quarters.

"The Rhea1 CPU is in its 12-week bring-up process since May 13, and it works exactly as it was designed to do," said Craig Prunty, vice president of marketing and business development of SiPearl, in an interview with Tom's Hardware Premium. "The test version of Rhea1 will be available for testing by partners and EU collaborative projects at the end of the bring-up process. The general availability of Rhea1 is scheduled for end of 2026."

Getting the very first silicon to work correctly is a stroke of good luck, especially for the very first product from a startup that has never designed a complex CPU before. However, it has taken the company over five years to define and then develop its processor, an unacceptably long cycle. With Rhea, SiPearl not only built its processor, but it actually built the company, Craig Prunty admitted in an interview with Tom's Hardware. The company once tried to work with a contract chip designer, but eventually canceled the deal and formed five in-house development teams in Europe. Since these teams have never worked together before, the processor was delayed a number of times from 2023 to 2026. It, of course, gained eight additional cores in the meantime, but this hardly justifies a three-year delay.

"We have five development teams in Europe: Maisons Laffitte, Massy (both in the Paris region), Grenoble and Sofia Antipolis in France, Barcelona in Spain," Prunty said. "The Bologna team is currently being put together."

One of the reasons why SiPearl has so many locations is that it wants to shrink its development cycle to around 18 months to offer competitive CPUs.

Now, because it is 2026, HBM2E memory is extremely hard to get, which is why Rhea1 will be a limited-run processor only available to select clients and partners. In theory, this is not something that is going to happen to SiPearl's Athena processor for aerospace, defense, and government applications, which is essentially Rhea with 16, 32, 48, 64, or 80 Neoverse V1 cores and without onboard HBM2E, which will be sold based on market demand sometime in 2028. Though, do not expect Athena to have a very long lifespan. SiPearl hopes to tape out its 2^nd Generation Rhea (Rhea2) processor in 2027. That CPU will not have onboard HBM, so its derivatives for aerospace, defense, and government systems will probably follow shortly, making Athena1 obsolete.

Opening unexpected doors

To a large degree, the first-generation Rhea processor is more than just a product for SiPearl, as it is meant to put the company on the map of data center and supercomputer CPUs and proof that a European entity can develop a competitive processor. SiPearl originally intended to address European supercomputers and sovereign AI infrastructure with Rhea1. However, many commercial cloud providers in Europe and the Middle East plan to evaluate the platform and even deploy it (albeit not widely) as they want to ensure they have access to technology in the current geopolitical situation.

“Rhea1 is the unique European server CPU,” Prunty said. “We find that European sovereignty is exportable – customers not only in Europe, but also Middle East and Asia have expressed interest in our processor because it is designed in Europe. They like that we have full control of the source code. They want freedom from embargos, back-doors, kill switches."

While the CPU is the industry's third processor to use a hybrid memory subsystem comprising HBM2E and DDR5 (for which SiPearl deserves accolades), it is very late to market, so while it is natural that various sovereign AI and HPC deployments and Europe-funded supercomputers will deploy it, expecting commercial companies to deploy Neoverse V1-based machines in 2027 is pretty naïve. However, commercial companies will validate and test the platform, possibly do some software porting, and ensure that it works as intended. Some companies might even deploy Rhea1 in their data centers. As Craig Punty puts it, Rhea1 could open rather unexpected doors for SiPearl.

As it turns out, geopolitical tensions and export controls force big players to look for alternatives to American hardware, which is where SiPearl's processors could fit rather well. SiPearl is based in France, it has R&D centers around Europe, it licenses technologies from Arm, and produces its CPUs in Taiwan. The company cannot ship its CPUs to China due to export restrictions, but it can sell them to clients in Europe and the Middle East without restraint, which is its indisputable trump card. Assuming that SiPearl offers competitive performance, its CPUs are almost guaranteed to be adopted by sovereign AI and HPC deployments in Europe, which means guaranteed revenue.

One might argue that since SiPearl uses Arm's cores, it will inevitably compete against Arm's AGI processors eventually. Indeed, it will, once its CPUs address large CSPs. Which is why the company must stay ahead of Arm's own offerings in terms of performance and features, or at least be on par with them.

Seine reference server

For now, SiPearl is bringing up its Rhea1 processor in its labs. The company already has its Seine reference server design that is primarily designed for validation, testing, evaluation, and software porting. For AI and HPC deployments, Seine can be configured for one Rhea CPU and two accelerators; for more traditional supercomputer needs, two Seine motherboards can be installed into one chassis, then interconnected using PCIe cables with the CCIX protocol on top to get a classic 2-way server.

Speaking of the Seine motherboard, it should be noted that since SiPearl uses it for bringing up the CPU, it had to be made perfect so to exclude any possible problems on its side. To that end, it uses costly components and an ultra-expensive 26-layer printed circuit board to ensure signal integrity, reduce crosstalk, provide the best quality power possible, and ensure maximum mechanical stability.

The Seine server reference design will be used by Bull to build servers for the Jupiter supercomputer, according to Prunty. Other server suppliers may follow and adopt the same design to offer their servers based on Rhea1.

"We had also a partnership agreement signed with HPE to work together on European supercomputers tender offers," Prunty said. "Our CPUs will also equip other servers as part of European AI gigafactory project."

Sipearl's Rhea readies up

Developing a supercomputer-grade processor in Europe is already quite an achievement, but developing a CPU that works fine from the first silicon could indeed be considered a breakthrough for a startup. In addition, SiPearl tapes out its Rhea in a good time when potential customers may adopt it despite the fact that Neoverse V1 technology that powers the chip is outdated. As it turns out, export controls made not only sovereign AI and HPC deployments look in SiPearl's direction, but private CSPs in Europe and the Middle East also plan to evaluate its processors.

SiPearl admits that a five-year development cycle is too long for a modern CPU, though it remains to be seen whether it can indeed shrink it to 18 months. The company already has five development sites and is building another one, so it looks like it the company is on the right path. Yet, SiPearl must prove that it can develop Arm-based processors that are competitive against Arm's own AGI as well as other Arm-powered data center CPUs, something that will not be easy to do given the fact that SiPearl is a startup, whereas its potential rivals are billion-dollar companies.

Of course, SiPearl will always have a couple of trumps up its sleeve: the European Processor Initiative (EPI) as well as sovereign AI and HPC deployments that will always prefer locally developed CPUs no matter what. Whether such businesses are enough to build a world-class processor developer is something that remains to be seen, but at the very least, SiPearl will not vanish into oblivion like many other European CPUs makers.

Only a little while back, Phoronix got the chance to test-drive one of Nvidia's upcoming Arm-based Vera CPUs. In certain approved workloads, the chip put up an impressive showing, nipping at the heels of its Xeon and Epyc x86 competitors. In specific single-threaded scenarios, Vera "absolutely dusted the competition" (our words). But AMD had some things to say about the Phoronix test, firing back with its own metrics of a 3.3x performance gain over Vera for the projected output of a 100 kW rack of its hardware.

And Nvidia is already thinking about this future. It revealed that its next-gen Rigel Arm v9.2 CPU core, shipping as part of its Rosa CPU, will deliver even higher per-core performance than Vera's Olympus core within the same silicon footprint via "better instruction delivery," more L2 cache, and better memory handling.

Now, Nvidia is reasserting Vera's advantage for AI work by describing it with a new product category: a "max single-threaded CPU at scale" rather than a parallel-processing beast. Instead of simply maximizing the core count per socket, Nvidia says Vera's monolithic 88-core design is meant to provide strong performance per core under load, enough memory bandwidth per core to keep active cores supplied with data, and predictable latency.

Nvidia describes AI inference workloads as being bound by single-thread speed. For example, a reasoning AI will run the model for one step, and will run the model again as many times as it takes until the answer is generated. Since each step needs the output from the previous one, no amount of parallelism will help — the speed at which one thread can run is most important. The situation is similar in agentic workloads, as agent B can't get its work started without knowing what happened with agent A.

Vera's design, then, appears to be one aimed at both having and eating the proverbial cake: high single-thread speed with a large number of available threads. Vera is an 88-core design with SMT support for 176 total threads. And to supply each of those cores with adequate bandwidth, Nvidia says Vera talks to LPDDR5X RAM at 1.2 TB/s, and that its monolithic compute die keeps cores well fed and avoids bottlenecks thanks to 3.4 TB/s of core-to-core bandwidth. The company says the latter figure is 3x that of "any other data center CPU."

There are many ways to measure inter-core bandwidth, so direct comparisons are tricky at best, but given the bespoke design of Vera for AI inference tasks, the claim is at least plausible.

The company's latest blog post about the new silicon reiterates this point, claiming its new silicon delivers 1.8x higher performance versus its x86 competition in "loaded CPU workloads that represent agentic execution," 1.5x higher perf in coding workflows, and 3x faster work in database analytics.

The numbers Nvidia touts purportedly come from real-world scenarios, starting with those from Perplexity, whose usage of Vera in coding agent work delivered a claimed 1.5x performance increase over x86, and a 1.9x speedup running concurrent sandboxes.

The claimed speed increases are wider still in database workloads, with Starburst (federated database firm) clocking a 3x uplift in large-scale SQL analytics, while Redpanda's real-time analytics saw a claimed 6x latency drop. According to Nvidia, all this purported performance is delivered by Vera's particular architecture, one that aims to deliver maximal single-thread performance with high thread counts.

We should note that vendor-approved benchmarks should always be taken with a bucket of salt, particularly those for hardware in a field that can shuffle trillions of dollars in a single day. The company doesn't say which precise x86 chips it tested Vera against, but it's a fair guess that they're mid- to high-end Intel Xeon and AMD Epyc models.

Nevertheless, in the blog post, Nvidia describes a conundrum that's familiar to most any server administrator: big-iron server chips can pack obscene amounts of cores, making them ideal for processing many tasks at once. However, the more cores you add, the slower they need to be to keep thermal performance and power draw in check. But that scale is an obstacle for tasks that need to be done now, parallelization be darned.

And the architectural decisions involved in using chiplets to scale to high core counts aren't free, either. Nvidia calls this "chiplet tax", and it says that scaling using chiplets creates memory access and performance inconsistencies that Vera's monolithic design is specifically meant to avoid.

We've long emphasized the importance of high single-threaded performance for a fast and responsive experience for client PCs, and it seems like AI agents are going to end up placing similar demands on hardware as they do their thing. If that's how the agentic AI future plays out, Nvidia's particular design optimizations for Vera make greater sense than prioritizing core count above all, as it might be for a general-purpose server chip meant to satisfy different economic and customer demands.

We'll have to see if Intel and AMD respond with "max single-threaded CPUs at scale" of their own.

When Intel switched to a hybrid architecture with its 12th-Gen Alder Lake PUs, it removed AVX-512 support from the lineup entirely because the E-cores didn't support it. Since then, every subsequent generation has shipped without it... until now. Just today, a new Linux patch pushed in the RAID optimized path has revealed that AVX-512 is finally returning to Intel CPUs with Nova Lake, present on both P-cores and E-cores.

Intel has been working toward a unified AVX solution for the past few years, as it was originally a champion of the SMID extensions before running into the hybrid hurdle with Alder Lake. Getting past that hurdle is AVX10, which Intel first detailed a few years back. With AVX10.2, 512-bit instructions will run on the P-cores, while either core type can handle converged 256-bit instructions.

As such, the E-cores would have their processing width capped at 256-bit, while the P-cores would be open to the full 512-bit wide pipelines. Any thread could swiftly move between either core type with AVX10 implemented. Previously, if the scheduler shifted a 512-bit task running on a P-core to an E-core, the application would crash instantly because those E-cores couldn't process the instruction.

However, the new patches suggest that Intel has now mandated native 512-bit execution across both P-cores and E-cores, no longer requiring the latter to step down and process the data a bit slower. This is a major development over the standard we originally expected Intel to adopt; the E-cores are apparently becoming just as performant as the P-cores when it comes to SIMD instructions with Nova Lake and later.

Intel's original announcement showed that it had already uncoupled the software improvements of AVX-512 from the physical width of the register, so the new instruction features could remain present for both 512-bit and 256-bit execution. This includes things like masking, embedded broadcast for rounding math operations, and doubling the number of the registers themselves from 16 to 32.

It remains unclear if we will ever see this version of AVX10 on client CPUs, as it seems Nova Lake is going purely for 512-bit execution across both core types. AMD's current-gen Zen 5 processors also have full 512-bit wide registers, while the previous Zen 4 architecture divided a single 512-bit task across two 256-bit execution units over two clock cycles. This ensured execution remained disruption-free even if it took longer.

The last time we saw native AVX-512 support on an Intel client family was Rocket Lake (11th Gen), right before the hybrid era ushered in by Alder Lake. For modern AI workloads and other compute-heavy tasks such as encoding or simulations, AVX-512 instructions bring a huge performance benefit that's foolish to be left on the table. Keep in mind that this is just a Linux patch at the moment and that Intel hasn't officially announced native AVX-512 support for Nova Lake yet.

Intel's next-gen desktop family, Nova Lake, is expected to take a generational leap in terms of performance, and a big part of that is the rumored introduction of bLLC (Big Last Level Cache). It would be the company's answer to AMD's X3D chips, but implemented even more aggressively, from what we can tell so far. Now, a new leak from Jaykihn says Intel has added two new Core Ultra 5 SKUs with bLLC to the lineup, both featuring 22 cores in total.

The leaker accidentally tweeted out the wrong specs at first, so we've only embedded their correction post. According to the leak, each chip features a 22-core config comprising 6 P-Cores, 12 E-Cores, and 4 LP-E cores on a single tile. That would access to up to 144MB of bLLC. The rumor indicates one unlocked SKU with a 125W TDP and a locked SKU with a 65W TDP; there are seemingly no other differences between the two.

Dual-tile variants of Nova Lake-S could push the cache count up to 288MB, but those will likely be reserved for truly high-end SKUs. The Blue Team still offering a competitive midrange option with a lot of cache to help with gaming performance would be a welcome development. As a reminder, Nova Lake will likely use the Coyote Cove architecture for its P-cores and the Arctic Wolf architecture for its E-cores, according to leaks and rumors.

Previously, we covered how the alleged 42-core Nova Lake-S silicon could actually have 44 cores, combining 2x 8P+12E tiles, perhaps freeing up 6P+12E tiles that could be used for cheaper bLLC-equipped chips. At the time, we predicted that a Core Ultra 7 SKU with 22 cores could be the recipient of this silicon, but now this leak points toward it being used for Core Ultra 5 SKUs instead. Of course, the prospect of bLLC being limited to unlocked K-series models seems to have withered away at this point.

Nova Lake-S is shaping up to be a vast and expansive family for Intel with several SKUs that might not even make it to market by the time it launches as the Core Ultra 400 series next year. It remains to be seen how the company will name all these different models. So far, rumors indicate a CES 2027 announcement is imminent for Intel's next-gen family, but the ongoing component crisis could throw things off.

Intel on Friday confirmed that it had increased prices of some of its consumer and server CPUs, citing market dynamics, rising costs, and soaring demand for these products. While select enthusiast processors increased from $30 to $50, data center-grade products increased by hundreds, if not thousands, of dollars. Intel is among many suppliers that have recently hiked prices of their products, citing increasing costs and demand that exceeds their supply.

"The recent pricing updates reflect current market dynamics, including rising supply chain costs and strong demand for our Intel Core Ultra 200S Plus processors," an Intel spokesperson told Tom's Hardware. "These updates are in line with recent price increases for other Intel product families based on similar factors."

This week it turned out that Intel had quietly increased recommended customer prices (RCPs) of its latest Core Ultra 200-series Plus processors for desktops — the Core Ultra 7 270K Plus and the Core Ultra 7 250K Plus — by $30 - $50, depending on the model. Both processors belong to the Arrow Lake family and, like the rest of them, are produced by TSMC. Yet, Intel's original 'non-Plus' Core Ultra 200-series processors did not increase their MSRP. The flagship Core Ultra 9 285K still carries a $599 RCP, just like it did at its launch in Q2 2024. Something similar applies to the least advanced Arrow Lake processor for desktops — the Core Ultra 5 225 — that has an RCP between $183 and $236, which is a bit lower than its launch RCP of $241.

If Intel did see supply-chain inflation, it would be reasonable to expect the company to adjust prices of the whole family. Instead, the company raised prices only on select products that apparently had become unexpectedly attractive to customers who can afford them and who have probably demonstrated willingness to buy them above recommended prices. This means that we are not dealing with a simple cost pass-through, but rather with a price hike associated with strong demand for specific SKUs.

When it comes to data center-oriented processors, we see rather massive price hikes. While higher-end Xeon 6 'Granite Rapids' CPUs cost less than they used to at launch in 2024, they are noticeably more expensive after Intel slashed their recommended prices in 2025, and they can be twofold higher when compared to retail prices from mid-2025. Perhaps the biggest surprise is that select Xeon 8000-series 'Emerald Rapids' processors now carry higher RCPs than they used to when they were released in late 2023.

All Intel Xeon processors are produced internally (so Intel cannot blame higher costs on TSMC), and while Intel gets raw materials from its partners, it is doubtful that overpriced photoresist can significantly affect RCPs of CPUs that sell for thousands of dollars. Meanwhile, Intel has been saying for several quarters now that demand for its Xeon processors exceeds supply. Therefore, it makes a lot of sense for Intel to finally capitalize on that and increase RCPs of popular models.

There is a caveat, though. Actual prices of data center hardware tend to differ from list prices as they depend on many factors, including volumes and strategic relations between suppliers and consumers. To that end, while it is evident that Intel has increased RCPs of its Xeon CPUs, it remains to be seen how this affects its average selling prices (ASPs) for the ongoing quarter and for the whole year.

Intel released its new Core Ultra 200S Plus desktop processors back in March 2026, and it appears that the company has quietly increased their prices. According to Intel's official product page, the Core Ultra 270K Plus now has a recommended customer price of $339–$349, up from the previous $289–$299. Similarly, the Core Ultra 250K Plus has increased from $189–$199 to $219–$229.

