The rack design is standardized, based on the Open Compute Project’s ORv4 specification, which is slightly wider and taller than a standard IT rack. Helion is designed to support 44 OpenU trays, which are 21 inches wide and 48mm tall, slightly larger than a standard EA1-310 tray. AMD says the reference design supports both air-cooled and direct liquid-cooled deployments.

The design is also compliant with OCP’s DC-MHS (Data Center Modular Hardware System) standard. MHS is designed to bring “interoperability between key elements of data center, edge, and enterprise infrastructure,” according to OCP.

Although AMD has renders of the design you can see in the presentation below, it won’t actually sell racks. This is a reference design that AMD is contributing to OCP, and OEMs and ODMs will actually bring that design to market. AMD didn’t share partners supporting Helion during a pre-briefing with the press. The company says CEO Lisa Su will announce partners during the lead keynote at the OCP Global Summit later today.

AMD also didn’t share deployment partners, though it suggested that Helios customers will also deploy Helion racks.

Unlike Helios, which is built for AI training and inference, Helion is a more flexible rack solution primarily focused on CPUs. In a CPU-only deployment, the design can hold up to 128 EPYC 9996 CPUs, bringing that impressive 32,768 core count per rack. This configuration would likely look something like an agentic AI sandbox system, which is then connected to an inference system.

This model for personal AI agents is in use today, both in Meta’s Muse and OpenAI Dots. Each user has a sandbox with a slice of compute from the host system, with the backbone of an inference server. Some workloads are handled on those CPU-only servers, such as building software, but the heavy-duty AI inference happens elsewhere.

Presumably, this is the model AMD is aiming at with Helion as hyperscalers build out CPU capacity on the back of agentic AI. However, the design is flexible, supporting up to 16 MI350P GPUs and 102 Tbps of networking, though at the cost of CPU capacity. This hybrid configuration is targeted at enterprises, says AMD, giving them an all-in-one rack for data management, networking, and AI usage.

AMD says the Helion rack will be available starting in the first half of 2027, likely riding the wave of Zen 6 deployments that are expected to start ramping up later this year.

Full AMD OCP 2026 presentation

Intel, with its Pentium 4 processors, was clearly winning the MHz race at the time, while a CPU’s clock speed had long been established by marketers as a key performance metric. Unfortunately for AMD, its CPUs of the era didn’t hit such high clocks, even though the application and gaming performance was just as good (if not better) than Intel's.

Thomas Pabst's view

Our tests from October 2001 noted that AMD wisely undersold its chips, using the newly devised PR naming system. Anything less and it would have faced strong opposition and derision. Our tests clearly showed that a new “AthlonXP 1800+ is able to beat Intel's flagship Pentium 4 2 GHz in the majority of benchmarks,” wrote the original Tom, Thomas Pabst, in a 13-page feature about AMD’s newest CPUs. Moreover, it was observed that “AthlonXP at 1533 MHz is able to reach or even beat Pentium 4 running at an almost 33% higher clock speed.” Pabst concluded that AMD’s clocks/performance situation forced this concerted stand against the MHz myth.

As we broke the news about AMD’s Performance Rating system, some almost straight-away alluded to it being a ‘Pentium Rating’ system, but it was very early days. We didn’t know whether it would become a long-term success. It was known that AMD planned to establish an independent institution to create a new performance measure that is fairer and more realistic than the current MHz system in the following year. But, of course, Intel wouldn’t be interested in validating that. Pabst concluded that “I might not like it too much,” but admitted AMD had little alternative marketing-wise in the Pentium vs. Athlon era at the beginning of the millennium.

Please stop, it hertz

The AMD PR system would be abandoned roughly five years after it was initiated. Subsequent AMD (and Intel) architectures stopped the rating from making practical sense as the market evolved. For example, Intel’s big MHz advantage was lost in the transition from Pentium 4 Netburst to Pentium M/Core chips. Then, the Athlon 64 was so well received by reviewers and enthusiasts that AMD probably felt it might be wise to finish with its inferiority complex PR system that tied it to its obvious rival. Lastly, by the end of the AMD PR era, the marketing fight was clearly going multi-core. CPU marketing and naming completely sidelined any hint at CPU clock speeds later in the noughties.

"We are going to talk about our broader partnership," said Lisa Su when asked by reporters about what she planned to discuss with Samsung's executives. "We have many partners at Samsung."

Before visiting Samsung headquarters, Lisa Su and Jack Huynh, senior vice president and general manager of AMD's Computing and Graphics Group, attended the AMD Gamer Day event where Su and Huynh answered some questions from South Korean journalists and called local memory makers 'important strategic partners,' something that reflects the significance of data center-grade DDR5 and HBM memory for the company's AI and data center businesses.

"Samsung Electronics and SK hynix are key supply-chain partners for AMD," Su said, Chosun reports. "AI demand is extremely high, so we are planning our long-term supply chain with our partners. […] It is important to plan three to five years ahead with our partners."

The non-binding MOU that AMD and Samsung signed this March covered four areas and remained in effect for a period of three years:

• Supply of next-generation memory products, including HBM4 and advanced DDR5;
• Co-development and optimization of memory solutions for Al and high-performance computing systems;
• Exploration of foundry manufacturing opportunities for the future AMD platforms;
• Co-optimization of advanced packaging and related semiconductor technologies.

While Samsung has become a principal supplier of HBM4 for AMD's Instinct MI400-series accelerators, expansion of collaboration beyond memory supply is considerably more complicated, as it involves considerable efforts to design chips for Samsung Foundry's process technologies as well as risks associated with a product ramp at a new contract chipmaker. Truth to be told, AMD already uses Samsung Foundry and UMC to make its "ICs in the form of programmable logic devices" however, moving production of more advanced products to Samsung is risky. Meanwhile, Samsung badly needs big customers like AMD, and the latter's dependence on HBM supply certainly gives the former leverage.

"If we transition the production of some of our products to new manufacturers, we may experience delayed product introductions, lower yields or poorer performance of our products," AMD wrote in its most recent Annual Report.

AMD has already started shipments of its Instinct MI455 accelerators with 432GB of HBM4 memory onboard as well as rack-scale Helios AI systems to interested parties.

"We have started shipping our next-generation MI455 and Helios systems," Su told journalists, according to Chosun. "These are all systems that use HBM4."

So, it looks like the company is negotiating for more memory supply. The question is, will it take a risk and adopt Samsung Foundry for something more serious than niche programmable chips in a bid to get more high-bandwidth memory and sell more expensive AI accelerators? Only time will tell.

AMD's Verano is quite different from conventional EPYC processors. The CPU will feature up to 72 Zen 6 cores operating at up to 5 GHz and a massive 24-channel LPDDR5X memory subsystem that uses replaceable SOCAMM2 modules. AMD positions Verano as a host processor for AI servers, where CPUs primarily orchestrate workloads running on AI accelerators rather than provide maximum general-purpose performance themselves. Such specialization could explain why AMD developed SB1 instead of using SP7, which accommodates up to 256-core EPYC 'Venice' processors, or SP8, designed for models with up to 128 cores. Meanwhile, AMD has not disclosed SB1's specifications, dimensions, or capabilities just yet.

In fact, given Verano's highly specialized role as a host processor for AI accelerators, AMD may not need to offer a broad range of Verano SKUs typical of its general-purpose EPYC families. If Verano is deployed in systems with a predetermined number of Instinct accelerators, AMD could offer just one or a small number of CPU configurations to ensure a consistent ratio of CPU cores, memory bandwidth, and host-to-accelerator connectivity per GPU and simplify power, cooling, and system qualification, something that Nvidia does with its Grace and Vera processors.

If AMD decided to take a page from Nvidia's book, this explains why it intends to use a dedicated SB1 socket instead of using existing ones. Rather than supporting a broad range of CPU configurations compatible with SP7 and SP8 sockets and systems, SB1 could be optimized for one or a few CPU SKUs, their 24-channel LPDDR5X memory subsystem, and a particular system topology. Note that AMD has not announced how many Verano SKUs it plans to offer, so consider our thoughts speculation for now.

Dynatron's preliminary SB1-4U-Active cooler provides the first indication of what cooling Verano will require. The 4U unit measures 128.05 × 106.6 × 130.7 mm and uses an aluminum fin stack combined with a vapor chamber and heat pipes. The unit is equipped with a double-ball-bearing fan that operates at up to 6,000 RPM and delivers up to 91.7 CFM of airflow, which makes it a quite powerful device.

It is noteworthy that these specifications are virtually identical to those of Dynatron's J26 4U cooler for AMD's much larger SP7 CPUs, to the degree that even the dimensions are similar. This may either indicate that Dynatron reused the J26 design for SB1, or simply copied and pasted J26's specifications onto the SB1-4U-Active cooler's product page to use them as placeholders until the actual cooler is finalized.