While the increase amounts to an additional $30–$50 (depending on the model), it is worth noting that these higher prices have already been reflected on Amazon. We have reached out to Intel to confirm the pricing changes and will update this article if and when the company responds.

The Core Ultra 270K Plus and the 250K Plus are part of Intel’s Arrow Lake Refresh lineup and were introduced with a relatively low price compared to their non-Plus predecessors. Intel claimed up to 15% improvement in gaming performance at 1080p compared to stock Arrow Lake chips thanks to several key improvements aimed at addressing the architectural bottlenecks of the original Arrow Lake processors.

The Core Ultra 7 270K Plus comes with 24 cores, including 8 P-cores and 16 E-cores, which is similar to the more expensive Core Ultra 9 285K. The Core Ultra 5 250K Plus comes with 18 cores across 6 P-cores and 12 E-cores, which is slightly below the 20 cores available on the Core Ultra 7 265K. Intel also increased the die-to-die interconnect frequency by 900 MHz to reduce latency and improve gaming performance.

The new chips also offer native support for faster DDR5-7200 memory, as opposed to 6400 MT/s on the non-Plus Arrow Lake chips. Then there’s Intel's new Binary Optimization Tool, a free software utility that analyzes executables and automatically recompiles them with CPU-specific optimizations. According to Intel, this alone is responsible for a significant portion of the gaming performance gains.

While the increase in price may hamper the value advantage that these chips initially offered, there is no doubt that they bring meaningful performance to the table. In our testing, we found the Core Ultra 7 270K Plus to deliver excellent productivity performance with a noticeable uplift in gaming over the original Arrow Lake chips. In fact, it is currently the best Intel CPU on the market. As for the Core Ultra 5 250K Plus, it offers one of the strongest value propositions at its price point. It also remains our top pick for the best budget CPU, thanks to its ability to compete with similarly priced gaming processors while excelling in heavily threaded workloads.

AMD has submitted Linux kernel patches including support for its new low-power CPU cores that will likely emerge in its future heterogeneous processors. The new patch clearly distinguishes between high-performance cores, efficiency cores, and low-power cores, so it is safe to say that AMD's upcoming CPU platforms will use three types of cores, with the low-power one serving light workloads, reports Phoronix.

AMD's heterogeneous processors identify CPU types using CPUID Function 0x80000026 (Extended CPU Topology), as EBX bits [31:28] carry the core classification. Up until recently, AMD only classified its cores as Performance and Efficiency, while the latest patch adds Low-Power cores. The patch enables Linux to distinguish between Performance, Efficiency, and Low-Power cores efficiently, and the latter are also correctly supported by AMD's performance management.

According to AMD engineer Vishal Badole, these cores are designed specifically for background and idle tasks where reducing energy consumption is more important than offering high performance.

In recent years, both AMD and Intel released heterogeneous processors featuring both high-performance and energy-efficient types of cores in a bid to wed performance and low power consumption. With its latest CPU platforms, Intel introduced its low-power cores located in the SoC tile to offload light tasks and prolong the battery life of laptops. As it turns out, AMD is going the same route. Although AMD uses two different core types, the underlying architecture is the same. It offers a "dense" core offering that's optimized for space, while Intel uses entirely different microarchitectures.

Beyond the description of the Linux patch, AMD disclosed little about the low-power cores. The company only described them as being optimized for the lowest possible power consumption during background processing and idle operation, but did not reveal how they differ architecturally from today's dense Zen 5c cores. In addition, the kernel patches introduce no new scheduling policies or optimization logic beyond identifying the additional CPU category.

AMD also did not reveal whether these cores are based on the Zen 5, Zen 6, or other future microarchitecture. It should indeed be noted that AMD has traditionally preferred to use the same microarchitecture within one CPU, albeit with different optimizations when it comes to die size (or rather floorplan) and clocks. Such an approach greatly simplifies software development and performance management, but at the cost of higher power consumption compared to what a simplified microarchitecture would have offered.

Loongson Technology has announced the Loongson 3C3000, a new 16-core server processor aimed at low-cost general-purpose server systems. Unveiled on June 26, 2026, via a public corporate release, the chip is based on Loongson’s in-house architecture and is designed for small- and medium-sized business workloads, including file, database, web, and business process servers. Loongson says the 3C3000’s general-purpose computing performance matches that of the company’s earlier 3C5000 server processor.

The Loongson 3C3000 — which is based on a 64-bit architecture and supports the LoongArch instruction set — uses the company’s LA364E processor core design and comes in an FCBGA1371 package measuring 37.5mm by 37.5mm. The chip is pin-compatible with the Loongson 3B6000 processor, which should make it easier for system builders to reuse existing platform designs.

The processor features 16 physical cores and 16 threads, with clock speeds ranging from 1.5 GHz to 1.8 GHz. Each core supports 128-bit vector instructions and three-issue out-of-order execution. Loongson says each core integrates two fixed-point units, one vector unit, and two memory access units.

Cache and memory support are modest by modern server standards, but in line with the chip’s low-cost positioning. Each core includes 64KB of private L1 instruction cache and 64KB of private L1 data cache, while all 16 cores share 16MB of L2 cache. The integrated memory controller supports two 72-bit DDR4-2400 memory channels with ECC support, giving the processor server-class error correction for business and infrastructure workloads.

For expansion, the 3C3000 provides two PCIe x16 interfaces, totaling 32 PCIe lanes. These can be split into up to four x4 or x8 interfaces, depending on platform requirements. The chip also includes another PCIe x16 interface that can be configured as LCL (Loongson Coherent Link) for dual-processor interconnection. Other interfaces include SPI, UART, three I2C interfaces, AVS, and 16 GPIOs.

Loongson lists a typical power consumption of 40W when running at 1.5 GHz. The processor supports dynamic shutdown of the main module clock and dynamic frequency adjustment of the main clock domain, helping reduce power consumption under lighter workloads. It also integrates a Loongson-developed security and trust module that supports Chinese national cryptographic algorithms for encryption and decryption.

Unlike the 36C000, which Loongson says matches Intel’s Xeon, the company is positioning the 3C3000 as a low-cost, high-performance server CPU for customers that need general-purpose compute rather than high-end acceleration or AI performance, slotting it below the higher-core-count 3C6000 server lineup the company launched a year earlier. Support for China’s local software and hardware ecosystem also appears to be a selling point. The company has not publicly disclosed pricing, a common move for server processors, which are often sold through system builders and negotiated enterprise deals rather than as boxed retail chips.

Intel is expected to push the boundaries on power draw with its upcoming Nova Lake series processors, which will rival the best CPUs. According to newly leaked information, the flagship 52-core desktop variant is expected to feature a dual-compute tile architecture with a massive PL2 limit of 474W. The information was shared by LC Tech Leaks and confirmed by Jaykihn, who has a pretty solid track record with Intel hardware.

PL2, or Power Limit 2, represents the maximum power a CPU can draw during short boost periods. That said, a PL2 target of 474W remains quite demanding, although a previous rumor suggests Intel may also have a PL4 emergency power limit over 700W. It is important to note that these power limits may only apply to the top-end models with the dual-tile architecture.

Additionally, the leak also sheds light on the upcoming platform, including the previously rumored LGA1954 socket. We already know that Nova Lake-S will require a new generation of motherboards. Motherboard vendors are expected to classify their boards by sustained PL1 power levels, with configurations for 35W, 65W, 125W, and 175W CPUs. Enthusiast-grade motherboards, likely the Z990 series, are also rumored to feature three EPS 8-pin CPU power connectors instead of the traditional two. While vendors will have the option to include a third connector, its primary purpose would be to support extreme overclocking and would not affect the CPU's rated performance profile.

The upcoming Nova Lake-S lineup is expected to carry the ‘Core Ultra 400S’ moniker and will be Intel's biggest desktop CPU overhaul in years. We’ve previously reported leaked specifications indicating configurations ranging from 6 to 52 cores, with support for DDR5-8000 memory. The flagship 52-core model is expected to feature 16 performance cores, 32 efficiency cores, and a new Big Last Level Cache (bLLC) design to take on AMD's 3D V-Cache gaming dominance. The company is also rumored to introduce integrated Xe3 graphics, Thunderbolt 5, PCIe 5.0 connectivity, and an upgraded NPU for AI workloads.

While these specifications are unconfirmed, it is clear that Intel is targeting substantial gains in gaming, multi-threaded performance, and overall platform capabilities with its next-gen processors.

AMD has brought back its gaming champion from four years ago. The Ryzen 7 5800X3D has been revived in 2026 to breathe new life into the AM4 platform. The Zen 3-based CPU was the best CPU for gaming of its time, thanks to the first-generation 3D V-Cache technology. Since then, however, the competition in our CPU benchmark hierarchy has become more fierce.

Today's competition is Intel’s Core i7-14700K, based on the Raptor Lake Refresh architecture. At the time the Ryzen 7 5800X3D released, Intel’s 12th-gen Alder Lake CPUs were its main competition. Here, we revisit the comparison with Intel’s newer Core i7-14700K, which is available around the same price of $350.

The focus of this faceoff is to determine which CPU is the superior all-around chip. We will put the two CPUs through a series of tests spanning different categories to ultimately determine which CPU you should buy for your system.

This faceoff breaks down how two CPUs compare to each other in a head-to-head battle. If you'd like to read more about either processor, as well as see our full suite of tests, make sure to read our AMD Ryzen 7 5800X3D re-review and Core i7-14700K faceoff.

Features and Specifications: AMD Ryzen 7 5800X3D vs Intel Core i7-14700K

The Ryzen 7 5800X3D was first launched in April 2022 as a part of the Vermeer desktop CPU family. It is based on the Zen 3 architecture and built on TSMC’s 7nm production process. The CPU features 8 cores and 16 threads, with a TDP of 105W and a PPT of 142W. It has a base clock of 3.4 GHz and can boost up to 4.5 GHz.

The 5800X3D only supports DDR4 memory at a rated speed of 3200 MT/s over a dual-channel interface. It is compatible with the AM4 socket, with support for 300-series, 400-series, and 500-series AMD chipsets (though check support with your specific motherboard). It also supports 20 lanes of PCIe Gen 4. However, the 5800X3D does not have integrated graphics.

On a more positive note, the Ryzen 7 5800X3D was the first CPU to employ the new 3D V-Cache technology. As a result of stacking the cache vertically on the die, the 5800X3D has a total L3 cache of 96 MB. Of this pool, 64 MB is part of the 3D V-Cache stack. Core overclocking is disabled on the Ryzen 7 5800X3D due to its 3D V-Cache layout; DRAM overclocking still remains available.

Its competitor, Intel’s Core i7-14700K, uses a vastly different layout. It features the Raptor Lake Refresh architecture, which is a refined version of the 13th-generation Raptor Lake base architecture. The Core i7-14700K was launched in October 2023 and was built on a 10nm production process (Intel 7).

Intel’s 14th-generation CPUs use a hybrid core layout with performance-focused “P-cores” and more efficient “E-cores.” The 14700K also follows this structure, featuring 8 P-cores and 12 E-cores, for a total of 20 cores. In the 14700K, Hyper-Threading is only available on the P-cores, so the CPU has a total of 28 threads. The chip can boost the P-cores up to 5.6 GHz, while the E-core boost clock is 4.3 GHz.

Interestingly, the Core i7-14700K supports both DDR4 and DDR5 memory at 3200 MT/s and 5600 MT/s, respectively. The CPU is compatible with the LGA 1700 socket featured in the 600-series and 700-series Intel motherboards. There is also support for 16 PCIe Gen 5 lanes and 4 PCIe Gen 4 lanes.

The Core i7-14700K has a TDP of 125W, with a higher PL2 limit of 253W. Integrated graphics are also offered in the 14700K in the form of UHD Graphics 770. There is 33MB of shared L3 cache on the chip. Perhaps more importantly, the Core i7-14700K is fully unlocked for overclocking, which is a big advantage over its competitor for today, though that requires a Z-series motherboard.

Zooming out a bit, it is clear that the Core i7-14700K is vastly superior to the Ryzen 7 5800X3D on paper. It is a newer CPU, so it has a better feature set, including PCIe Gen 5 and DDR5 support. It offers more cores, a higher boost clock, integrated graphics, and an unlocked multiplier for overclocking.

⭐Winner: Intel Core i7-14700K

Nothing is decided on paper alone, but the Core i7-14700K offers much better specs, newer features, and even has overclocking support. It takes this round quite easily.

Gaming Benchmarks and Performance: AMD Ryzen 7 5800X3D vs Intel Core i7-14700K

AMD claims to have “re-engineered” the Ryzen 7 5800X3D for its 2026 re-release, so we have tested it again, along with a whole bunch of worthy competitors, including the 14700K. We chose the 1080p resolution for our 16-game test suite in order to maximize the performance differences between the various CPUs. The graphics card used was the GeForce RTX 5090 to keep potential GPU bottlenecks to a minimum. Let’s get into the results.

Starting off with our 16-game FPS geomean at 1080p, the Core i7-14700K dominates the Ryzen 7 5800X3D with an average result of 166.7 FPS across our tested games, compared to the 145.6 FPS result of the Ryzen 7 5800X3D. That is a 14.5% difference in favor of the 14700K in our performance geomean. In 1% lows, the 14700K leads the Ryzen 7 5800X3D by 20% on average, putting out 114 FPS against the Ryzen’s 95.

However, there is another side to this benchmark table. The Core i7-14700K supports both DDR4 and DDR5 memory, so we tested it in both configurations. With DDR4-3200 memory, the 14700K’s advantage vanishes, and instead the Ryzen 7 5800X3D leads by 1.04%, or just 1.5 FPS. The 1% lows are in favor of 14700K by only 3 FPS (3.15%), which is astonishingly close.

When the Intel CPU is paired with DDR5 memory, the Ryzen 7 5800X3D’s cache advantage seems to be struggling against the Core i7-14700K’s raw core count and higher boost clock (along with far faster memory speeds). Looking at individual titles, we see a similar pattern with the Core i7-14700K holding a consistent lead over the 5800X3D.

In 007 First Light, the 14700K paired with DDR5 memory has a 25.7% lead on average over the 5800X3D. That lead shrinks to 21.5% in Crimson Desert, and the difference is 13.7% in favor of the 14700K in Cyberpunk 2077. Interestingly, the Core i7-14700K leads the entire pack in Flight Simulator 24, establishing a 26.6% lead over the 5800X3D in this title. The DDR5-equipped 14700K also leads the 5800X3D in Spider-Man 2, Starfield, The Last of Us Part One, Baldur’s Gate 3, and Counter-Strike 2.

However, when the Core i7-14700K is paired with DDR4-3200 memory, the picture changes completely. The Ryzen 7 5800X3D leads the i7-14700K with DDR4 memory in Baldur’s Gate 3 by 11.7%. In Crimson Desert, the lead is 3.2% for the 5800X3D, and 2.6% in Cyberpunk 2077. The Ryzen 7 5800X3D sits between the DDR5 and DDR4 versions of the 14700K in a few other titles, including Counter-Strike 2 and DOOM: The Dark Ages.

There are also some titles in which the Ryzen 7 5800X3D leads both the DDR4 and DDR5-equipped versions of the Core i7-14700K. In F1 2024, the Ryzen 7 5800X3D leads the DDR5 14700K by 5.6%, and the DDR4 14700K by 13.7% on average. The same pattern can be seen in Final Fantasy XIV, with a 6.6% lead over the 14700K using DDR5 memory, and in Minecraft RT, with a 18.5% lead over the 14700K using DDR4 memory.

It is certainly all over the place when you put both configurations of the 14700K into the mix, but the two behave more like separate CPUs. The long and short of it is that the 14700K with DDR5 memory provides the best gaming performance on average, followed by the Ryzen 7 5800X3D. The DDR4-equipped 14700K is ever-so-slightly slower than the 5800X3D, but it really just depends on the game you’re playing.

During our testing, the Core i7-14700K averaged 4.93 GHz with DDR5 memory and 4.88 GHz with DDR4 memory. The Ryzen 7 5800X3D could only manage 4.34 GHz, but it sipped only 77.5 watts during our gaming tests. The 14700K DDR5 averaged 132.4 watts, while the DDR4-equipped 14700K averaged a whopping 155.1 watts during gaming. This is why the 14700K with DDR4 reached an average temperature of 80 °C, compared to 62 °C for the 14700K with DDR5 and 59 °C for the 5800X3D.

In addition to running the coolest, the Ryzen 7 5800X3D is also the most efficient CPU of the bunch. The 5800X3D had an FPS-per-watt output of 1.88, compared to 1.26 for the Core i7-14700K with DDR5 memory, and just 0.93 for the DDR4 version. It is amazing how much the Core i7-14700K suffers when paired with DDR4 memory.

Lower overall performance also hurts the value proposition of the DDR4-equipped 14700K, as it puts out just 0.39 FPS-per-dollar, compared to the 0.45 of the DDR5-equipped 14700K. Astonishingly, the Ryzen 7 5800X3D falls between the two 14700K versions, delivering 0.42 FPS per dollar. This makes the Core i7-14700K the value king, but only if you pair it with DDR5 memory. I suspect that will be tricky in the current market.

⭐Winner: Tie

While the Ryzen 7 5800X3D does slightly pull away from the DDR4-equipped 14700K, both of these setups get demolished by the 14700K when it is paired with DDR5 memory. We're calling this round a tie considering the massive price disparity between DDR4 and DDR5 memory right now.

Productivity Benchmarks and Performance: AMD Ryzen 7 5800X3D vs Intel Core i7-14700K

Productivity performance spans single-threaded and multi-threaded workloads, so we tested the CPUs across a range of benchmarks covering both categories. Just like in our gaming tests, we tested the 14700K with both DDR5 and DDR4 memory, since it does impact the performance significantly.

Intel’s hybrid architecture has historically been quite strong at multi-threaded workloads due to E-cores, and that pattern appears here too. In our multi-threaded performance ranking geomean, the Core i7-14700K scores 492 points, a massive lead of 116.7% over the Ryzen 7 5800X3D that could only manage 227 points on average. Even when the Core i7-14700K is paired with DDR4 memory, it has a 105% higher average score than the Ryzen 7 5800X3D.