For now, Dynatron does not specify the cooler's maximum supported processor power, so its listing cannot currently be used to determine Verano's TDP with confidence. Nevertheless, a large vapor-chamber-based 4U active cooler suggests that Verano will be a server processor with a rather high power consumption despite having considerably fewer cores than the higher-end EPYC 'Venice' CPUs for general-purpose workloads.

AMD plans to launch Verano in the second half of 2027.

Short of a few questionable Geekbench leaks, we haven’t seen any performance results for the RTX Spark yet, so AMD isn’t using it as a comparison point. Rather, it’s comparing the Ryzen AI Max+ Pro 495 to Intel’s Core Ultra X9 388H. These chips aren’t in the same class of device, though AMD would argue that it’s comparing its top-of-stack part to Intel’s top-of-stack part.

AMD used ComfyUI to measure generative AI performance. AMD averaged multiple runs of various models, comparing total throughput to Intel’s competition. AMD used its top-spec 192GB configuration of the Ryzen AI Max+ Pro 495 and compared it to the Core Ultra X9 388H in a system with 64GB of memory (Panther Lake supports up to 128GB).

The performance advantage ranges from 1.1x up to 32.2x, though the end point is a clear outlier. We searched for Yuve on Hugging Face and didn’t find any results. It’s possible this delta comes down to an optimization issue, or that it’s simply too big to run on the Panther Lake machine.

We don’t have gaming or general application performance, but we don’t expect a major swing compared to last-gen Strix Halo chips. Gorgon Halo is largely a refresh of that range.

Gorgon Halo isn’t getting into the ring with Panther Lake, however. It’s going mainly against the RTX Spark, and to a lesser extent, Apple’s larger M-series SoCs. This category of agentic PCs, as AMD calls it, is seemingly expanding, though it’s still far smaller than some of the hype around the RTX Spark would have you believe. AMD bragged in a prebriefing with the press about shipping “10s of millions” of AI PCs (read: laptops) before clarifying that it had slipped “over half a million” agentic PCs. Presumably, those are numbers for Strix/Gorgon Halo devices.

The matchup between Gorgon Halo and the RTX Spark has, up to this point, focused mainly on memory capacity. RTX Spark devices top out at 128GB of unified memory, same as the GB10 in the DGX Spark (the two chips are nearly identical). AMD, on the other hand, supports up to 192GB with Gorgon Halo. Higher capacity means running larger models locally, though at a lower performance level.

In the GLM 5.3 Flash with 320 billion parameters, AMD saw peak throughput of 20 tokens per second, though using Unsloth's UD-IQ4_XS mixed-quantization format. For context, we clocked peak token throughput on the DGX Spark running GPT-OSS 120B with 4-bit quantization at 64 tokens per second, and the last-gen Ryzen AI MAx+ 395 at 56 tokens per second.

Note that although GLM 5.3 Flash has 320 billion total parameters, only 18 billion are activated for each token. Similarly, GPT-OSS 120B is another mixture-of-experts model with 120 billion total parameters, though only around 5 billion are active per token.

AMD also shared Qwen 3.8 Flash Next performance, a multimodal MoE model with 125 billion main parameters, 51 billion embedding parameters, and about 4 billion parameters for multi-token prediction (MTP), with 5 billion parameters active per token. AMD says the Ryzen AI Max+ Pro 495 achieves up to 42 tokens per second with this model, once again using Unsloth’s dynamic 4-bit quantization and MTP.

That’s solid performance, but in both cases, “up to” carries a lot on its shoulders. The story of token throughput is told as the context length increases, showcasing what happens when someone actually runs these models locally, not just boots them up cold. Performance drops at higher context lengths, naturally, which could pose some issues for the larger models. If GLM 5.3 Flash provides up to 20 tokens per second, it could very easily decline into unusable territory as the context length increases.

We largely know what performance to expect out of the Ryzen AI Max+ Pro 495, and Gorgon Halo more broadly. It’s a refresh of Strix Halo, with notable spec changes being the bump up to 192GB of unified memory from 128GB, as well as a 100 MHz jump on boost clocks for the 495. Otherwise, the range is using identical core counts and microarchitectures as previous-gen Strix Halo chips.

With these proxies — Strix Halo for Gorgon Halo, and GB10 for RTX Spark — we can already get a good idea about how these parts will stack up. As you can see in our Ryzen AI Halo review (packing the Ryzen AI Max+ 395), AMD’s part universally underperformed compared to the DGX Spark in both time to first token and tokens per second across three models. More unified memory will allow you to run larger models, but that doesn’t mean those models will run faster.

The first Gorgon Halo devices are available for sale now, such as the Minisforum MS-S1 Max-P495. For the top-line configuration, prices sit around $7,000 right now, though we expect a broad range of prices once different devices are available, likely driving above that $7,000 mark.

Full presentation

We’ve also seen increases in “weird” memory configurations — 24GB has grown to 2.56%, increasing by 0.34%, while 48GB inched up to 1.46% after a 0.26% increase.

This comes as a surprise, as some manufacturers have reintroduced laptops with 8GB of RAM even as Microsoft called 16GB the new “practical starting point,” and recommended 32GB as the future-proof “no worries” configuration for gaming. It should be noted, though, that the company quietly erased this suggestion after it released 8GB base models for its Surface laptops a few months later.

It’s quite possible that memory capacities were always destined to increase for gamers, especially as titles become more demanding. Minecraft, which hasn’t updated its system requirements for 17 years and is known for being potato-PC-friendly, even increased its recommended memory to 16GB earlier this year, forcing players on older systems to switch to the Java Edition or play retro games instead. Because of that, the memory shortage may have merely postponed the inevitable reign of 32GB RAM on gaming PCs.

Another metric we’ve been following in the survey results is AMD’s slow but relentless march toward taking away Steam Survey market share from Intel. Team Red’s 48.37% market share across all platforms is the highest we’ve seen so far this year, with Team Blue now at just 51.63%. The May 2026 Steam Survey showed AMD at a little under 45% — at this pace, it’s possible that it will overtake Intel in just a matter of months.

If we break it down by operating system, 52.59% of Windows users are on Intel, while 47.41% use AMD. Intel may have an advantage here compared to the overall result, but it’s just a tiny margin, and the trend shows that AMD is gaining ground month after month. The difference between Intel and AMD is more glaring when it comes to Linux users, with an overwhelming 66.18% preferring Team Red over Team Blue’s 33.71%.

The Steam Survey is far from scientific and may not even reflect the broader movement of the overall PC market. But with Steam reporting a peak of more than 38 million concurrent players at the time of writing, the survey still offers a useful snapshot of what’s happening in the gaming PC market.

“The way we approached that design was really in terms of de-risking and being able to do that design fast. Vera, as a standalone, is a little bit behind on that maturity level. The Turing device is strong. It did what we needed to do, and partly our partners had some experience with it,” Ho told Tom’s Hardware Premium. “For the Jalapeño program, we were trying to make very pragmatic decisions. We wanted to be aggressive on the goals of the performance and the cost, but we didn’t want to take unnecessary risks. That felt like a good design decision that would fit within the parameters of how we make these design decisions."

There are many Arm-based CPUs on the market, many of which are deployed internally at different hyperscalers, such as Google Cloud’s Axiom and AWS’ Graviton, but Vera and Arm’s own AGI have been described as agentic CPUs, purportedly accelerating the complex reasoning involved in agentic loops compared to their x86 counterparts from AMD and Intel. We asked Ho about this dynamic, and why Turin was the right choice given the close working relationship between OpenAI and Nvidia, who responded with the quote above.

Arm has long touted the virtues of the AArch64 ISA compared to x86, going as far as to claim that its own AGI CPU provides more than twice the performance of modern x86 platforms. That claim is based on internal estimates, not real benchmarks, however. The company has been naturally bullish on AGI’s adoption in the market, though even with an impressive $2 billion in commitments, analysts say the market penetration will still be in the low single digits after two years.

With the explosion of server CPU demand this year and evolving agentic workloads, Nvidia has been at the forefront of messaging, comparing its Arm-based Vera to the x86 competition, boldly claiming a 1.8x improvement over a competing AMD Turin chip; a figure that’s extracted from an overall benchmark suite showing Vera just 3% ahead of AMD’s EPYC 9755.

Many semi-custom Arm designs, from Graviton to Azure’s Cobalt, have focused on core density and efficiency for cloud workloads. Vera and AGI are a shift toward messaging against peak performance. In early Vera benchmarks, Nvidia’s CPU looks impressive, though we’ve yet to see the chip in action in a wide variety of workloads, much less benchmarked against AMD’s upcoming Venice CPUs or Intel’s Diamond Rapids.

What’s interesting about OpenAI’s decision to use Turin is that it’s one of Arm’s deployment partners for AGI, as well as among the list of customers “exploring the Vera CPU,” according to Nvidia. OpenAI, other frontier labs, and hyperscalers keep a wide variety of hardware in their fleet. Pairing Jalapeño specifically with Turin came down to reducing risk, with Ho pointing to the maturity of the platform.