The superior core count of the 14700K is proving to be the difference maker in this category. In the Cinebench 2024 multi-core test, the 14700K with DDR5 memory is a whopping 126.6% faster than the Ryzen 7 5800X3D. Even the DDR4-equipped 14700K secures a 107% lead over the 5800X3D in Cinebench. The lead for the 14700K is about 137% in POV-Ray, and it shrinks to 135% when using DDR4 memory.

Blender tests were also favorable for the 14700K, but we didn’t see a big difference between DDR4 and DDR5 systems in these benchmarks. In Junkshop, the DDR5-equipped 14700K leads the 5800X3D by 116.4%; in Monster, by 116.6%; and in Classroom, by 118.3%. The DDR4 variant follows closely behind, by 1 or 2 percentage points.

The memory generation again comes into play when we look at HandBrake x265 10-bit encoding, with the DDR5-14700K leading the 5800X3D by 90.5%, while the DDR4-14700K manages a 82% lead. The gap is even larger in x264 encoding, with the DDR5 variant gaining a 105% lead over the 5800X3D, while the DDR4 variant can only manage a 63% lead.

That still makes the Core i7-14700K far better than the 5800X3D in productivity workloads, regardless of the memory generation. However, we still have single-threaded results to look at. Our single-threaded performance ranking geomean puts the Core i7-14700K 36.6% faster on average than the Ryzen 7 5800X3D. Interestingly, there is no difference in single-threaded performance between the DDR5 and DDR4 variants of the 14700K.

The same trend is seen in individual benchmarks as well. The 14700K is about 25% faster than the 5800X3D in Lame’s audio encoding test, and the DDR4 variant is in the same ballpark as well. Curiously, the DDR4-equipped 14700K is slightly faster than the DDR5-14700K in Cinebench 2024 and also outperforms the 5800X3D by 36.6%. Safe to say, the RAM difference doesn’t really come into play in these tests.

Overall, though, the winner is quite clear. The Ryzen 7 5800X3D is a gaming-oriented chip with only 8 cores and 16 threads, so it is no match for the 20-core 14700K in productivity workloads. Whether you go for DDR4 or DDR5 is your decision, but the productivity champion of this faceoff is the Core i7-14700K.

⭐Winner: Intel Core i7-14700K

With an average lead of 116% over the Ryzen 7 5800X3D in multi-threaded tasks, the Core i7-14700K sweeps the productivity round quite easily.

Overclocking: AMD Ryzen 7 5800X3D vs Intel Core i7-14700K

Overclocking has never been a strong suit of AMD Ryzen processors; however, the Ryzen 7 5800X3D doesn’t support core overclocking at all. AMD cited the 3D V-Cache technology as the reason the 5800X3D can’t be overclocked, and they were right to assume so.

Due to the vertically-stacked cache, heat transfer from the CPU die to the heatspreader was a real issue. An overclocked 5800X3D would have sipped more power and produced more heat. Therefore, an efficient heat-transfer system was needed but could not be developed in time for the first-generation V-Cache product.

AMD has since reinstated overclocking support for Ryzen 9000 series X3D CPUs by flipping the cache layout, so it no longer hinders heat transfer. However, the Ryzen 7 5800X3D’s core multiplier still remains locked, but you can still tune the DRAM and Infinity Fabric clocks.

The Core i7-14700K, on the other hand, is tailor-made for overclocking. Being a K-series SKU, the 14700K comes with an unlocked multiplier and all the Intel bells and whistles for overclocking. It can reach 6.1 - 6.2 GHz on individual cores with proper cooling, and users can expect a 5.6 - 5.8 GHz all-core overclock on most setups.

Its overclocking toolkit features traditional multiplier adjustments, voltage controls, and established BIOS interfaces that most enthusiasts are already familiar with. The Core i7-14700K also has a significant amount of power headroom to play with, although temperatures become a concern as soon as the power consumption ramps up.

By all overclocking metrics, the Core i7-14700K is the superior CPU for tinkerers. The Ryzen 7 5800X3D can’t be manually tuned, at least not in the traditional sense, and therefore doesn’t really stand a chance in this round.

⭐Winner: Intel Core i7-14700K

The 14700K is the real deal when it comes to overclocking support. The Ryzen 7 5800X3D is locked and therefore can’t be overclocked, so Intel sweeps this round.

Power Consumption and Efficiency: AMD Ryzen 7 5800X3D vs Intel Core i7-14700K

The Ryzen 7 5800X3D has a base TDP of 105W and a PPT of 142W. Intel’s is much higher, with the 14700K clocking in at 125W TDP and a PL2 limit of 253 watts. However, TDP numbers don’t give us a good idea of real-world power consumption, so we ran our own detailed tests.

First, at idle, the 5800X3D consumed only 5 watts, while the 14700K consumed 27 watts. In an active-idle situation, such as YouTube playback, the Core i7-14700K consumed 28 watts with DDR5 memory and a concerning 39 watts with DDR4 memory. The 5800X3D, on the other hand, sipped only 9 watts, making it anywhere from 67% - 76% more efficient than the 14700K.

Moving on to all-core workloads, in our y-cruncher multi-threaded AVX power test, the Ryzen 7 5800X3D consumes 119 watts, while the Core i7-14700K clocks in at a staggering 335 watts, a 181.5% higher figure. Even the Core i7-14700K with DDR4 memory consumed 307 watts, which is still a 158% increase over the 5800X3D’s power consumption.

In Linpack, the Ryzen 7 5800X3D is again more reserved, with the 14700K consuming 168.6% more power than the Ryzen. The DDR4 setup was not much better, with a 137.2% higher power consumption than the 5800X3D in this test. The gap widens even more in Cinebench 2024’s multi-core render and our Blender tests, which show the 14700K consuming anywhere from 250% to 285% higher power than the 5800X3D.

In our encoding tests, the situation remains pretty much the same. In Handbrake x264, the DDR5-14700K consumed 242% more power than the 5800X3D, while the DDR4-14700K consumed nearly 200% more. Similar numbers were seen in Handbrake x265 and SVT_AV1 encoding, with the 5800X3D being the clear winner.

We even looked at single-threaded workloads to determine the power consumption of those tasks. In y-cruncher’s single-threaded AVX power test, we saw the 14700K consume 157% more power when paired with DDR5 memory, and 132% more when using DDR4 memory. Safe to say, the Intel CPU does not fare any better in these workloads either.

To determine the efficiency, we calculated the watts-per-FPS number in Handbrake x265. The 5800X3D was 43.4% more efficient than the 14700K with DDR5 RAM, and about 41% more efficient in this task than the 14700K with DDR4 memory. The pattern can again be seen in Cinebench 2024’s efficiency test, where we look at points-per-watt. The 5700X3D is anywhere from 62% to 68% more efficient than the 14700K in this task.

We can also visualize the efficiency differences using our handy scatter plots. In the Linkpack power efficiency plot, the 5800X3D is towards the bottom left of the chart, while the 14700K is more towards the top. This means that the 14700K uses substantially more energy to deliver marginally higher performance than the 5800X3D. Ideally, you want to be towards the bottom right of this graph.

So, the Ryzen 7 5800X3D consumes much less power in both single-threaded and multi-threaded productivity workloads, and as we saw in our gaming tests, it runs cooler as well. The Core i7-14700K has a distinct performance advantage in all-core workloads, but the power consumption ramps up quickly once it gets going. Still, the Ryzen 7 5800X3D is the clear winner in this round.

⭐Winner: AMD Ryzen 7 5800X3D

The Ryzen 7 5800X3D consumes between 150% and 300% less power than the Core i7-14700K in all-core workloads, making it the definitive winner in this round.

Pricing: AMD Ryzen 7 5800X3D vs Intel Core i7-14700K

The pricing situation is a bit of a wildcard in this comparison, since these are not exactly “new” CPUs. The Ryzen 7 5800X3D was recently re-released at $350, which is $100 lower than its 2022 price tag. The Core i7-14700K is currently priced at $370 at the time of writing, which makes the 5800X3D $20 cheaper in a direct comparison.

However, comparing the two CPUs is more than just comparing their sticker price. We must also look at the platform costs of the two CPUs. The Ryzen 7 5800X3D uses the fan-favorite AM4 socket, which has a whole heap of chipsets in all price brackets. You can pair the Ryzen 7 5800X3D with a mid-range B550 or a high-end X570 motherboard, but older 400-series motherboards are also compatible, depending on the board.

As far as the price goes, B550 motherboards can be purchased for $80 - $180, while higher-end X570 motherboards range from $150 - $300. Some premium models can even go beyond $400, but those are not really needed for our CPU since it doesn’t support overclocking. A nice mid-tier B550 or X570 motherboard will be more than enough for our needs.

Memory is where the price difference really grows. The Ryzen 7 5800X3D only supports DDR4 memory, so it is relatively safe from the ongoing DRAM crisis. A nice 32GB DDR4-3200 kit can run you about $140 - $160, which is definitely higher than DDR4 prices of the past, but nothing compared to current DDR5 rates. The Ryzen 7 5800X3D also needs an aftermarket cooler since it doesn’t come with one, and that can cost you about $100 - $150 too.

For the Core i7-14700K, you have the option of either a DDR4 or a DDR5 motherboard. Even then, you still have to choose between a 600-series or a 700-series chipset. For the sake of this comparison, let’s go with a Z790 motherboard since the 14700K is unlocked and we want those overclocking capabilities. A basic Z790 motherboard can be found around the $150 mark, but we would want to go with something that has decent VRMs. That can cost around $200-$250 at current prices.

Of course, as evidenced in our benchmarks, DDR5 memory is the best way to maximize the 14700K's performance. Due to the RAMpocalypse, DDR5 memory is ridiculously expensive, and a 32GB DDR5-6000 kit can cost between $390 - $550 at the time of writing. Going with DDR4 would require a motherboard swap, but it would save you between $300 and $350 on the system based on these two components alone.

For cooling, the 14700K requires special consideration, as we have the option to overclock. Even a stock 14700K sips more power and produces more heat than a 5800X3D, but if you plan to overclock, the thermals can get out of hand pretty quick. You’ll ideally use a solid 360mm AiO liquid cooler for the 14700K, which can add about $100 - $150 to the cost of your build.

Another factor to consider when determining the value of a CPU is the longevity of its platform. AMD’s AM4 platform has been going strong for a decade, and AMD has continued to support it through updates and releases such as the 5800X3D. However, it would be hard to see AMD releasing more CPUs for the AM4 platform going forward.

On the other hand, Intel’s LGA 1700 socket was already semi-retired, but new reports suggest that Intel will bring this platform back in early 2027. New “Raptor Lake Next” CPUs will reportedly be available on the same socket and the same motherboards, so there is certainly a better upgrade path on Intel’s side.

When we put everything together, the Core i7-14700K is a bit hard to recommend from a value perspective. The motherboards for the 14700K are more expensive on average, and if you want to maximize its performance, you will have to take a massive hit to your wallet with DDR5 memory. Moreover, it is more expensive to cool, too. Its platform looks more future-proof in light of recent rumors, but that can’t guarantee it a win in this round.

⭐Winner: AMD Ryzen 7 5800X3D

The 5800X3D is cheaper to get up and running, since you only need an affordable B550 motherboard and some DDR4 memory to get started. The 14700K can be cheap, but that requires you to leave serious performance on the table and go with a DDR4 setup.

Bottom Line: AMD Ryzen 7 5800X3D vs Intel Core i7-14700K

After a grueling 6-round back-and-forth, we finally have our winner. The Intel Core i7-14700K is the superior CPU of the two. Now, it is not as black-and-white as the 4-3 score might suggest, but the 14700K is still the winner of this faceoff.

The Core i7-14700K delivers better gaming performance on average than the 5800X3D. Sure, there are some titles that favor AMD’s 3D V-Cache, but those wins were not as frequent. However, AMD’s 5800X3D has a better chance if the 14700K is limited by DDR4 memory.

Intel’s 14700K is also vastly superior in productivity and has support for manual overclocking. AMD’s main selling point for the 5800X3D in 2026 is its low price, both upfront and in terms of platform costs. It is also an easier CPU to maintain since it runs cooler and consumes less power.

Interestingly, the choice also depends heavily on your memory generation of choice. It is better to save a few bucks and go with a 5800X3D if you plan to stay on DDR4 for now. However, if you are willing to make the (difficult) jump to DDR5, the 14700K is the clear choice.

Potential buyers who want to stick to gaming should still prioritize a Ryzen 7 5800X3D over a Core i7-14700K with DDR5 memory. On the other hand, if you regularly run any type of productivity workload, the 14700K blows the Ryzen out of the water.

⭐ Winner: Intel Core i7-14700K

Nonetheless, the overall winner of our faceoff is Intel’s Core i7-14700K.

Check Out More CPU Faceoffs

• Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7800X3D
• Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 9700X
• AMD Ryzen 7 9850X3D vs AMD Ryzen 7 9800X3D
• Intel Core i7-14700K vs Intel Core Ultra 7 265K
• AMD Ryzen 9 9950X3D2 vs Ryzen 9 9950X3D

Scalpers are moving in on AMD's Ryzen 7 5800X3D 10th Anniversary Edition. The new CPU, which is identical to the original but "re-engineered" for a new bonding process, is on sale officially today, June 29. We've been keeping track of inventory and haven't seen the chip available for more than a few minutes — and amid the flurry, scalpers are stepping in and asking for $600 or more for the chip on the secondhand market, sometimes doubling the CPU's $350 MSRP.

Over on eBay, most listings for the chip are above $600, with some asking for $750 or more. Worse, they've sold for that price. Looking at sold listings, at least two of the CPUs sold today, one for $540 and another for $585. Most of the listings just use the stock product photo, but some scalpers have an actual CPU in-hand.

Micro Center has the chip available for sale, standalone or as part of a bundle with a motherboard and RAM, and it's exclusively available in Micro Center stores. Presumably some scalpers went to buy a chip at Micro Center this morning to flip it on eBay in the afternoon.

Although it's easy to get caught up in the rush of a new release (or old release, in this case), we don't recommend giving into scalper prices. Rumors have circulated that the 10th Anniversary Edition is a limited-time run, but that's not the case. AMD says it plans on continuing to make the CPU, so we should see inventory stabilize eventually.

We have some placeholder listings available. Again, we've seen inconsistent inventory throughout the morning, so check back at these retailers regularly:

• 5800X3D 10th Anniversary Edition at Newegg
• 5800X3D 10th Anniversary Edition at Best Buy
• 5800X3D 10th Anniversary Edition at B&H

We've been checking retailer listings all morning, and we've only briefly seen the chip in stock. Online, some buyers claim they've secured a CPU through Newegg, but we haven't been able to successfully move through the checkout process yet. We also haven't seen the chip come back in stock on AMD's direct web store yet.

Because of that, it looks like retailers are doing something of a staggered rollout. There's a good chance we'll see more inventory released throughout the day, so make sure to check back at retailers.

We just re-reviewed the Ryzen 7 5800X3D to evaluate how the CPU stacks up to the current market at its new $350 suggested retail price. If you already have an AM4 motherboard and memory to go with the CPU, it's a good choice at $350. Otherwise, there are better options.

AMD's newer Ryzen 7 7600X3D is just as fast in applications, despite sporting fewer cores, and it's around 15% faster in games. Meanwhile, Intel's Core i7-14700K paired with DDR4 matches the 5800X3D in games and offers around twice the multithreaded performance. With DDR5, the Core i7-14700K wins across the board.

AMD answered the demands of gamers and re-released the Ryzen 7 5800X3D, though not without compromise. Although the return of Zen 3 X3D has been a good idea for months, given the limited time we saw those chips on the market, this re-release comes with a surprisingly high price, considering the silicon and how it compares to the best CPUs for gaming.

Price is the biggest issue for the Ryzen 7 5800X3D. AMD shaved $100 off the original MSRP for the 10th Anniversary Edition re-release, but that puts it in very competitive waters, even considering current RAM prices. The CPU is flanked on one side by the Core i7-14700K that also supports DDR4 memory, and on the other by the Ryzen 5 7600X3D, which offers superior gaming performance and a lower price to offset the cost of a DDR5 platform.

The chip mainly appeals to those who already have an AM4 motherboard and memory to go with it, and who were unfortunate enough to miss the small window when you could buy the Ryzen 7 5800X3D a few years ago. In that situation, just about any price is a deal compared to the competition.

Otherwise, the Ryzen 7 5800X3D is about $70 to $100 too expensive, and even that lower price would be questionable if DDR5 prices weren’t out of control. Although the chip has earned its legendary status among gamers, revisiting it in 2026 shows clearly that it maxes out what DDR4 platforms are capable of in games, and it falls far too short of the DDR4 competition in applications.

The island of AM4 users stranded without a clear upgrade path will love the 5800X3D re-release. But the chip is not nearly as impressive as it once was if you have to buy a motherboard and/or RAM alongside your CPU, however.

Some notes on this re-review

We don’t normally re-review products here at Tom’s Hardware, much less update existing reviews outside of some extraordinary circumstance. We will follow up reviews with additional coverage as needed, but our reviews are as much buying advice at the time they’re written as they are historical context years down the road. Reviews exist in the context in which they’re written.

That’s important because, especially with PC hardware, some good products can become worse over time and bad products can become good over time. Even in this past generation, AMD had several stumbles with Zen 5, which it addressed post-launch through a combination of firmware updates and exposing additional settings in the BIOS. Intel had some major regressions in performance with Arrow Lake, which it partially addressed after release with Core 200S Boost.

These products are better now than they were at launch, but it’s still important to know that they had issues at launch. That’s the function of our reviews. They’re a snapshot of how a particular component performs and compares to the rest of the market at a certain point in time. Our list of the best CPUs for gaming and CPU benchmark hierarchy pagesare where you’ll find the consistent updates on which chips are best at any given time.

That preamble is to say that this re-review of the Ryzen 7 5800X3D does not replace our original review, which is why this is a separate piece of content and not merely an update. We’re re-reviewing the chip because AMD is re-releasing it, and we need to compare the chip to the current market it exists in.