Broadly, x86_64 is more mature than AArch64; ARM has been around since the mid-1980’s, though a proper 64-bit extension didn’t arrive until 12 years after it was introduced on x86. And although Arm has worked its way into data centers over the past two decades, there’s still a much deeper x86 foundation reaching back several decades.

It doesn’t seem Ho’s comments were specifically on the delineation between AArch64 and x86, however. Rather, the comments are focused on the dynamic between Turin and Vera, with the latter being Nvidia’s first foray into a custom CPU core for the data center. Although the early benchmarks of Vera are impressive, it nonetheless represented an unnecessary risk for OpenAI’s rack-scale Jalapeño deployment.

Ho’s comments about OpenAI’s partners having experience with Turin expose where the thinking for the hardware team was at, as well. There are some differences in the practicality of servicing the host system between Turin and Vera, most notably the fact that Vera is board-mounted while Turin chips are socketed; swapping chips isn’t common in a server regardless, but it’s easier with Turin.

Given that OpenAI has a fleet of hardware available, it’s possible Jalapeño will be deployed with a different host system in the future, potentially paired with Arm’s AGI or Nvidia’s Vera. For now, however, Turin was the right choice.

OpenAI bills the autonomous agents as having “their own cloud computer, their own browser, and the apps you’ve connected” in its introduction. This is not a new concept: Cursor says its cloud agents get their own virtual machines; SpaceXAI says every Grok Bot shares one persistent cloud computer; and Meta’s Muse gives each agent a sandbox.

Maddie D. Reese posted on X on launch day the apps her dot said came preloaded. The lineup includes Chromium, Blender, GIMP, Inkscape, Kdenlive, Godot, FreeCAD, OpenSCAD, KiCad, QGIS, ParaView, 3D Slicer, with Python, Node.js, and Git. In the same reply, the dot said its cloud computer runs “Debian Linux.”

The post went up the evening of launch day, captioned “OpenAI dots Geekbench 7 Score,” with a screenshot showing a single-core score of 1,667 and multi-core of 9,435. The result in the screenshot shows that it was uploaded Sept. 25, or four days before the launch. The 9,435 is almost 5% higher than the post-launch best and about 10% higher than the median. The pre-launch run is potentially a bit of an outlier, but performance remains within the same ballpark. It is uncertain who ran the original benchmark.

Tom’s Hardware previously reported that rival Meta’s Muse runs on AMD EPYC Turin with two cores and 8GB per sandbox. The 10 Muse runs in Geekbench’s database have median scores of about 1,041 single-core and 1,394 multi-core. The dot’s post-launch runs, with median scores of 1,570 and about 8,550, put it at about 1.5x the single-core and 6x the multi-core score.

Some of this gap is accounted for by the nine-core vs. two-core configuration; the rest of the gap likely comes from a difference in clock speed, partially offset by Muse’s newer architecture. Geekbench lists a 2.6GHz base clock for the dot’s EPYC 9V74 against 1.5GHz for Muse’s EPYC 9D25.

Muse’s runs also report 7.75GB of memory. Our independent run for this article came back with a single-core score of 1,000 and a multi-core score of 1,433, in line with the other Muse runs.

Data center CPU demand for always-on agents has surged as high-core-count CPUs are needed at sustained load. It is unknown whether the dots, as tested, hold their cores while idle. Such agents are rising in popularity: Muse reportedly passed 500,000 daily active users, and SpaceXAI plans standalone Nvidia Vera CPUs for Grok’s agentic workloads. It may only be a matter of time before a new CPU arms race begins. OpenAI has already said that, in the future, users will be able to add more dots and “scale the output of each dot by ... increasing its speed.”

Traditionally, CPU developers start development of a new processor project with a particular ISA — such as Arm, RISC-V, or x86 — already selected. Although modern high-performance processors translate instructions into internal micro-operations and therefore separate the ISA from much of the underlying execution pipelines, the instruction set still influences numerous architectural decisions, including instruction decoding, register handling, memory ordering, and, of course, software compatibility.

Nuvacore says its Core First approach reverses at least part of this process. Instead of designing WarpCore around the requirements and limitations of a predetermined ISA, the company's engineers are developing a significant portion of the foundational CPU IP independent of a particular ISA. To a large degree, one can compare such an approach to the development of a car around its engine and not trying to fit an engine into an already developed platform.