That market includes high memory prices, which is a driving force behind the re-release of the Ryzen 7 5800X3D in the first place. We’re paying especially close attention to memory in this review, both in terms of price and performance. However, we’ve also brought some price-competitive DDR5 chips into the mix, including some of AMD’s own CPUs.

Finally, we’re reviewing the original Ryzen 7 5800X3D here. AMD says that the new 10th Anniversary Edition should be identical to the original model, but it’s using a slightly different bonding process, which could have a minor impact on power and thermals, in particular. We’ll be getting a 10th Anniversary Edition into the lab in order to find out, but we don’t expect major performance differences between the original and re-release versions.

AMD Ryzen 7 5800X3D specifications and pricing

It’s difficult to evaluate the specs of the Ryzen 7 5800X3D given the current market, so this is a refresher of what the processor offers and how it compares to some of the current options featured in our test suite. It’s an eight-core / 16-thread chip sporting AMD’s Zen 3 architecture, and it boosts up to 4.5 GHz, with a base clock of 3.4 GHz.

The chip is fabricated on TSMC’s 7nm FinFET process, with GlobalFoundries stepping in to fab the I/O die on its 12nm process. Of course, the main draw of the CPU is the 64MB chunk of SRAM that’s bonded to the compute die, giving the processor access to a total of 96MB of L3 cache.

In recent years, we’ve seen both AMD and Intel increase cache sizes broadly, not just on X3D CPUs. For instance, the Ryzen 7 9700X has the same 32MB of on-board L3 that we can see all the way back to Zen 3, but it has double the L2 cache. Intel has traditionally split L2 and L3 more evenly, and we’ve seen an increase in both with Arrow Lake and Arrow Lake Refresh.

Still, the huge boost in L3 helps a lot here. It comes with some thermal trade-offs, however. Although the Ryzen 7 5800X3D is a very efficient CPU, it also has careful power management. The SRAM sits on top of the compute die, insulating the cores from the IHS. This thermal design means the Ryzen 7 5800X3D has relatively low peak clock speeds out of the box, and it doesn’t officially support AMD’s Precision Boost Overdrive.

AMD has addressed that issue in newer X3D chips, riding the efficiency of Zen 4 with the Ryzen 7 7800X3D and moving to a new bonding process that situates the SRAM below the compute die with the Ryzen 7 9800X3D.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

Although the Ryzen 7 5800X3D was a revelation when it first released, it’s important to remember that it wasn’t leagues faster than Intel’s competing Alder Lake chips, at least not on the level of the 30%+ delta we see today with Arrow Lake and the Ryzen 7 9800X3D. The pedigree that 3D V-Cache has built comes in part from the newer X3D chips, and that’s clear when looking back at the Ryzen 7 5800X3D.

Even more clear is the split between DDR4 and DDR5. Now that we have Raptor Lake (and Refresh) as a comparison point, the Ryzen 7 5800X3D positions itself as the peak of what DDR4 platforms are capable of in games. It’s marginally faster than the Core i7-13700K and Core i7-14700K with DDR4 memory, 17% ahead of the Core i7-12700K with DDR4, and even 4.5% ahead of the Core i7-12700K with DDR5.

This wall that you can see, around 145 fps in our geomean, directly translates into a handful of the games we tested. Especially among the newer titles in our suite, simply moving to DDR5 memory results in more than a 31% increase in performance on the same CPU. That led to a handful of situations where both Raptor Lake CPUs perform worse than the Ryzen 7 5800X3D with DDR4, but offer double-digit improvements with DDR5.

Based on our RAM price tracker, a 32GB kit of DDR5-6000 runs between $400 and $450 currently, while a 32GB kit of DDR4-3200 will run you between $200 and $250. There are plenty of exceptions, for better and worse, but we’re going to call the price difference between DDR4 and DDR5, right now, about $200.

Establishing that number is important because of one CPU: AMD’s own Ryzen 5 7600X3D. It’s on the AM5 platform and requires DDR5, but it’s also $230, $120 less than what AMD is re-releasing the Ryzen 7 5800X3D at. Assuming our lowest RAM prices, that means the Ryzen 5 7600X3D is around 14% more expensive than the Ryzen 7 5800X3D when memory is brought into the price. But the Ryzen 5 7600X3D is also 18% more performant.

That’s the biggest hurdle standing in the way of the Ryzen 7 5800X3D. Even if you already have DDR4 memory, the Ryzen 5 7600X3D and 16GB of DDR5-6000 is only around $80 more expensive, and much more performant. Plus, it gets you on an AM5 platform, setting up cheaper future upgrades (AMD says AM5 will receive support through at least 2029).

In Intel’s camp, the two Raptor Lake chips with DDR4 run up against a similar wall as the Ryzen 7 5800X3D, but offer around a 15% jump with DDR5. The Ryzen 7 5800X3D still makes sense if you already have an AM4 board. However, if you have to buy a motherboard, the Raptor Lake chips offer similar gaming performance with DDR4 and much better application performance, which we’ll get to next.

Elsewhere, there aren’t a lot of surprises. The Ryzen 7 5800X3D is as efficient as ever, drawing just 77.5W on average in our testing. AMD has pushed efficiency even further now, but it’s remarkable to see the Ryzen 7 5800X3D offering similar performance as the Core i7-14700K with DDR4, while consuming half the power.

We also have our value geomean here, which is deceptive. Obviously the price of memory is a huge influence here, not only on total platform cost, but also on performance. The value geomean here just represents a true CPU-to-CPU comparison of value, devoid of RAM context.

007 First Light Benchmarks

The first game in our suite is the newest, which is 007 First Light. The Ryzen 7 5800X3D surprisingly struggles in this title, which is strange given how well IO Interactive’s previous title, Hitman 3, took to 3D V-Cache CPUs. Still, the Raptor Lake chips are ahead here, even with DDR4, and they claim top slots with DDR5.

Baldur’s Gate 3 Benchmarks

Baldur’s Gate 3 favors X3D chips, which is clear based on the fact that the Ryzen 7 5800X3D is just a touch behind the much newer Ryzen 7 9700X. The 5800X3D is also around 11% faster than the Raptor Lake chips with DDR4 memory. However, the Raptor Lake chips with DDR5 are about 20% faster than the Ryzen 7 5800X3D, while the Ryzen 5 7600X3D is around 35% faster.

Crimson Desert Benchmarks

In Crimson Desert, the 5800X3D outpaces the Raptor Lake competition by a few frames, but this game clearly favors faster memory. It also scales oddly well on Raptor Lake chips, as evidenced by the fact that the Core i7-14700K outclasses the Core Ultra 7 270K Plus.

Counter-Strike 2 Benchmarks

Counter-Strike 2 is more competitive, with only the Ryzen 7 7800X3D offering a clear lead above the rest of the pack. The Ryzen 7 5800X3D is marginally ahead of the Raptor Lake chips with DDR4 here, and even moving to DDR5 doesn’t offer a significant improvement.

Cyberpunk 2077 Benchmarks

Cyberpunk 2077 is another clear example of the memory divide that’s growing between DDR4 and DDR5. The Ryzen 7 5800X3D tops the DDR4 rankings by a few frames, but moving to DDR5 on Raptor Lake offers roughly a 15% performance jump.

Doom: The Dark Ages Benchmarks

When we brought up a 31% gap in DDR4 and DDR5 performance earlier, we were referring to Doom: The Dark Ages. This is a fairly recent game that’s clearly designed with DDR5 in mind. The Ryzen 7 5800X3D does surprisingly well, though, offering a 10.8% jump over the Raptor Lake competition with DDR4.

F1 24 Benchmarks

F1 2024 slants heavily toward AMD processors, and it’s a game that scales well with 3D V-Cache. Here, the Ryzen 7 5800X3D is around 5% faster than the Core i7-14700K with DDR5, and 13% faster with DDR4. The Ryzen 5 7600X3D spoils the fun a bit, though, offering a solid 10% jump over the 5800X3D.

Far Cry 6 Benchmarks

Final Fantasy XIV Benchmarks

Final Fantasy XIV scales well with X3D chips, with the three 3D V-Cache CPUs in our test pool topping the rankings. The 5800X3D takes a clear backseat to the DDR5 CPUs, but it still manages a solid 6.6% improvement over the Core i7-14700K with DDR5 and a 16% jump with DDR4.

Flight Simulator 2024 Benchmarks

Flight Simulator 24 leans back toward Intel, with the 5800X3D only managing to outclass the Core i7-12700K with DDR4. In this game, the Ryzen 5 7600X3D is around 18% faster, while the Core i7-14700K is 11% faster, even with DDR4.

Hogwarts Legacy Benchmarks

We can see a similar situation in Hogwarts Legacy, with the Core i7-14700K paired with DDR4 outpacing the Ryzen 7 5800X3D by about 12%. The Ryzen 5 7600X3D, meanwhile, is more than 20% ahead, while the Core i7-14700K with DDR5 marks a 25% lead.

Marvel Rivals Benchmarks

Marvel Rivals is an Unreal Engine 5 game, and it’s mostly bound by the GPU, which explains the stair-step pattern you can see in our data. We can see a clear divide between DDR4 and DDR5 platforms here, with even the Ryzen 5 7600X offering superior performance.

Minecraft RTX Benchmarks

Intel struggles in Minecraft, while X3D chips run away with performance. The Ryzen 7 5800X3D ends up in the middle of the pack, offering a clear buff of around 18% over the Raptor Lake competition but falling short of even AMD’s weaker DDR5 options.

Spider-Man 2 Benchmarks

Starfield Benchmarks

The Last of Us Part One Benchmarks

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

The Ryzen 7 5800X3D was a weak CPU for productivity when it was released, and it hasn’t aged particularly well since then on that front. The clock speed is limited compared to non-X3D Zen 3 chips. Add to that the limitations of DDR4 platforms more broadly in non-gaming applications, and it's clear the 5800X3D was never destined for high marks here.

Even compared to the Ryzen 5 7600X3D, the 5800X3D is only 4.6% faster, which is bad considering the latter CPU has access to 33% more threads. Intel’s hybrid architecture boosts core counts, allowing Team Blue to dominate the top of our multithreaded performance rankings.

The CPU to call out here is the Core i7-14700K, particularly paired with DDR4 memory. It’s more than twice as fast as the 5800X3D in our multithreaded geomean, while offering similar gaming performance. Again, we can see this dilemma for the 5800X3D, where the 14700K is a better all-around CPU with DDR4, while the Ryzen 5 7600X3D is a superior DDR5 CPU for not much more money (even considering the price of DDR5).

Turning to single-threaded performance, the playing field levels out, short of the 5800X3D, which even loses out to the Ryzen 7 5700X due to its limited clock speed (the 5700X boosts 100 MHz higher).

Rendering Benchmarks

Rendering apps factor heavily in our overall multithreaded geomean, so the individual results here largely mirror what you can see in the geomean above. In Cinebench 2024, the Core i7-14700K with DDR4 still offers more than double the performance of the Ryzen 7 5800X3D, and it pushes higher with DDR5. We can see a similar situation in the newer Cinebench 2026.

Single-core results are especially weak considering the Ryzen 7 5800X3D’s limited boost potential. DDR5 isn’t much of a factor here. Although we can see some scaling on the Raptor Lake platforms when moving to DDR5, the Ryzen 7 5800X3D is limited in rendering apps, even by DDR4 standards.

Blender shows similar disparities, with the six-core Ryzen 5 7600X3D largely matching the Ryzen 7 5800X3D, despite sporting a 33% reduction in thread count.

The Ryzen 7 5800X3D is a gaming CPU, pure and simple, so we never expected any miracles here. However, the Ryzen 5 7600X3D is offering comparable performance on a DDR5 platform for less money, while shooting ahead in games. And the Core i7-14700K offers much better productivity performance and comparable gaming performance, even when paired with DDR4.

Encoding Benchmarks

Encoding is similarly a major factor in our geomean, though with a more even split between lightly- and heavily-threaded workloads. Starting with the latter, Handbrake is an all-out workload for any consumer CPU that leans on high core counts, power limits, and clock speeds. With an x265 10-bit encode, even the Core i7-12700K with DDR4 is 39% ahead of the 5800X3D, while the Core i7-14700K with DDR4 shoots ahead, scoring 82% lead. AV1 and X264 show similar gaps.

The LAME audio encoder provides a closer look at single-threaded encoding workloads, and once again, the Ryzen 7 5800X3D struggles. It was the worst performer in the test pool running a standard LAME encode, and the gap between it and the more performant options in our test pool only grows in the extended run.

Creator App Benchmarks

Perhaps the biggest area of crosstalk with gaming is creator applications. The Ryzen 7 5800X3D is a bit more impressive here compared to our rendering and encoding workloads, often matching the Core i7-12700K with DDR4 memory. Still, it ends up at the bottom of our rankings, with newer Raptor Lake and DDR5 CPUs offering better performance.

Photoshop is an app that leans toward AMD chips, though that’s mainly with newer Zen 5 chips. For the 5800X3D, the Core i7-14700K offers a 10% performance jump with DDR4. Things are a bit more competitive in Premiere Pro, with the limitation of DDR4 platforms becoming abundantly clear in our data.

Outside of the Adobe suite, we have DaVinci Resolve, which isn’t as lopsided as Photoshop with recent AMD CPUs. Here, even DDR4 Raptor Lake options are near the top of the chart. The 5800X3D secures a minor win over the 7600X3D and 5700X, but not by a meaningful margin.

Finally, After Effects offers a look at VFX performance. AMD’s 5800X3D especially struggles here, with the Core i7-12700K with DDR4 offering a 13% boost in overall performance. With DDR5, that lead jumps to 22.9%.

Web and Office Benchmarks

In lighter web and office workloads, the Ryzen 7 5800X3D still struggles. Most basic productivity and web applications are lightly-threaded, so it makes sense to see the 5800X3D would fall below the marks we can see in creator apps. Across all of the workloads here, the 5800X3D is either at the bottom of the pile or just narrowly outclasses the 5700X.

That doesn’t mean you’ll run into major performance issues. The workloads here are relatively light, and they’re suitable for far weaker CPUs than the Ryzen 7 5800X3D. We’re mainly looking at general application performance for browsers and lightly-threaded apps here, which is an area where the 5800X3D struggles.

Chess Engines, Compilation, Compression, AVX, and Other Benchmarks

Finally, we have a range of benchmarks examining code compilation, chess engines, database workloads, and far more. The 5800X3D is a gaming CPU, and these workloads are highly targeted at specific applications. There are some workloads where the 5800X3D shines, such as the ebizzy web server test, SQLite database test, and Linpack. However, it mostly lands at the bottom of the pile when looking at workloads more broadly.

We’ve certainly seen a benefit from a larger L3 cache for specialized workloads (see our Ryzen 9 9950X3D2 review for more), but that’s not the draw here. The Ryzen 7 5800X3D is weak in most productivity apps, so unless you plan on running a server and somehow gaming on the same machine, I wouldn’t weigh these results too heavily.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

AMD’s X3D CPUs are known for their efficiency, and although those efficiency gains mainly show up with the Ryzen 7 7800X3D, the 5800X3D is no slouch. Power is carefully limited on the chip, with it peaking at just 119W in our y-cruncher test. In the same test, the Core i7-14700K drew 335W.

Y-cruncher leverages AVX instructions, pushing the Ryzen 7 5800X3D to consume more power compared to the 7800X3D. In an all-out, non-AVX workload like Cinebench 2024, the two CPUs are in lockstep. Again, we can see the Raptor Lake competition dancing with 300W across configurations.

Blender and Handbrake tell a similar story, but efficiency is what’s important here, given how broad the performance window is for the chips in our test pool (particularly in applications). The 5800X3D is much more efficient than the 13700K and 14700K. The margins are also narrower than raw power consumption.

The efficiency picture is clearer with a scatterplot, where we can see a clear separation between the AMD and Intel chips in our test pool. Short of the Core Ultra 270K Plus, there’s a compromise. AMD’s options are much less performant but consume far less power, while Intel's options are generally the opposite.

One upside of the 5800X3D, in particular, is its low idle power consumption. With Zen 4 and Zen 5, we saw a massive increase in power consumption in both idle and active idle (YouTube playback) scenarios. With both the Zen 3 chips in our test pool, we can see a return to single-digit power draw under idle circumstances.

Test Setup

We try to minimize the differences between our test platforms to keep our results consistent, which is especially important here given that we used so many different platforms. Even with big differences, we used the same OS image that’s specifically tailored for testing and frozen to avoid updates skewing our results as we test.

In addition to standardising our OS, we standardise our BIOS settings. We test with XMP/EXPO enabled on memory kits that we’ve validated for stability on the platforms we use. We also disable VBS in the BIOS and turn on ReBAR.

Modern AMD and Intel CPUs come with sophisticated boosting algorithms, but they aren’t always covered by warranty. AMD doesn’t cover PBO, for instance, and Intel doesn’t warranty the “Extreme” power profile that’s common on motherboards. Because of the lack of warranty coverage, we test with PBO disabled and Intel’s power profile set at its default settings.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

At $350, the Ryzen 7 5800X3D really only makes sense if you already have an AM4 motherboard and DDR4 memory to go along with it. That’s not a small audience, especially considering the relatively short time the original Ryzen 7 5800X3D was available on the market. But if you need to buy a motherboard and/or memory for your upgrade, there are better options at this price.

The 5800X3D re-release is a victim of poor pricing. The chip taps out DDR4 platforms in gaming performance, but we’re closing in on half a decade with mainstream DDR5 platforms, and we’ve seen much more powerful CPUs in that time. And yet, the Ryzen 7 5800X3D’s price hasn’t moved significantly away from its original $450 MSRP.

If RAM prices weren’t out of control, a $350 price tag on the Ryzen 7 5800X3D would look insane given the current options around that price. As I’m writing this, the Ryzen 7 7800X3D is on sale for less than $350. And even considering current RAM prices, the Core i7-14700K and Ryzen 5 7600X3D are compelling alternatives.

On the productivity front, the Core i7-14700K is a clear winner, even with DDR4 memory. There are some titles where the Ryzen 7 5800X3D is faster than the Core i7-14700K is when paired with DDR4. But overall, they offer similar gaming performance. Especially now that AMD has turned its attention to AM5, the Core i7-14700K is a better all-around option if you don’t already have an AM4 motherboard.