"Rather than beginning with the constraints of an existing architecture and iterating from there, we are starting with the core itself," the company said in its statement.

As a result, the team may be able to better focus on the core's microarchitecture and optimize it for performance, power efficiency, and sustained workloads typical of modern data centers and AI infrastructure, the target application for the company. Nuvacore intends to select an instruction set later based on what works best for the systems that it and its partners plan to build.

There are obvious limits to how ISA-independent a CPU design can be. Eventually, WarpCore will need front-end logic designed around its chosen instruction set, and those architectural requirements will inevitably affect other parts of the processor. Nevertheless, substantial pieces of a modern CPU — such as execution units, branch-prediction units, data paths, caches, and portions of the memory subsystem — can be developed and evaluated before every ISA-related implementation decision is finalized.

In addition to not disclosing its ISA, which is arguably the biggest unanswered question surrounding the project, Nuvacore also has not disclosed the project's schedule, configuration, manufacturing process, power, or performance targets, though given the lack of the ISA commitment, one could say that the remaining details are not that important.

For now, the only thing that Nuvacore discloses about WarpCore is that it is "a new class of general-purpose CPU designed specifically for the sustained performance and power-efficiency requirements of AI infrastructure and contemporary data centers."

AMD's ambidextrous strategy returns?

Nuvacore's approach is not entirely without precedent. Under CEOs Rory Read and Lisa Su, AMD spent part of the 2010s pursuing an "ambidextrous" strategy intended to let it address x86 and Arm customers by sharing substantial processor and platform IP.

Arguably the most detailed part of the initiative was AMD's SkyBridge platform, announced in 2014. Under SkyBridge, AMD envisioned pin-compatible 64-bit Arm and x86 systems-on-chips that were to use AMD's own Puma+ x86 cores and Arm's Cortex-A57 cores with common platform infrastructure. Separately, AMD developed its custom 64-bit Arm K12 core under Jim Keller that was supposed to share microarchitecture features with Zen. Ultimately, due to lack of resources, SkyBridge was ultimately scrapped, while K12 never became a commercial product as AMD only pursued x86 Zen cores. The difference is, of course, that Nuvacore appears to be taking the idea further by deliberately postponing the ISA decision for WarpCore.

Since Nuvacore is still just a startup, we can only wonder whether its business plan includes marketing of CPUs at all. Nuvia, which was also founded by Williams and Bruno along with Manu Gulati, was acquired by Qualcomm and ultimately shaped the company's Snapdragon X Elite SoCs. In that light, its approach could give the company considerable commercial flexibility.

By developing much of its WarpCore design without tying it to a particular ISA, the company could potentially adapt the technology to the requirements of a future customer or partner. An Arm licensee could use the underlying core technology for an Arm implementation; another customer could choose RISC-V or another ISA. An established x86 vendor could potentially acquire or license the underlying microarchitecture and create its own x86 CPUs if it needs to.

Nuvacore, of course, has not indicated that such a transaction is its objective, but keeping WarpCore ISA-agnostic for as long as possible potentially broadens the number of companies to which its CPU technology could be valuable.

All in all, the Core First announcement tells us considerably more about how Nuvacore intends to develop WarpCore than what the processor itself will eventually look like. But if the company really does keep its ISA options open until relatively late in development, WarpCore represents an unusual attempt to separate development of a high-performance CPU microarchitecture from one of the decisions that normally defines a processor project from its earliest stages.

Su says building compute platforms for the next generation of AI requires “a deep understanding of how models are evolving,” and Li says advancing the technology “requires close collaboration across model research, systems and compute,” and wrote in a separate post that World Labs needs to get “closer to the hardware,” without which AI is “hobbled in efficiency.”

Before founding World Labs, Li had already established herself as a luminary in AI research and development. She led the vital ImageNet project in the 2010s and has served as the chief scientist of AI and machine learning at Google Cloud and the director of the Stanford Artificial Intelligence Laboratory. Thanks to her many contributions to the field, she has been dubbed the "Godmother of AI" by several institutions.

World Labs develops “spatial-intelligence models that generate, reconstruct and simulate interactive 3D environments,” which are useful in robotics and simulation applications. Its world models build or simulate 3D scenes rather than being text-focused like LLMs. The company’s first commercial product was Marble, launched late last year, which creates persistent, downloadable 3D worlds from source material.

AMD and World Labs had already been working together to optimize training and inference for its models, and AMD had invested in World Labs’ $1 billion funding round in February. During Su’s keynote at CES in January, Li said that part of World Labs’ model ran on AMD's Instinct accelerators, and her team used both Instinct hardware and the ROCm software stack to achieve performance improvements for their applications. This tuning experience positioned World Labs as one of AMD’s showcase customers, and AMD says the deal will “help shape its future technology roadmaps.”

World Labs is AMD’s second-largest deal after FPGA designer Xilinx, which came out to about $50 billion at the close in 2022. AMD's other big buys have been smaller: ATI for $5.4 billion in 2006, AI systems designer ZT Systems for $4.9 billion in 2025, and networking solutions developer Pensando for $1.9 billion in 2022.

AI titan Nvidia was also an investor in World Labs in the round earlier this year. However, Nvidia has already been developing Cosmos, its own family of open-source world models. It agreed to buy Hugging Face, a platform for hosting and distributing open models, earlier this month for $12.93 billion.

World Labs' team will continue to do model research. The companies say that the deal will strengthen an “open AI ecosystem,” per the release, and World Labs said in its own post that it remains committed to “widely accessible open models.”

Atlas, World Labs’ latest world model, will power future versions of its Marble product, the company says. Li’s co-founders Justin Johnson and Ben Mildenhall will continue to work with Li in leading World Labs, including after Li takes on the new role at AMD after the expected close of the transaction by the end of the year.

A GitHub project called DLSS-NR-on-AMD, designed to run DLSS 5 Neural Rendering on AMD's Radeon RX 9000 and RX 7000 series GPUs, is already seeing a rapid improvement in performance. It uses a ground-up, full reimplementation of the neural network runtime for AMD GPU architectures. As per the developer, the project currently works with DirectX 12 games that use FSR, with Cyberpunk 2077 and GTA V Enhanced among the tested titles.

Testing done by YouTube channel Ancient Gameplays shows Cyberpunk 2077 running at over 80 FPS on a Radeon RX 9070 XT without DLSS 5 Neural Rendering. Enabling the unofficial implementation, however, caused performance to drop around 11-12 FPS. Not only is that unplayable, but it also highlights just how much optimization was still needed to make the technology practical on AMD hardware.

Development has been rather quick, as an Alpha 0.3.3 version was released with a claimed 23% performance improvement, followed shortly by Alpha 0.4.0, which reportedly added another 42% improvement over version 0.3.3. The changes essentially amount to roughly a 74% performance improvement in just 24 hours. The project has also become easier to install thanks to the AMDNR launcher that can automate much of the setup process, meaning that users no longer need to manually copy and replace DLL files for each game. The launcher also supports different runtimes and provides additional options through the OptiScaler interface.

Ancient Gameplays tested the latest Alpha 0.4.0 build in Cyberpunk 2077 at 3440 x 1440 resolution and found the Radeon GPU reaching around 50 FPS. That is a substantial improvement over the roughly 30 FPS achieved with the earlier build, although it remains well below the performance of the game without Neural Rendering enabled.

There is still a long way to go before DLSS 5 on AMD hardware can be considered a practical alternative to native implementation. The project remains in alpha, which means performance and compatibility can vary considerably depending on the game and GPU. That said, the developer has informed Ancient Gameplays that they are working on another major update, which should bring a substantial performance boost on Radeon GPUs.

Both Kim and Hoffman asked Muse about the VM's specs, and in both instances, Muse revealed that it's running on AMD EPYC 9D25 CPUs, a high-density Turin chip with up to 128 cores (two of which are generally fused off or reserved). The VMs are running on Ubuntu 24.04 and using Linux kernel 7.0. The systems hosting Muse don't include GPUs. The AI agent revealed that Meta uses separate GPU servers for inference, isolating the agent to CPU-only sandboxes.

The agent suggests that each user gets their own private sandbox that's persistent, which allows us to do some math on how many people an individual tray can host. Assuming a 2P system that offers up to 510 vCPUs with 2TB of memory, hosting up to 254 Muse users. Muse has reportedly passed over 500,000 daily active users as of a few days ago, which would come out to somewhere around 2,000 server trays with dual EPYC 9D25 CPUs and 2TB of memory.

This is just some rough napkin math; don't take it as law. It's possible Meta has CPU-only servers deployed with multiple different chips to host Muse, and it's also possible there's overhead in the configuration. Turin chips support up to 6TB of memory with high-density DIMMs, for instance. Still, EPYC hosts seem popular for this use case, mainly because of their core density, as even a dual-core sandbox can add up quickly when multiplied across hundreds of thousands (or even millions) of users.