Although AMD almost always comes out on top with platform longevity, Intel is surprisingly ahead in that regard in the Ryzen 7 5800X3D versus Core i7-14700K matchup. Intel is reportedly set to launch Raptor Lake Next on the LGA 1700 socket next year, though I suspect any performance benefits that come along with it will require an upgrade to DDR5.

On the gaming front, the Ryzen 5 7600X3D does a lot of work. It requires DDR5, but that’s partially offset by the fact that it’s $120 cheaper than the 5800X3D. Factoring in RAM or not, spending a bit more on the Ryzen 5 7600X3D not only nets you much higher performance in games, it also gives you an AM5 platform that you can easily upgrade in the future.

That doesn’t discredit the island that AM4 users are currently on. The Ryzen 7 5800X3D left the market fairly quickly, and given the popularity of AM4, it stands to reason that there’s a large group of people for whom the 5800X3D is a significant upgrade. For that group, it’s the CPU of legend you’ve heard so much about. But if you’re planning on buying a motherboard and/or RAM with your new CPU, shop around a bit more.

• MORE: Best CPU for gaming
• MORE: CPU Benchmark Hierarchy
• MORE: Intel vs AMD
• MORE: How to Overclock a CPU

Asus has started rolling out beta BIOS updates that restore Transparent Secure Memory Encryption (TSME) support to several AM5 motherboards, making it one of the first board vendors to implement AMD’s promised fix after the company was criticized for quietly removing the feature from non-Pro Ryzen CPUs.

According to VideoCardz, the beta BIOS files — which cover several ROG Crosshair, ROG Strix, TUF Gaming, and ProArt boards based on AMD’s X870, B850, and X670 chipsets — were reportedly shared through the ASUS ROG forum by overclocker SAFEDISK and include support for “GNR Transparent Secure Memory Encryption,” with GNR referring to Granite Ridge, AMD’s Ryzen 9000 desktop CPU family.

The BIOS updates are based on AGESA ComboAM5 PI 1.3.0.1b Patch A and appear to restore TSME support for non-Pro Ryzen 9000 processors earlier than AMD’s previously stated July timeline. X870 boards mostly move to BIOS 2401; B850 boards move to BIOS 1686; and X670 boards move to BIOS 3901 or 3886, depending on the model.

AMD officially confirmed to Tom's Hardware last week that it will reinstate memory encryption on Ryzen 9000 CPUs via a BIOS update, following “valuable community feedback.” AMD users had strongly expressed disapproval after the company silently removed TSME support from Non-Pro CPUs. TSME is a security feature that protects CPUs against physical exploits by encrypting the data stored in memory, making it unusable to physical attackers.

A user discovered that the feature was no longer available on his Ryzen 7 9700X system, even though it was enabled in the BIOS. Further testing involving MSI showed that consumer Ryzen chips could report TSME support under older firmware, but not after a newer AGESA update, while Ryzen Pro processors continued to support it. After countless reactions, AMD moved to fix the issue, setting July as the timeline for reinstating the feature via a BIOS update.

The Asus update now suggests the fix is beginning to arrive earlier than AMD’s July timeline, positioning the company as one of the first board makers to package the reinstatement into actual motherboard firmware. However, this is not yet the broad, stable rollout most users will be waiting for. The files are beta BIOS releases shared through the ASUS ROG forum, so users who specifically need TSME may want to track them closely, while anyone running a production or stability-critical system should probably wait for final BIOS builds.

Unless you have several extra hundred dollars to burn, it’s virtually impossible to build a PC right now, even as Prime Day deals start rolling out. You’ll spend more than twice as much (sometimes three times as much) on a kit of DDR5 memory as you would just a year ago, and SSD prices have shot up so much that a decent NVMe drive can rival the price of a graphics card. The key to skirting the worst part of these price hikes is to shop for bundles.

If you regularly browse Tom’s Hardware, you’ve probably seen us cover bundle deals before. We’ve seen a massive increase in their availability and discount rate over the past few months, specifically when a CPU and motherboard are bundled with RAM and/or an SSD. That isn’t an accident. It’s a concentrated effort by retailers, motherboard vendors, and CPU brands to move product in a time when RAM prices have hamstrung the consumer market.

Those bundles are starting to proliferate online. Originally starting at Newegg online and Micro Center in-store, we’ve seen an uptick in the number of bundles available just about everywhere. Some of these bundles will unlock a reasonable price for building a rig. Others offer little more than some pocket change for the trouble of shelling out money at inflated prices. We’re going to help you separate the wheat from the chaff.

We’re going to run down a few bundle deals that are live for Prime Day to kickstart your next build, as well as provide a bit of context on why we’re seeing so many bundles and why the discounts on them are so high. Our focus here is DDR5, as that’s the biggest hurdle you need to overcome when building a new PC right now. DDR4 bundles are available, as well, though much of the focus is on newer platforms.

Bundle deals typically sell out quickly. We will try to keep this list updated with new bundles as we go throughout Prime Day, so check back throughout the week for fresh deals.

A midrange DDR5 bundle with the Ryzen 5 9600X and 1TB of storage for $635

If you want to get started on a DDR5 build, this bundle from Newegg has everything you need. It includes the Ryzen 5 9600X at the heart of the machine with six Zen 5 cores and 12 threads, alongside a 16GB kit of Team Group memory clocked at DDR5-6000 and a 1TB Patriot P410 SSD. The motherboard is a MSI Pro B850-S, which is decent considering the entry-level design.

The motherboard is a full ATX design, so it should fit in most PC cases outside of small form factor options. I/O is fairly limited, but you still get an internal USB-C header, along with two USB 3.2 Gen 1 ports, a USB-C 3.2 Gen 1 port, four USB 3.0 ports, and 2.5Gb Ethernet. The board also has two M.2 slots attached, along with a block of four SATA 6 Gb connectors for additional storage.

As for the Ryzen 5 9600X and Teamgroup memory, it’s all you need to get started with a modern gaming rig. 32GB of memory is ideal, but the 16GB kit included in this bundle only occupies two of the four DIMM slots on our motherboard, so sizing up down the line is always an option. The Patriot P410 is a PCIe 4.0 SSD and it comes with 1TB of storage, giving you plenty of room to install your OS and apps.

For a kicker, Newegg includes a free $80 Cooler Master Elite Liquid 240 all-in-one liquid cooler with the bundle. Although the discount may not seem like much, this bundle essentially cuts the DDR5 price down to normal levels while maintaining the list price of the Ryzen 5 9600X and Gigabyte motherboard.

A high-end Intel build with the 270K Plus for under $800

For most builds, 32GB of DDR5 memory is the sweet spot, so it’s no surprise that we see bigger discounts on bundles that include 32GB. Such is the case with this Core Ultra 7 270K Plus bundle, which includes the chip, a midrange Z890 motherboard from ASRock, and 32GB of G.Skill Trident Z5 memory at DDR5-6000. The bundle is 22% off, but that really sells short how good of a deal it is; the memory alone normally costs $510. That extra $260 buys you a Z890 motherboard and Intel’s most impressive CPU to date.

Although Arrow Lake has earned rightful criticism, the Core Ultra 7 270K Plus is among the best CPUs for gaming. It’s one of Intel’s two ‘Arrow Lake Refresh’ CPUs that come with massive performance improvements over the base range. Even at its affordable price, the 270K Plus is Intel’s fastest productivity CPU around, and it’s just a hair short of outclassing the Core i9-14900K in games.

The motherboard is ASRock’s Z890 Pro RS, which critically includes a Z-series chipset, allowing you to overclock the 270K Plus. It comes with dual Thunderbolt 4 ports, along with two USB 3.2 Gen 1 ports and four USB 2.0 ports. There are four M.2 slots on board, one of which supports PCIe 5.0 x4. Two others support PCIe 4.0, while the last one can operate in either PCIe 4.0 x4 or SATA3 modes. This particular board lacks Wi-Fi (and by extension Bluetooth), so you’ll need to purchase a separate add-in card.

Finally, the Tridzen Z5 RGB memory. This is the memory we use on our own CPU test beds here at Tom’s Hardware, and it’s on the Qualified Vendor List (QVL) of most major motherboards (including the board included here).

Top-shelf gaming performance with the Ryzen 7 7800X3D for $900

The gold standard for a gaming PC these days is one of AMD’s 3D V-Cache CPUs, and in particular, the Ryzen 7 7800X3D or 9800X3D, AMD’s most recent eight-core X3D parts. This bundle includes the last-gen 7800X3D, which, despite being slower than the 9800X3D, still beats everything else in our CPU benchmark hierarchy. Alongside the chip is a 342GB of Corsair Vengeance RGB DDR5-6000 memory and an MSI MAG X870 Tomahawk Wi-Fi motherboard.

Despite coming from the Tomahawk range, this MSI motherboard is solid. It includes Wi-Fi 7 and 5Gb LAN for networking, along with a PCIE 5.0 x16 slot for your GPU and two Gen5 M.2 slots (alongside two Gen4 slots). Around the back, you get two USB 3.2 Gen 2 ports, three USB 3.2 Gen 1 ports, and four USB 2.0 ports, alongside dual 40Gb USB-C connections.

For the memory, we’ve found that 6000 MT/s is the sweet spot for Ryzen CPUs, and even then, X3D chips don’t need super-fast memory. The kit here is a 2 x 16GB kit, giving you two free DIMM slots to expand in the future.

Why bundle deals are so popular, and why we’re seeing more of them

Most bundle deals aren’t anything to get excited about. How they typically work is that a retailer wants to move more product, and in order to do so, they offer a small discount on multiple products in a bundle. This move is a major factor in why physical Micro Center locations are still so popular; bundle a CPU and motherboard together, and get them for $20 or $30 off. What we’re seeing today is different.

Outside of retailer bundles, we occasionally see bundles that are subsidized elsewhere. Maybe Intel or AMD, or a motherboard vendor, is running a promotion, and so they offer a rebate on components bundled on sale to a retailer. With current RAM and SSD prices, we’re seeing everyone pitch in.

Vendors tell Tom’s Hardware that motherboard sales have been hit hard by RAM prices, and industry reports suggest sales have dropped as much as 37% at some retailers. Motherboard vendors are in a unique position. Unlike the GPU shortages of years past, motherboards are downstream from RAM. If you can’t afford RAM, you’re probably not going to buy a new CPU or motherboard, even if the prices on those components are good (and they are right now).

So, AMD and Intel, alongside motherboard vendors and retailers, are working together to subsidize these kits. Module houses (the companies that package memory) might be kicking in, as well, but it’s hard to say. Most companies I’ve asked tell me that the specific details of each deal are different, so it’s some combination of all these companies working together on bundle deals.

Regardless, it’s in the interest of AMD, Intel, motherboard vendors, retailers, and, to a lesser degree, even module houses, to get prices down. As long as they’re inflated, the downstream sales impact hits all of them, and bundles allow all of these companies to split the burden of subsidizing high RAM prices.

If you're a gamer, the best CPU you can buy is going to come from AMD right now, and the AMD Ryzen 7 9800X3D is going to be on your shortlist. While it isn't technically the most powerful chip on the market anymore, the 9800X3D is still an absolute powerhouse performer and our top pick for best gaming CPU in 2026, and it's just hit its lowest ever price in the UK on Amazon, costing just £339.99.

● Check out this deal on Amazon

The data from Camelcamelcamel confirms what we know about Amazon's price, and an early check across the other big tech retailers in the UK points to a minimum £30 price drop compared to the cost elsewhere. There are few CPUs that offer the reputation that the AMD Ryzen 7 9800X3D. Until very recently, it was the fastest CPU for gaming in the world, and it's still our best pick overall. This 8-core, 16-thread processor features AMD's newest Zen 5 architecture and comes equipped with a boost clock speed of up to 5.2 GHz.

The performance upgrade, however, is really noticeable in one area: the cache. These 3D V-cache chips feature a boosted 96MB L3 cache capacity, stacked beneath the CPU cores, closer to the integrated heat spreader on a reconfigured die. For gaming, that means that it doesn't need to drop down and rely on the slower system RAM, reducing latency and ensuring much higher frame rates in-game.

The 9800X3D comes fully unlocked, making overclocking a possibility, something which wasn't available on earlier 3D V-cache chips. That said, base performance is pretty stellar on its own, and there really isn't a more efficient gaming CPU on the market at this level.

You aren't having to juice for power here. It offers the best performance per watt of any gaming CPU on sale, besting newer upgrades like the AMD Ryzen 7 9850X3D. Our CPU benchmark data confirms that, showing that the FPS difference between the earlier AMD Ryzen 7 9800X3D and newer Ryzen 7 9850X3D sits at only 3% at 1080p. You're probably running your games at 1440p or 4K with a chip like this one, too, which will close the gap even further.

Pair the 9800X3D with a modern GPU, and you won't find too many bottlenecks from the CPU end. If you combine it with a top-class option like the RTX 5080 or RTX 5090, however, and it'll smash through anything. High frame rates, high resolutions, even in CPU-heavy games like Cyberpunk 2077. The 96MB of L3 cache will give you the performance upgrade you need to perform at any level.

Another bonus for the 9800X3D is its lifespan. This is a long-term CPU, with performance that sits near the top, and which is very much expected to continue to dominate gaming for years into the future. This is an AMD AM5 processor, too, and that's a platform that AMD has publicly committed to support for several years to come.

The £339.99 sale price for the AMD Ryzen 7 9800X3D makes it a must-buy if you're due for a CPU upgrade or you're planning a new build. While RAM and SSD prices have shot through the roof in recent months, this 9800X3D deal is an unmissable discount ahead of the Prime Day sales event, and one you'll want to grab quickly before the stock or discount runs dry.

If you're looking for more savings, check out our Best PC Hardware deals for a range of products, or dive deeper into our specialized SSD and Storage Deals, Hard Drive Deals, Gaming Monitor Deals, Graphics Card Deals, Gaming Chair, Best Wi-Fi Routers, Best Motherboard, or CPU Deals pages.

Most all you hear about "running an AI model" involves a GPU of some sort, but not every AI task is suited to that hardware. Smaller models or single-user latency-sensitive operations can benefit from running on the CPU instead, as it avoids the overhead of shuffling data to and from the GPU. There are also many situations where there is no GPU available to begin with, or it's a meek integrated affair with limited capabilities. Intel and AMD have recently released the full specification for the ACE CPU extensions that make it easier and more power-efficient to run the aforementioned AI tasks on x86 processors.

ACE comes in by offering a technical standard that leverages the existing AVX10 registers but adds silicon dedicated to matrix multiplication. This brings multiple benefits, but the key advantages are better power efficiency, easier development and optimization, and leveraging AVX's 512-bit inputs. The latter makes for easy integration with existing designs by eschewing the need for ACE-specific inputs.

Matrix multiplication is the cornerstone of AI workloads: take a table of numbers, and run a multiplication-addition loop over the whole thing. This has always been possible with most any CPU, though at limited speed. Even today, running these loops uses a lot of power, even when leveraging x86's AVX10 multiply-accumulate instructions — something that's technically a hack, as AVX wasn't designed with 2D matrix operations multiplication in mind.

For the same number of input vectors, ACE can perform 16x as many operations, compared to AVX10. Note this doesn't necessarily mean a 16x speedup, as that will depend on each individual implementation, but it's reasonable to expect that Intel and AMD will dedicate more silicon to this task in future designs to improve performance. Plus, as each ACE instruction performs more work than its equivalent AVX10 loop, there's less CPU instruction overhead and potentially better RAM bandwidth usage right off the bat.

The benefits go far beyond just using fewer instructions for the same thing. ACE is intended to be implementation-agnostic, meaning that ML frameworks and their underlying libraries (PyTorch, TensorFlow) can just write one code path instead of having multiple variations depending on the underlying hardware and its degree of AVX support.

ACE native supports most every data type used in ML operations (including but not limited to INT8, INT32, FP8, FP16, FP32, BF16), but it also can use Open Compute Project's MX block-scaled formats natively, something that AVX10 does not provide. Developers will also be able to move some NPU-specific workloads back to CPU when they need something done now and fast. In those situations, not having to deal with the fact that each NPU is different is a huge boon, too, as ACE offers a consistent target across x86 hardware.

AMD has told Tom's Hardware that it will reinstate Transparent Secure Memory Encryption (TSME) on desktop Ryzen 9000 processors in July (we have the full statement further below). The feature is branded as Memory Guard for AMD's Ryzen PRO lineup, but it's available on non-PRO CPUs, as well. Earlier this year, AMD quietly removed the feature with AGESA 1.2.7.0, which Ars Technica reported on earlier this week. AMD tells Tom's Hardware that it's bringing TSME back to non-PRO Ryzen 9000 chips "based on valuable community feedback."

TSME is a firmware-level encryption feature for memory. It allows the processor to generate a key in order to encrypt data stored in RAM, serving as a layer of protection against cold boot attacks, where a sudden shutdown can allow a physical attacker to extract sensitive data stored in memory.

According to the Ars Technica report, AMD confirmed TSME support on consumer CPUs as far back as 2020 with the Ryzen 7 3700X. The author of the story, Ben Kilpatrick, discovered TSME's removal after running a security audit on a new machine with the Ryzen 7 9700X. After discovering that TSME was no longer supported, Kilpatrick worked with MSI (his motherboard vendor) to confirm that TSME had previously been supported but was disabled in AGESA 1.2.7.0.

Following the discovery, Kilpatrick raised a bug report on AMD's GitHub repository, where Mario Limonciello, a senior principal software engineer at AMD, eventually responded: “My apologies, but I don’t have any more information to share on this topic."

Without any comment from AMD, it appeared as though the company disabled TSME through firmware on its consumer parts in order to differentiate its PRO lineup. TSME isn't a critical security feature for most consumer desktops, as it protects against attacks where the attacker needs physical access to the device. Still, if it was previously a capability, there's no reason TSME should be disabled through firmware.

Now, AMD has responded to Tom's Hardware with the following statement:

"We take the security of our customers’ data very seriously.

AMD Memory Guard (Transparent Secure Memory Encryption, or TSME) is a hardware-based memory encryption technology available on our Ryzen PRO desktop and mobile processors where supported in silicon. It is a foundational security feature, and we have no plans to remove support from our Ryzen PRO lineup. This commitment holds now and in the future.

Regarding certain non-PRO Ryzen 9000-series desktop processors, a BIOS option to enable Memory Guard was previously available but was removed in a recent update. Based on valuable community feedback, we will reinstate this option in an upcoming BIOS release in July."

We've already heard quite a lot about AMD's next-generation EPYC processors, code-named Venice, and we've heard a fair bit about the consumer Zen 6 parts, too, particularly in their Olympic Ridge and Medusa Point guises. Now, thanks to Budapest-based low-level programming enthusiast InstLatX64 (@InstLatX64 on X), we now have the first confirmed details about AMD's Zen 6-based Threadripper CPUs, code-named Mustang Peak. There's no shockers here, and there's not a ton of information to work from, but we can still do some interesting speculation.

The key details that we actually have confirmed are this: Zen 6-based Threadripper parts are internally codenamed "Mustang Peak," and they're going to use Core Complex Dies (CCDs) built on a 2nm-class process at TSMC. We can infer that it's almost assuredly using the same CCDs as Olympic Ridge and Venice, but more on that in a bit. We know that it's going to be based on DDR5 memory and PCI Express 6.0, and likely because of the latter, it's going to require a new platform, known as "TR6."

InstLatX64 found this information in AMD's technical documentation portal, which we unfortunately don't have access to, though you can clearly see the front page of the document above. That's most of the story right there, but there are, nevertheless, some pretty interesting details we can deduce from this information.

For starters, the core counts. AMD's Zen 6 processors are expected to use "Powderhorn" CCDs that raise the core count per chiplet from 8 cores to twelve. This raises the max number of CPU cores on a desktop Ryzen processor from 16 to 24, and it increases the potential max number of cores on a Threadripper Pro CPU from 96 to a whopping 144 cores. Furthermore, we can speculate that clock rates will likely increase considerably, as Zen 6 is said to be a design that aims for clock rates "significantly above" 6 GHz.

A chip with 288 hot-clocked full-power CPU threads is likely to draw massive power, but it's absolutely going to want monstrous memory bandwidth. How do we feed the beast? Not with DDR6, as that standard isn't even finished yet, despite the fact that LPDDR6 is a go. It's possible AMD may be increasing the channel count from 8, as it's known that EPYC Venice is increasing memory channels from 12 on Turin to fully sixteen 64-bit channels (a 1024-bit memory bus). In combination with second-generation MRDIMMs, that's going to give Venice some 1.6 TB/second memory bandwidth.

Could AMD add MRDIMM support to Threadripper? It's certainly possible. Threadripper has always been derived from EPYC technology, and it doesn't seem a stretch to think that the rank-interleaving RAM could solve the memory bandwidth struggle that high-end Mustang Peak chips are going to slam into in massively multi-core workloads. After all, you're not going to reach 12.8 GT/s (the transfer rate of second-generation MRDIMMs) through regular old EXPO overclocking.

Besides that, PCI Express 6.0 will deliver up to 256 GB/second bidirectionally (128GB/second unidirectionally) on a by-sixteen link, which should probably suffice for nearly any sort of device you care to plug into one of these machines. I almost wrote "PCs," but with 144 CPU cores and PCI Express 6.0, we're hardly in "personal computer" territory anymore. Indeed, it's clear that Mustang Peak is going to be an absolute monster when it eventually arrives next year, likely in mid-to-late 2027.

According to a report by Ars Technica, AMD has quietly stripped a critical security feature from its lower-end CPUs, leaving unaware users potentially vulnerable to physical attacks. Following a months-long investigation tracked on GitHub, Ben Kilpatrick confirmed that the Transparent Secure Memory Encryption (TSME) feature — which protects CPUs against physical exploits that siphon data from connected memory chips — was suddenly no longer available on AMD CPUs outside the company's Pro lineup.

As the exhaustive inquiry, which involved conversations with AMD engineers, board vendors, and other CPU users, was coming to a head, an AMD engineer abruptly cut discussions short, stating, "My apologies, but I don't have any more information to share on this topic." As of this report, AMD has neither officially acknowledged nor explained the disappearance of the security feature.

TSME is a protection feature that encrypts the data stored in memory, making it unusable to physical attackers. AMD initially added this feature to its high-end CPUs, then later extended it to lower-end CPUs. Eventually, the feature became a given, leaving lower-end chip users assured in its availability as part of the chip package. However, without prior notice, AMD appears to have scrapped the security feature in these processors.

According to the Ars report, the company's only official reaction to the matter — not counting the GitHub discussions — is an email response stating that TSME "is a security feature only applied to PRO CPUs as part of AMD PRO Technologies," notably the first time the company has publicly stated such a restriction, despite the feature having worked on consumer chips for years. However, it remains unclear whether the disappearance is an intentional policy decision by AMD to reserve TSME for Pro chips or an unintentional regression that was introduced in AGESA 1.2.7.0, a newer firmware release.

Another concerning aspect of the removal is that the feature's disappearance is completely undetectable on Windows machines and requires significant technical work to identify on Linux. That means the security feature was removed, leaving users unaware that anything had changed.

Kilpatrick, a self-described "privacy-conscious Linux hobbyist" who first reported the change, was installing a new operating system on his machine running a Ryzen 7 9700X from the Zen 5 architecture. To confirm that all his security protections were enabled, he ran Host Security ID (HSI), an auditing feature that evaluates a system's firmware and hardware security configurations. To his surprise, HSI reported that TSME was no longer supported — even though he had enabled it in his BIOS settings all along. The contradiction sent him searching for answers.

His first instinct was to reach out to MSI, his motherboard’s manufacturer, but the company didn't initially provide a definitive explanation. He also filed a bug report on AMD's public engineering GitHub repository, where two AMD engineers eventually responded: Tom Lendacky, an AMD fellow software engineer, and Mario Limonciello, an AMD senior principal software engineer.

Interestingly, neither engineer appeared to have a clear answer for why the feature had disappeared. Their advice was basically the same: disable and re-enable the option in the BIOS, and if that didn't work, take it up with the motherboard manufacturer, making it clear that people directly at AMD were just as in the dark as the user reporting it.

It was only after this that Kilpatrick pressed MSI harder, eventually convincing its engineers to run controlled tests. They found that consumer Ryzen chips had TSME enabled under an older firmware version but showed it as "not supported" under a newer one (AGESA 1.2.7.0), while Pro versions of the CPU supported the feature regardless of the firmware or motherboard used.

This leaves the big question of whether AMD deliberately restricted TSME to its Pro chips, or whether the change was an accidental regression — a firmware bug introduced in that newer AGESA version. Either way, the silicon appears to have been capable of running the feature. The difference is whether users are looking at a bug that AMD should fix or a quiet product-segmentation decision that AMD has not properly explained.

Kilpatrick took these MSI findings back to the AMD engineers and resumed the discussion six weeks later. MSI's product marketing team, he reported, had been told directly by AMD that TSME is exclusively supported on Pro series processors. He also relayed MSI's test results: an internal AGESA flag that controls whether TSME activates during boot returned FALSE on consumer chips regardless of the BIOS setting, but TRUE on Pro processors when the feature was enabled.

Kilpatrick then brought up something especially awkward. He reminded Lendacky of a comment that the engineer had made back in 2020, confirming that a Ryzen 3700X, a consumer CPU, “should support TSME.” In a later 2025 comment in the same discussion, Lendacky again recommended using TSME, while noting that the motherboard BIOS provider had to expose the option. So there it was, AMD's own engineer, years earlier, acknowledging the feature working on exactly the kind of lower-end chip now stripped of it, proving that Ryzen support was not some fantasy users invented.

After some more back-and-forth, Kilpatrick asked bluntly whether the flag being set to FALSE on consumer chips was a silicon-level limitation or a firmware policy decision — since one is permanent and the other is potentially reversible. Limonciello’s reply effectively closed the chapter. “My apologies, but I don’t have any more information to share on this topic,” he wrote.

To be fair to AMD, there is no clear indication that the company ever publicly advertised TSME as a consumer Ryzen feature. AMD has long said that a related memory protection, Secure Memory Encryption (SME), is available only in the Pro and EPYC CPU tiers. SME is OS-managed, using a single key and allowing the OS to selectively encrypt individual memory pages. TSME, by contrast, is firmware-managed, encrypting all RAM with no OS involvement. When active, it guards against physical attacks such as cold-boot exploits, DRAM interface snooping, and memory module removal, and it activates silently once enabled in the BIOS, making it the more practically useful of the two protections.

For now, AMD has said nothing official. It hasn't confirmed what happened, why it happened, whether anything actually changed, or what users of its consumer chips should now expect. Given the years of TSME quietly doing its job on these lower-cost processors — and the AMD engineers' supposed own past comments treating it as supported — users had every reason to regard it as part of the package.

For most consumer Ryzen users, the practical impact of the change is narrow. TSME protects against physical attacks, meaning scenarios in which someone has physical access to the machine or its memory hardware and attempts to extract secrets directly from RAM. The feature is more important for people carrying sensitive laptops, handling confidential work, relying on full-disk encryption, or operating in environments where seizure, theft, or hardware tampering is a realistic concern. Anyone who genuinely needs memory encryption on AMD hardware now appears to need a Ryzen Pro or EPYC system, unless AMD clarifies the situation or restores support.

We learned a lot about Intel’s upcoming plans for desktop CPUs at Computex 2026. In classic Intel fashion, we’ve already heard a lot about the company’s next-gen CPUs, codenamed Nova Lake, even while the recent Arrow Lake Refresh CPUs are still warm from the oven. But on the ground in Taipei, we heard not only more about Nova Lake and the Z990 platform it’s arriving on, but also how Intel intends to handle the rollout and how it will fill the gaps in its lineup with “Raptor Lake Next,” which is supposedly slated to launch next year.

Trade shows are the best opportunity to learn details about unreleased products before they show up in a press deck, and simultaneously the worst venue to do so. With jet-lagged representatives and reporters, thousands of people whizzing past, and the threat of Jensen Huang showing up to sign components and shut down a floor on a moment’s notice, it’s easy for things to get lost in the shuffle. So, we’re going to work through everything we learned about Intel’s upcoming plans in stages, starting with details that are confirmed, and working toward more speculative murmurs.

Intel has a fairly aggressive consumer roadmap, which the company itself would tell you – and the company told us as much at Computex, as a matter of fact. Both Nish Neelalojanan, senior director of client product management, and the recently joined Alex Katouzian, executive VP and GM of client, played up Intel’s roadmap to Tom’s Hardware, and for good reason.

Chronologically, Intel’s plans look something like this: We’ll see the first Nova Lake SKUs roll out at CES 2027. A few months later, we’ll see a refresh on the LGA 1700 socket with “Raptor Lake Next” CPUs, and come Computex next year, Intel will launch a 52-core flagship Nova Lake SKU. None of that is confirmed by Intel, and we have varying degrees of confidence in each step of the roadmap, so take it as speculation for now. We’ll dig more into the details we have and what’s simply rumored below.

What about AMD?

Before Intel, we should at least look at why we’re not talking about AMD’s next-gen desktop plans. Basically, we don’t have a ton of information on Zen 6 CPUs yet, and even less information about Olympic Ridge, the desktop consumer lineup of Zen 6 chips. Computex didn’t change that fact.

At Computex, AMD revealed the Ryzen 7 7700X3D, relaunched the Ryzen 7 5800X3D, and brought the RX 9070 GRE to the rest of the world. Unlike previous years, AMD didn’t hold a keynote, where we might’ve seen a more concrete tease of Olympic Ridge; AMD has already teased Zen 6 broadly several times. Bigger Zen 6 news is likely at the company’s Advancing AI event next month.

Although AMD hasn’t said when Olympic Ridge will launch, we originally expected it in late 2026. Now, 2027 is very likely. AMD has shifted the Zen 6 conversation toward its EPYC Venice chips, and confirmed production ramp on Venice in May. Although AMD traditionally leads with a consumer launch at the turn of a new microarchitecture, it’s unlikely that Olympic Ridge will launch before Venice. Demand for CPUs is spiking in the data center for agentic AI workloads, after all, and AMD is adjusting accordingly.

Olympic Ridge probably isn’t top of mind right now, from both AMD itself and its partners. AMD laid the groundwork for a unified CPU architecture generations back, and Intel’s approach has been a bit more disparate across client and data center (although that’s been changing with releases like Xeon 6 and Xeon 6+). We don’t know when Olympic Ridge news will arrive, but it almost certainly follows far greater detail about Zen 6 in the context of Venice.

What’s confirmed

Let’s start with the concrete details about Intel’s future CPU plans. These are things we have direct evidence for, be it photos, our own hands-on time, or sources we’re extremely confident in. At least two Z990 motherboards were at Computex, a third is rumored, and we saw (and held) what looked to be a near-production model in a closed-door meeting. And from that, we can already tell a lot about Nova Lake.

First, the LGA 1954 socket, which has now been pictured (we were told not to take pictures, but someone else did the dirty work, it seems). It’s the same size as the LGA 1851 socket, measuring 45 mm x 37.5 mm, and it retains compatibility with existing coolers, which we were able to confirm at Computex. It features more pins, as the name reveals, and uses the 2L-ILM, or two-lever Independent Loading Mechanism. The picture of the socket circulating matches what we saw at Computex.

The motherboard we saw featured dual 8-pin EPS connectors, along with an 8-pin PCIe connector near the bottom of the board, which is said to provide auxiliary power to the CPU. We’ve seen a leaked photo of the Z990 PCH now, which is said to draw more power due to broader PCIe 5.0 support. The Z990 board we saw, at least, had three PCIe 5.0 M.2 slots, along with three PCIe 5.0 expansion slots. Short of perhaps specialized designs with extra M.2 slots, we expect Z990 to support PCIe 5.0 across the board.

As for the chips themselves, all that is confirmed from Z990 motherboards is that Nova Lake can scale up to a high-end power design. We’ll speculate more on specific numbers later, but we’ve seen auxiliary power beyond two 8-pin EPS connectors on two Z990 motherboards now, and the motherboard we held had an extremely high-end VRM design; we can’t say more than that at this point.

An important caveat here is that we’re dealing with high-end motherboards and discussing how high the platform can scale, not how it will scale. Plenty of ink has been spilled about Nova Lake’s supposedly high power draw, but we really don’t have details about the chips themselves, rather just the tippy-top of the platform that will support them.

Outside of Z990 boards, Intel has confirmed that Nova Lake is “coming at the end of 2026.” That’s what CEO Lip-Bu Tan said at the company’s full-year 2025 earnings call back in January. What we were told by multiple vendors at Computex is Q1 2027, with a portion of those vendors specifically pointing to CES 2027. Similarly, with Z990 motherboards, some vendors said Q1 2027 while others said Q4 2026 (one even hinted at Q3). Believe it or not, these timelines actually all match up.

What’s lost in translation here is when the sale is happening. Before Nova Lake launches publicly, Intel and motherboard vendors will need to sell products into the channel, which, a few months later, will be available for sale at retailers for you to buy. What we’re likely looking at is sales into the channel in Q4, a public launch of Nova Lake at CES 2027, and retail sales in Q1. When Tan says Nova Lake is coming at the end of 2026 to a group of investors, he’s likely referring to selling into the channel, not the final retail sale.

What’s likely

Now, we’re getting into a bit more speculation. These are some of the details we heard about at Computex, or confirmations of previous rumors that we don’t have any concrete evidence for. Given the conversations we had at Computex, and a healthy dose of critical thinking, these are the details that are likely but not confirmed. There’s always a chance we’re just blind men touching an elephant on some of these points.

First, Nova Lake. For nearly a year now, it’s been rumored that the highest-end Nova Lake SKU will scale up to 52 cores. That’s the number we heard at Computex, as well, but not as a typical flagship. Rather, we heard that Intel plans to lead Nova Lake with a 28-core flagship, which will launch at CES 2027, and introduce a high-end 52-core model later in the year. The timeframe we heard was Computex 2027, but if anything is subject to change, it’s a release date that’s a year away. For now, let’s call it later in 2027.

The 52-core SKU will apparently come with 16 Coyote Cove P-cores, 32 Arctic Wolf E-cores, and a cluster of 4 LP-E cores; we didn’t hear that at Computex, nor anything to the contrary, but that’s what has been previously rumored. That model will reportedly come with two compute tiles, so the 28-core model with a single compute tile will likely look like an 8 + 16 + 4 split. That’s pure extrapolation at this point, however.

As for the 52-core model, we were told it comes with a PL1 of 175W and a PL4 of up to 700W. The PL1 number is what’s important here. Although that is a sizable increase over the 125W PL1 of both the 285K and 14900K, 52-core Nova Lake doesn’t sound like a direct replacement for those parts. Given the timing and extra power demands, it looks more like a spiritual successor to Intel Extreme Edition chips, targeting enthusiasts with deep pockets and the HEDT crowd.

Nova Lake is treaded ground at this point, however. Something new we learned about from Computex is “Raptor Lake Next.” After hearing the name, we asked Intel, which declined to comment on Raptor Lake Next at this time. Apparently, however, it will be the third refresh of Raptor Lake CPUs on the LGA 1700 socket, particularly targeting budget-conscious builders while Nova Lake satiates the enthusiast crowd.

There are some pieces of circumstantial evidence that point to a reintroduction of LGA 1700 CPUs. First, this has been previously rumored. In April, prolific leaker Jaykihn hinted at another Raptor Lake refresh coming in 2027. We’ve now heard that the range is called Raptor Lake Next from multiple sources, and it’s specifically coming in the first half of 2027, some months after the initial Nova Lake launch.

Additionally, multiple motherboard vendors told us that they’re ramping production of LGA 1700 motherboards, including DDR4 boards, though they didn’t say it was in relation to any new CPU releases. Intel itself has dropped a few hints, as well. Earlier in the year, Intel’s Robert Hallock said that Raptor Lake will be “abundantly available” in the market, and at Computex, Intel’s Nish Neelalojanan told Tom’s Hardware that Intel “will continue to make sure that there are products which can take care of older memory technologies.”

It would certainly make sense for Intel to refresh Raptor Lake a third time. Although data center demand is offsetting it, the decline in desktop sales from high memory prices hits Intel and AMD on the balance sheet as well. Just about everyone we spoke with at Computex talked about the state of memory prices, and Intel has a DDR4 platform that it’s still actively selling on the market. AMD, with a hard switch to DDR5 with Zen 4, has to reach back further to revitalize DDR4 options, but Intel already has a small ecosystem of DDR4 motherboards and CPUs available now, which it could easily bolster. We’ve heard that bolster is coming in the opening months of next year.

What that range looks like remains a mystery, however. It could be a proper refresh, or it could simply be an infusion of 14th-gen stock (and LGA 1700 motherboards) into the market along with new price points; both Raptor Lake generations have slowly crept up in price since the end of last year. The important thing here is that it seems Intel is targeting LGA 1700 for the lower end of the market, as Arrow Lake, with its underperformance and high price due to exclusively using DDR5, won’t provide the last-gen value bridge that previous generations have.

After Tom's Hardware originally broke the news about Raptor Lake Next, we followed up with Jaykihn, who provided a few specs.

*Naming unconfirmed by Intel, specifications rumored

The specs we've heard about are for the four SKUs above, which would comprise the main lineup of chips with integrated graphics enabled; apparently, Raptor Lake Next will include options with the iGPU disabled, as well as mobile chips. The final branding is unconfirmed, but we've heard that Intel intends to launch under the Core Ultra 200 name.

Out of the four SKUs, the 16-core Core 5 looks like Intel's breadwinner. Throughout 12th- to 14th-Gen, Intel topped out Core i5 models at 6 P-cores. You'd have to step up to a Core i7 for 8 P-cores. If these specs are correct, Intel is stepping down to an 8 P-core configuration a tier in branding, which will hopefully come with a cut to price.

What’s still up in the air

Some of the finer details of Nova Lake are still up in the air. That is, we don’t have any direct evidence for them, nor any corroboration from Computex. That’s not to say that the details here are false. Rather, we just need more information to say, for sure, that some of these details are a part of the Nova Lake lineup.

First and most obvious is bLLC, or big Last Level Cache. This is one of the earliest Nova Lake rumors that is still circulating, and for good reason. Intel hasn’t found an effective counter to AMD’s 3D V-Cache CPUs in more than four years. We’re closing in on half a decade where AMD has entirely owned the high-end of PC gaming, which has continually eaten away at Intel’s market share. bLCC is, apparently, Intel’s counter to 3D V-Cache, using its own Foveros 3D hybrid bonding to stack additional last-level cache.

Tom’s Hardware asked Intel CEO Lip-Bu Tan and a panel of executives at the company how it plans to address X3D CPUs, and Alex Katouzian, a 20-year Qualcomm veteran who recently joined Intel in a leadership role over the client group, said the following: “When I first came in and started reviewing road maps for the team, I was very pleasantly surprised. So, stay tuned, a very strong roadmap [is] coming, and we will be gunning for that section of the market as well. And so, please stay tuned.”

Context is important, but Katouzian is really only saying that Intel is gunning for high-end gamers with its roadmap, which, of course, it is. Otherwise, bLLC has entirely been a topic of the rumor mill. Intel has indirectly teased it with PR hits about its packaging capabilities, but that extends far beyond bLLC. Hybrid bonding, especially from a foundry perspective, has far greater legs in the data center.

Although Intel has the packaging and bonding capabilities, the scale of them for a mass-market product like Nova Lake is questionable. Intel would need to bond the SRAM to the logic tile with Forveros and package the chip with EMIB, creating the “EMIB 3.5D” combination that Intel has talked about previously. We first saw EMIB 3.5D on the Ponte Vecchio data center GPU, but most recently and relevantly on Clearwater Forest, Intel’s first foray into putting 18A in the data center. The capability is there, but if Intel can scale that up to a consumer range with more limited die space and higher per-core performance remains to be seen.

One advantage of Intel’s hybrid bonding and advanced packaging is that it can package dies from other foundries, not just those from Intel foundries. That brings us to the second finer point about Nova Lake, which is the node. Originally, the assumption was that Intel would use 18A for Nova Lake. We have 18A on mobile with Panther Lake, in the data center with Xeon 6+, but not on the desktop. Further, Intel has previously commented about reshoring its manufacturing for consumer chips after a brief stint with TSMC for logic tiles in both Lunar Lake and Arrow Lake.

Around this point last year, however, rumors started circulating that Intel is using TSMC’s N2 for Nova Lake. The source of the rumor is flimsy, however. Well-known reporter Charlie Demerjian of SemiAccurate reported in July 2025 that Intel taped out a major product. The report didn’t mention what product, what foundry, or even include “TSMC” anywhere on the page. Still, other outlets took the story, claiming that not only was Demerjian talking about Nova Lake, but also that he was talking about TSMC N2.

There are reasons Intel could use TSMC for the logic die. The company has reiterated that it’s shifting wafer capacity toward the data center, so if TSMC can fill additional capacity on the desktop, we could see TSMC on the main logic die. It’s also possible that TSMC is manufacturing other tiles on Nova Lake. Intel has consistently blended nodes in recent generations, so even if Intel were to confirm that it’s tapping TSMC for Nova Lake, that doesn’t necessarily mean the Taiwanese giant is manufacturing logic.

And, just as easily, Intel could absolutely be using TSMC for logic. That’s the point here; we really don’t know at this point, outside of vague reporting, getting swept up in the rumor mill, and taking on a life of its own. The Cinderella story for Intel would be Nova Lake on 18A, but given the struggles on 18A yields, it wouldn’t be surprising to see TSMC at the helm for Nova Lake once again.

Hurry up and wait

Intel needs a much more aggressive roadmap on the desktop than AMD, frankly, and that roadmap is starting to take shape. Although AMD and Intel compete on the finer points of performance, Team Red has almost exclusively taken market share away from Intel, quarter over quarter, for the past decade. There are only a handful of quarters in that time when AMD has lost market share, which it has always rebounded from in the quarter that follows.

Even if Intel still represents the majority of the desktop market — and it does based on the latest market research — the trend is abundantly clear. Add on top of that clear fumbles like Arrow Lake, and it’s obvious that AMD doesn’t need to move the needle much to continue swiping customers. Intel needs to make big moves to recover.

We should have more official details about those plans soon. Intel mostly sat Computex out on the consumer front, short of the Arc G3 range that, although exciting for gaming handhelds, is destined to be a niche product given the high prices of the devices those chips are going in.

For the past four years, Intel has held its Tech Tour event in the fall, taking the place of its previous Architecture Day, which took place in the late summer (most of those details have shifted to the Hot Chips conference in August). Intel has already told us that Hot Chips will have more details about Diamond Rapids, Intel’s next-gen P-core Xeons. That leaves Tech Tour for when we’ll likely get a full architectural deep dive on Nova Lake. Intel has yet to confirm Tech Tour 2026, but we have no reason to believe the company will sit out the rest of the year at this point. It also lines up with what we’re hearing about Nova Lake’s release — architectural details in the fall, a launch at CES 2027, and availability in Q1.

Regardless of when the exact dates fall, Computex made it clear that Intel is readying Nova Lake for a release soon. Multiple motherboard vendors brought Z990 motherboards to Computex and actively showed them to the press; I can’t imagine that was sanctioned by Intel.

As for Raptor Lake Next, Computex is the first quasi-confirmation we’ve heard of the range. That name apparently appears on Intel’s roadmap at some point in the first half of next year. With Nova Lake at the high-end and Raptor Lake Next in the midrange, Intel might have a one-two punch strategy to earn back some spots in the market, especially as AMD turns its Zen 6 focus toward the data center and prioritizes older architectures on desktop, given high DDR5 prices. Now, we just need to wait and see how those internal plans materialize as the rest of the year goes on.

We managed to grab pics of the newest upcoming server sockets from both AMD and Intel at Computex 2026. Both AMD and Intel are preparing to launch their next-generation server platforms that use all-new sockets, which enable new levels of performance, functionality, and power delivery.

AMD is a bit ahead with its SP7 platform in 2026, while Intel’s gargantuan 9324-pin socket will be used for Xeon ‘Diamond Rapids’ in 2027. While the platforms are entirely different, what makes them similar is the massive dimensions of CPU sockets and coolers.

AMD’s SP7 is the company’s next-generation socket that will support AMD’s 6^th Generation EPYC ‘Venice’ processors with up to 256 cores. The socket is huge and is rumored to support 16 DDR6 memory channels using 12.8 GT/s MRDIMMs as well as up to 96 PCIe 6.0 lanes (with the CXL protocol on top, though this is a processor, not a socket feature).

Based on information from Auras, the SP7 socket will be able to handle CPUs with a peak power consumption of up to 1,400W, so Auras and other companies are prepping liquid cooling solutions for these parts. In person, the socket is strikingly large, occupying most of my palm and overshadowing today’s server CPU packages.

Given the fact that the socket must support so many memory channels and PCIe lanes, it is not surprising that it is that large. Despite its enormous dimensions, the socket is still compact enough to enable dual-socket server designs, so AMD’s partners will be able to offer systems with up to 512 x86 cores as soon as its next-generation EPYC processors arrive later this year.

Meanwhile, for those systems that do not need so many cores and memory channels, AMD is prepping the SP8 platform that is set to offer fewer cores and DDR5 channels. Interestingly, Auras is working on water blocks for SP8 sockets as well, which means that the platform will still be quite mighty in terms of power consumption.

But while AMD’s SP7 is huge, Intel’s 9324-pin socket easily dwarfs it, as it is noticeably longer than the palm of my hand. The socket will work with Intel’s Xeon ‘Diamond Rapids’ processors with up to 192 cores, a 16-channel DDR5 memory subsystem supporting MRDIMMs, and PCIe Gen6 lanes.

Intel is yet to announce the processor base power of Diamond Rapids processors, though, since Auras is prepping water blocks for these CPUs, we're talking about circa 300W – 500W PBP and over 1 kW peak power consumption. Meanwhile, given that the socket is so massive, we would not be surprised if Intel’s 9324-pin socket will also support the Coral Rapids processors, presumably due in 2028 – 2029.

AMD on Monday announced that it had acquired MEXT, a startup that developed a memory tiering technology that makes NAND flash memory appear as DRAM to the operating system, which enables operators of data centers to save money on DRAM. AMD expects the acquisition to help customers improve system efficiency, lower operating costs, and deploy large-scale workloads more quickly.

As AI models continue to expand and datasets grow larger, memory availability has become an increasingly important factor affecting overall system performance. In many cases, memory resources, not CPUs or GPUs, are becoming a performance bottleneck. Meanwhile, in many cases DRAM is used inefficiently.

MEXT addresses memory efficiency challenges with an AI-based memory tiering technology that moves infrequently accessed data from expensive DRAM to NAND storage, which costs orders of magnitude less per unit of capacity, and in a way that's transparent to applications. MEXT's Predictive Memory Engine continuously analyzes memory access patterns and uses AI models to anticipate which data stored in flash will be needed next. Those memory pages are proactively transferred back into DRAM before applications request them and enable software to access data as though it were in main memory, thus preserving performance levels.

By increasing the amount of usable memory available to applications, MEXT's technology aims to improve utilization of existing infrastructure and at the same time reduce needs for expensive DRAM. This approach can potentially lower total cost of ownership for cloud providers and enterprise customers and enable larger workloads to run on existing hardware. AMD believes that these capabilities can benefit both traditional data center applications and modern AI deployments, where access to large memory pools is often critical for efficiency and scalability.

AMD plans to incorporate MEXT's technology into its data center product portfolio and expand its capabilities to address memory-hungry AI workloads. The company already offers integrated solutions that combine processors, accelerators, networking technologies, and software, so MEXT's Predictive Memory Engine will complement the already broad portfolio.

As an added bonus to the technology itself, AMD gains a team with expertise in memory architectures, infrastructure software, and large-scale computing systems. Terms of the deal are unknown.

Yesterday, we broke news of a third Raptor Lake refresh in the works from Intel, dubbed "Raptor Lake Next." Following that, new leaks have emerged from reliable tipster Jaykihn confirming the codename and detailing the upcoming family. Raptor Lake Next is slated for early 2027 and will land on the LGA 1700 socket as reported previously, but some outlets say it will also retain the existing Core 200 branding.

According to the rumors, Raptor Lake Next doesn't use Bartlett Lake silicon, which is Intel's special lineup for edge and embedded devices with only P-cores. That means a 12 P-core SKU has been ruled out for now, and that Raptor Lake Next will stick to the usual hybrid config. It's expected to scale across mobile and desktop with up to 125W parts and HX models as well, but it won't bring any new notable features.

Essentially, we're looking at rebranded 14th Gen CPUs that are likely being re-released because of their high yields. Intel spent a considerable amount of time and effort fixing the stability issues with 13th and 14th Gen families, so it makes sense the company wants to maximize its returns here. The Core-i9 14900K is still the most performant gaming CPU Intel has ever released, so calling this leftover silicon doesn't feel right either.

Jaykihn also provided further specs for the Raptor Lake Next, claiming it will only spread across Core 3, Core 5, and Core 7; no Core 9 SKUs in sight at the moment. At the top sits a Core 7 part with 8 P-cores and 12 E-cores, totaling out to 20 cores rated at 65W, same as the Core i7-14700. Then there's a 16-core 125W processor, which is identical to the Core i7-13700K, but it's counted as "Core 5" for Raptor Lake Next.

There's a special 10-core 65W part with 24MB of L3 cache because the previous 10-core silicon only had 20MB. It looks like the cache from disabled core clusters will still be accessible in certain Raptor Lake Next SKUs, and this might become a trend going forward. Lastly, we have a Core 3 CPU with 4 P-cores and no E-cores, running at 65W as well. Since all these chips are based on Raptor Lake silicon, they use Raptor Cove P-cores and Gracemont E-cores.

Keep in mind that RTL and RTL-R both came with Intel HD 700 series integrated graphics; the new Arc Xe architecture debuted with Arrow Lake on desktop. Nonetheless, Raptor Lake Next will work with DDR4 memory because it uses the LGA 1700 platform. That could help alleviate at least some of the burden of building a PC today as DDR5 RAM doesn't seem to be getting cheaper anytime soon.

Raptor Lake Next is apparently going into production in January 2027 so it could be available in Q1 2027, coexisting with Nova Lake as perhaps the value-focused option from the Blue Team. All rumors point toward Nova Lake being delayed to next year, but officially we still expect an announcement later this year. Regardless, the two families sure look like they'll overlap, which is why the Core 200 vs Core 400 distinction would be important for consumers.

Some of the most interesting topics at Computex happen around the announcements themselves, and that’s certainly the case when it comes to Intel’s plans beyond Nova Lake. At the event, Tom’s Hardware heard that Intel has plans to launch “Raptor Lake Next” CPUs within the first half of 2027. The CPUs would mark the third refresh of Intel’s Raptor Lake range, which first debuted with 13th-Gen chips. The chips will reportedly live alongside Intel’s upcoming Nova Lake range, which the company intends to introduce at CES next year.

We don’t know the proper name of the CPUs yet, just that they’re referred to as “Raptor Lake Next” and will arrive some months after the first Nova Lake chips have hit the market. At least two motherboard vendors confirmed to Tom’s Hardware plans to increase production of DDR4 motherboards, both for AM4 and LGA 1700, citing increased DDR4 demand. They did not comment on Raptor Lake Next.

At this time, we simply know the name and a rough timeline when the chips are set to release, which follows a rumor that circulated in April about Intel’s plans to re-introduce Raptor Lake CPUs, as well as comments from Intel’s Robert Hallock that Raptor Lake will be “abundantly available” in the market. At the moment, Raptor Lake Refresh still holds the distinction of being Intel’s best CPU for gaming, with the new Core Ultra 7 270K Plus just narrowly falling short of the Core i9-14900K in games.

The specifications remain a mystery, as well as what exactly the range will look like. Recently, Intel introduced Bartlett Lake for embedded and industrial applications, which use exclusively P-cores and slot into the LGA 1700 socket. The flagship Core 9 273PQE goes up to 12 P-cores, four more than the Core i9-14900K. Bartlett Lake chips are socket-compatible with Raptor Lake platforms, though not supported through software.

Nonetheless, some enthusiasts have managed to get Bartlett Lake chips working on consumer 600-series and 700-series motherboards. Although we don’t know if Bartlett Lake will make an appearance for consumer applications under a different name, the mere existence of the range confirms Intel continues to produce Raptor Cove-based wafers on Intel 7.

We’ve corroborated the name Raptor Lake Next, but we still don’t know if it will be an entirely new range of processors. AMD recently reintroduced the Ryzen 7 5800X3D, turning back to DDR4 amid memory shortages, and it makes sense for Intel to do the same. That could simply look like an infusion of stock into the market and new price points, however.

Keep in mind that plans can change. Although we’ve heard the name from multiple sources, as well as confirmed an LGA 1700 ramp with vendors, even small details can change weeks before launch. If, say, memory prices drop severely in the next few months, that would almost certainly change Intel’s plans. For now, however, this is the rollout we’ve heard about.

Intel declined to comment on Raptor Lake Next at this time.

Nvidia has told Chinese clients that its Arm-based Vera server CPUs could be available as soon as August and that orders can be placed now, Reuters reports, citing three sources familiar with the matter. Meanwhile, shipments of H200 AI GPUs to China remain frozen, months after CEO Jensen Huang said Nvidia’s market share in the country had effectively fallen to zero.

This August timeline runs in sync with what was said at GTC Taipei during Computex, when Nvidia indicated that Vera systems would reach customers through system builders and cloud partners starting this fall. Telling Chinese buyers they can have silicon in August, during a global server CPU shortage, suggests they’re sitting near the front of the allocation queue for a product line Nvidia expects to generate $20 billion in revenue by the end of its fiscal year in late January.

According to the report, Chinese cloud companies are already testing more than 300 Vera servers, and at least one major cloud provider plans to place an order. Initial deployments will be restricted to those companies' overseas data centers, one of the sources said.

If this goes ahead, Vera will reach Chinese buyers where Nvidia’s GPUs can’t. Server CPUs face far lighter U.S. export restrictions than the accelerators that underpin Nvidia's data center business, and the company's recent history in China shows that Washington is no longer the only obstacle. The U.S. licensed roughly 10 Chinese firms to buy the H200, but not a single unit has been delivered because Chinese officials, intent on nurturing domestic chipmakers, withheld approval on their side.

That dynamic helps to explain why the deployment of Vera CPUs will be restricted to overseas data centers: Chinese cloud providers want the hardware, but putting U.S. silicon into domestic data centers will obviously invite scrutiny and potential action from Beijing officials.

Vera began life as the CPU half of the Vera Rubin superchip, first shown at last year’s GTC event. Nvidia broke it out as a standalone product at GTC San Jose this March, launching it alongside a rack design that packs in 256 liquid-cooled Vera CPUs and sustains more than 22,500 concurrent CPU environments. Then, at Computex, Nvidia said the chip had entered full production, claiming 1.8 times faster task completion than x86 processors on agentic workloads. Its predecessor, Grace, has shipped nearly 2.5 million units to date.

Meanwhile, server CPUs are being tightly squeezed by the shift of AI workloads from training toward inference and agentic execution. Agentic AI leans heavily on host processors for tool calls, code execution, and data handling, and CPU demand has outrun supply as a result of the agentic explosion.

Intel has quoted Chinese customers lead times of up to six months, while AMD has said that the global CPU market is tight, with demand outpacing its forecasts and supply constraints expected to persist.

AMD has shared the first official benchmarks for its forthcoming EPYC 'Venice' CPUs, which will be the first chips to use the Zen 6 architecture. The flagship 256-core model hasn't been detailed in full, but AMD claims it offers 3.3 times the performance of the Nvidia Vera CPU in a rack-scale implementation with a fixed power budget of 100kW.

The constraints of this test completely change the framing of the results, so it's worth highlighting them first. AMD is looking at performance from the level of a rack, not an individual component on a single socket. AMD's results are modeled around a 100kW deployment, showing the performance across the rack rather than the performance on a single socket, or even a dual-socket system.

AMD did not, however, actually test all of these deployments. There's a lot of modeling going on here, which AMD details in its methodology paper that was published alongside the results. First, AMD estimated power based on the processor TDP and additional components, and it used that to calculate the number of nodes (2P system for each node) within a 100kW power budget. Then, it multiplied that number of nodes by single-node performance measured in a handful of benchmarks.

That's just the start of the stipulations. AMD doesn't have its hands on Vera, so the performance here is an estimate. AMD took benchmarks it had for Nvidia's Grace chip and multiplied them by a scaling factor of 1.63x based on Vera results published by Phoronix. AMD also says that its 256-core EPYC Venice results are derived from an estimated 1.7x scaling factor over the EPYC 9965, along with "internal testing."

AMD's methodology ends with this line: "[These results] are intended to provide directional comparison rather than direct measured rack benchmarks." As you can tell from the three paragraphs of stipulations above, that sentence carries a lot of weight on its shoulders. You can't simply scale the performance of a single "node" (a dual-socket system, in this case) up in a linear fashion. Interconnects, as well as thermal and power limitations, will become a factor as you scale up.

Regardless, AMD frames the results here around agentic AI, though the benchmarks it's using are focused on general-purpose data center tasks. The topline result is from SPEC CPU 2017, specifically looking at integer throughput. AMD used the following benchmarks, as well:

• Server-side Java based on SPECjbb 2015
• WRK Tool for load on an NGINX web server
• Redis-benchmark for in-memory workloads
• Memory caching with Memcached
• Database performance with TPROC-C on MySQL

These "results" are mostly a way for AMD to fire back at Nvidia after Phoronix published a list of results for Vera that were curated by Nvidia. AMD is laying the groundwork for its Advancing AI event next month, where we expect to hear more about Venice, Zen 6 more broadly, and AMD's enterprise roadmap in much greater detail.

Intel is broadening support for its Binary Optimization Tool, or iBOT, with seven new games. Across the newly-supported titles, Intel claims an average performance jump of 12%, and increases as high as 27%. To unlock support for the new games, you'll need the latest Intel Platform Performance Package. Then, you'll be able to select the games under the Advanced tab in the Intel Application Optimization GUI.

We originally saw iBOT roll out with Arrow Lake Refresh chips like the Core Ultra 7 270K Plus, but the feature works on Panther Lake chips, as well as Intel's latest HX mobile chips, Core Ultra 9 290HX Plus and Core Ultra 7 270HX Plus. Older Intel CPUs can't use iBOT due to hardware registers Intel must tap into for optimization. However, Intel says it plans on supporting the feature on future CPUs.

Intel initially rolled out iBOT with support for 12 games, which has now expanded to 19 titles. In our iBOT testing, we found that the feature offered an 8% average performance improvement, with the feature climbing as high as an 18% improvement in Shadow of the Tomb Raider.

Here are the games Intel added with the latest update:

• Metro Exodus Enhanced Edition
• The Callisto Protocol
• Homeworld 3
• Ori and the Will of the Wisps
• Little Nightmares III
• Warframe
• Hollow Knight: Silksong

Across the seven new games, Intel claims an average performance improvement of 12% with the Core Ultra 7 270K Plus. Intel tested at 1080p with High settings, and it used 32GB of DDR5-7200 memory and an Nvidia GeForce RTX 5090.

At the higher-end of the spectrum, Intel saw a 27% jump in Hollow Knight: Silksong and 16% improvement in Warframe. On the other end, Metro Exodus showed just a 2% uplift, while The Callisto Protocol jumped by 8%.

Intel's approach with iBOT might seem a bit scattershot, with 19 games now supported and no clear pattern between them. There are some popular and widely-played titles, such as Warframe and Cyberpunk 2077, but also less-demanding games like Silksong and older titles that don't see much play these games, like The Callisto Protocol. Even as the list expands, this strange mix is likely to stick around.

That's due to how iBOT works. It's essentially a translation layer, not dissimilar from Microsoft Prism. However, Intel isn't translating instructions from one ISA to another. Instead, it's optimizing x86 applications to ensure they're using the latest, most efficient instructions. Intel does this optimization with Hardware Profile-Guided Optimizations, or HWPGO, tapping into the silicon to see what's going on while code is executing. Once Intel has its optimizations, it packages them into a profile and ships.

Although iBOT is a unique feature, it's not universally applicable. Some games (and applications, as iBOT works outside of games) are already optimized enough that iBOT provides no benefit. We've also seen performance differences between chips, even between the Core Ultra 5 250K Plus and Core Ultra 7 270K Plus. It would be interesting to see how the feature works on more constrained chips, such as Panther Lake.

The AI revolution that shows no signs of stopping appears at times to have echoes of the gold rush. Whisper networks spread quickly through communities about new scarce commodities, and suddenly there’s a surge of interest as people snap up resources. For most of the ChatGPT era, you’ve struggled to get hold of a GPU for neither love nor money, with Nvidia practically able to manage its own waitlist, so great is the demand.

Much of the media’s attention – and plenty of investment – has been focused on the dash to grab as many GPUs as possible; most recently, memory has become a focal point.

But in recent weeks and months, there’s been a focus on ensuring that people have CPUs to match. For decades, the CPU has been the anonymous workhorse of the hardware stack, running operating systems, scheduling workloads, and keeping everything ticking over, rarely grabbing headlines unless there’s a supply crunch or a generational leap in performance.

Suddenly, it’s being talked about in the same breath as scarce-as-gold GPUs. What’s going on?

“AI deployment at scale has forced organizations to look at the infrastructure underneath the hype,” said Jason Beckett, chief technology officer in Europe, the Middle East and Africa at Hitachi Vantara, in comments to Tom’s Hardware Premium. As Beckett points out, while most of the attention is focused on GPUs because they run the AI models, the CPUs are vital because they handle “everything else”.

And as agentic AI becomes the norm, there’s a greater need for that CPU backbone to keep things running properly. “Always-on, multi-step reasoning systems don't create brief orchestration bursts around GPU workloads,” said Beckett. “They demand high-core-count CPUs running at sustained loads, continuously. The infrastructure requirement was always structural. It's just now unavoidable.”

Readjusting ratios

When data centers were previously being specced to deliver AI training and inference in the early days of the generative AI revolution, those building them accounted for a gargantuan bias in favor of GPUs. Chatbot conversations required between four and eight GPUs to every single CPU required, because the parallel equations required to meet user requests were GPU-inference heavy.

But as the main use case of AI changes from chatbots to agents, the requirements have also altered. A slight delay for in-depth inference while an AI model ‘thinks’ was seen as an acceptable interface choice. But as agentic AI requires rapid responses and the smooth coordination of tool calls and much more, latency can be a killer. Bolstering CPU counts can help avoid any problems that can quickly spin out into something more significant, breaking the entire agentic stack.

AMD, one of the major manufacturers of CPUs, has seen that shift first-hand. The company had previously forecast that the CPU market would grow at a rate of around 18% annually, but says that the change in requirements has materially changed the market. The rate of growth has now doubled to 35% a year, AMD claims, and will become a $120 billion market by the end of the decade.

“What AMD and Arm's results are telling us is that this is a structural, not cyclical requirement,” said Roger Cummings, CEO of PEAK:AIO, in an interview with Tom’s Hardware Premium. “In actuality, two structural shifts are driving the demand surge: the rise of agentic AI and the need for deterministic, predictable performance at rack scale.”

Much of that CPU demand is being driven by hyperscalers, who recognize the integral role that CPUs play in developing the AI clusters that are likely to power the economy in the years to come. “As GPU clusters scale, CPUs are taking on larger roles in orchestration, memory management, networking, storage coordination, and inference handling,” said Jeff Moore, vice president of strategic partnerships at Aegis Cooling, which specializes in next-gen liquid cooling solutions for AI and high-performance computing infrastructure, in an interview with Tom’s Hardware Premium.

There’s a rise in CPU-to-GPU ratios inside AI deployments, said Moore, “particularly because distributed AI workloads generate significant demand for general-purpose compute, memory bandwidth, and east-west data movement.” A recent TrendForce analysis points out that CPUs’ contribution to latency – accounting for nearly 91% of all the delay in responses – is something that AI deployments are trying desperately to counteract.

Changing designs

That shift is now visible not just in financial forecasts, but in the physical design of AI infrastructure itself. In early generative AI deployments, racks were often built around dense GPU configurations, with CPUs effectively treated as supporting components – enough to keep the system running, but not a bottleneck concern. Things are shifting now. “In the media, an AI rack is pictured as a giant box of GPUs,” said Hommer Zhao, founder of OurPCB, a PCB manufacturer with more than 15 years’ experience, in comments to Tom’s Hardware Premium. “But from a hardware design perspective, a GPU is just a very fast, very dumb engine. It cannot talk to the internet or pull data from a hard drive.”

Rather than a single host CPU loosely paired with multiple GPUs, hyperscalers are deploying configurations with higher core-count CPUs, more memory channels, and, in some cases, multiple CPUs per node to keep pace with data movement demands.

There are also thermal and power considerations shaping how racks are populated. High-core-count CPUs, especially those optimized for cloud workloads, are being selected not just for raw performance but for efficiency under sustained load. In liquid-cooled environments, CPUs are increasingly part of the same thermal design envelope as GPUs, rather than an afterthought cooled separately with air.

Financial signs of success

Recent results from AMD and Arm reinforce the idea that this is not a short-term correction but a deeper architectural shift. AMD has reported strong growth in its data center CPU segment, driven in large part by hyperscaler demand for its EPYC processors, which offer high core counts and memory bandwidth well suited to AI orchestration tasks.

Arm, meanwhile, is benefiting from hyperscalers designing their own custom silicon. “Arm accounts for close to half of all compute shipped to top hyperscalers in 2025, with over a billion Neoverse cores deployed,” said Beckett. “Those are rack-level architectural decisions made years ago.” AWS’s Graviton, Google’s Axion, and Microsoft’s Cobalt chips all reflect a move toward CPU architectures tailored for specific workloads: high-throughput, energy-efficient, and tightly integrated with networking and storage. Arm’s licensing model positions it at the center of this trend, and its recent financial results highlight how significant that hyperscaler-driven demand has become.

Both sets of results point to a change in how CPUs are being valued. In traditional enterprise contexts, the hardware was often general-purpose and interchangeable. In hyperscaler environments, it’s becoming a specialized infrastructure component, tuned for specific roles within AI systems, whether orchestration, inference at the edge, or data preprocessing.

Taken together, the changes in rack design and vendor performance suggest that CPUs aren’t a secondary consideration in AI infrastructure planning any more. Instead, they are becoming a critical factor in determining overall system efficiency and cost.

“The spotlight hasn't revealed something new,” said Beckett. “It's just finally illuminating what serious infrastructure teams never stopped building on.”

Probably the most important processor in PC history was introduced on this day 48 years ago. June 8, 1978, marked the birth of the x86 architecture with the arrival of the 16-bit Intel 8086 CPU. This lineage continues to the majority of PCs today, almost half a century later. Ironically, this chip and its x86 architecture, the result of 18 months of R&D, was meant only as a stopgap because Intel’s complex, clean-sheet 32-bit iAPX 432 project was delayed.

The Intel 8086 was designed by a team of four engineers and 12 layout people led by Stephen P. Morse. Reports indicate that the impetus behind this project was to provide a practical, timely alternative to upcoming 16-bit Motorola and Zilog CPU designs. The fabled 8086 processor was only meant as a stopgap, as Intel had bitten off a bit more than it could chew with the iAPX 432 project, begun a year prior. As a side note, the 432 finally shipped in 1981 and was deemed too expensive, too complex, and fatally too slow when it arrived.

The 8086, the founding CPU in the x86 lineage, was marketed as Intel’s first 16-bit processor. It benefited from a degree of backwards compatibility with prior Intel 8-bit designs like the 8008, 8080, and 8085. Notable advancements over its predecessors included microcode for multiply and divide assembly language instructions.

Looking closer at the hardware tech specs, the Intel 8086 had around 20,000 transistors (29,277 including ROM and PLA) and was manufactured using Intel’s HMOS (High performance MOS) manufacturing process, originally developed for manufacturing fast static RAM products. The resulting 40-pin chip measured 33mm², and the minimum feature size was 3.2μm. Over its lifetime, it was released in clock speeds ranging from 5 to 10 MHz.

While the Intel 8086 founded the x86 architecture, the subsequent 8088 design (1979) would become the beating heart of the first IBM PC (1981) and that particular storied lineage.

Direct 8086 successors like the 80286, 80386, and 80486 would spearhead the Wintel alliance and establish the PC compatible as the default choice for productivity, home computing, and computer gaming enthusiasts until being sidelined by the Pentium CPU (also x86) from the mid-90s onwards.

Intel released its Core i7-8086K 40th anniversary chip in 2018. What next?

To celebrate 40 years since the original 8086, Intel launched the Core i7-8086K 40^th anniversary chip in 2018. It looks like a pretty safe bet that we should get another tribute in 2028, marking the half-century anniversary. What will we get for the 50^th? Something that embodies the fun and enthusiastic side of PCs, we hope.

Also, it will be interesting to see if Arm processors begin to impinge upon the dominant x86 designs from the likes of Intel and AMD in the Windows PC market in the next couple of years. We’ve had Windows-on-Arm efforts from Qualcomm and Mediatek try to usurp x86 with muted success.

At the recent Computex 2026 there was a lot of buzz about Nvidia’s RTX Spark Superchip, a powerful new Arm platform designed to transform Windows 11 into an agentic AI operating system. Looking back two years from now, will Nvidia and its partners have started to turn the tide against x86?

The Core i7-6700K was once one of Intel’s best CPUs for gaming, but fast forward to 2026, and the decade-old chip is now hardly an optimal choice to pair with any modern GPU hardware, including some entry-level graphics cards. Despite this, YouTuber TrashBench put the 6700K to the test and challenged himself to see if he could overcome CPU bottlenecking with an RTX 3080 purely through overclocking.

In his tests, the YouTuber pushed the Core i7-6700K to its absolute limit, exceeding safe voltage limits in an attempt to achieve his goal. Three overclocks were tested: a 4.7GHz overclock at 1.4v, a 5GHz overclock at 1.56v, and a 5.1GHz overclock at 1.65v. The YouTuber even attempted 5.2 GHz and 5.3 GHz overclocks with a vcore of 1.7 V, but sadly, neither attempt was successful. As for cooling, the 4.7GHz overclock was achieved with an air cooler, while the 5GHz and 5.1GHz overclocks were achieved using a custom loop submerged in an ice bath.

At stock speeds, the 6700K was barely able to get the RTX 3080 above 60% utilization in Cyberpunk 2077, with an average of 103 FPS. With the 4.7GHz overclock, the bottlenecking situation noticeably improved, with the RTX 3080 achieving nearly 70% utilization, resulting in 13% higher performance compared to stock. At 5GHz, the 6700K improved the RTX 3080’s utilization to 74%, with Cyberpunk 2077 performance improving by 17%. Finally, at 5.1GHz, the fastest overclock unfortunately yielded the least impressive result, improving performance by just 1 FPS compared to Cyberpunk 2077 running at 5GHz on the 6700K.

The YouTuber also tested several other games featuring Shadow of the Tomb Raider, Hitman 3, and Far Cry 6. 3DMark Time Spy was also thrown in as a synthetic test. With air cooling at 4.7GHz, all of the aforementioned titles achieved an average performance gain of 7% compared to stock. In 3DMark Time Spy, the results were far different, with the 4.7GHz overclock achieving a whopping 19% performance improvement over stock. For the 5GHz ice-cooled results, performance improved by 11.3% on average over stock across the same three titles, and in 3DMark, the results showed a 24% improvement over stock.

The YouTuber unfortunately failed to achieve his goal of stopping CPU bottlenecking with a 6700K on an RTX 3080 through CPU overclocking alone. However, the overclocking results demonstrate that the Skylake flagship is still a capable gaming CPU, achieving around 100 FPS in all of the aforementioned games at 5 or 5.1GHz.