Muse isn't completely open. Hoffman shared an example where an attempted command failed due to improper permissions when Muse tried to query the kernel buffer. Presumably, sudo (admin) commands would be blocked as well.

However, there might still be some security loopholes. Hoffman says that Muse offered to set up SSH into the private Muse VM. With a reverse SSH tunnel — where the destination machine initiates the connection, bypassing the firewall — an attacker may be able to execute more damaging commands.

Meta's David Singleton says this is intended behavior, however, describing Muse as "a free computer in the cloud." Meta has an extensive white paper on the security architecture of Muse published on its research website.

Dylan Patel, the head of top semiconductor analysis company SemiAnalysis, has discovered that China's Puzhi Electronic Technology is crowdfunding development of a system based on AMD's Zync UltraScale+ XCZU47DR RFSoC, which is a U.S. export-controlled dual-use component. He called on the government to investigate AMD for treason.

"AMD found a business opportunity in dumping U.S. military chips for ¼ of the cost in China," Patel wrote in an X post. "U.S. list price is $36K for this chip, with $4K – 5K to U.S. companies at volume, but it is getting quoted $1K in China to crowdfunding campaigns. AMD needs to be investigated for treason. It is not just export violations, but it is literally selling military end use components to an adversary for less than your home country."

In a statement to Tom's Hardware, AMD denied selling or shipping the chips directly into the restricted region. The company says it investigates suspected diversion cases to determine whether any violation of law occurred, although it did not explicitly confirm that it had opened an investigation into this particular case.

"AMD is committed to full compliance with all U.S. and global export regulations and has a clear policy requiring our distributors, resellers, and customers to do the same," a spokesperson told Tom's Hardware. "Any shipment into restricted regions is a direct violation of our policy. This recent instance is unrelated to any direct AMD sales or shipments. We investigate all suspected diversion cases and if we find any violation of law, we will report it to relevant authorities and cease all business with the company."

AMD told Patel the same, according to another X post he made as a follow-up on the matter.

AMD's Zynq UltraScale+ XCZU47DR RFSoCs pack four general-purpose Arm cores, two Cortex-R5F real-time cores, an FPGA with 930,000 gates, and DSP engines with 4,272 DSP slices. It also has eight 14-bit analog-to-digital converters (ADC) capable of up to 5 GSPS and eight 14-bit digital-to-analog converters (DAC) operating at up to 9.85 GSPS to receive, process, and generate RF signals in real time without relying on separate data converters.

Devices based on Zynq UltraScale+ RFSoCs are used for software-defined radios, radar and phased-array systems, wireless and satellite communications, electronic test equipment, and other applications that require high-bandwidth, low-latency signal processing. Puzhi Electric officially positions its upcoming systems based on the Zynq UltraScale+ RFSoCs for 'professional engineers' and 'advanced researchers in universities,' though such systems can be used for completely different applications.

AMD's Zynq UltraScale+ XC-series RFSoCs are designed for commercial and industrial applications, but AMD also offers separate XQ-series defense-grade RFSoCs for aerospace and defense applications, although there the XCZU47DR does not have a direct XQZU47DR counterpart. Yet, the closely related ZU48DR is available in both XC and XQ versions. The latter adds defense-oriented qualification, ruggedization, and support rather than fundamentally different processing capabilities.

Therefore, while the XCZU47DR is technically aimed at civil, commercial, and industrial applications, the chip is a dual-use component featuring a 3A001.a.14 ECCN classification under the U.S. Commerce Control List (CCL) and therefore its exporting, re-exporting, or transferring normally requires a BIS export license to non-allied destinations (e.g., China) on anti-terrorist and national security concerns. This means that any shipment to China requires a license, but in practical reality, that license will almost certainly be denied. However, there is a license exception that allows shipping such items to commercial companies from the Group B countries, which means legitimate exports to dozens of states from where such goods could get to China illegally.

This might explain how Puzhi obtained its XCZU47DR devices, but it does not explain the enormous apparent difference between the prices available to Chinese and American buyers. The $1,000 figure, however, is Patel's reported price for the chip rather than a publicly advertised retail offer.

Xilinx, which AMD later acquired, introduced its 3^rd Generation Zynq UltraScale+ RFSoC family, including the XCZU47DR, in February 2019, and the devices reached production status in 2020, meaning that they have been on the market for almost six years now. Xilinx and then AMD have sold the unit to tens, if not hundreds, of companies in different countries, so determining the origin of the units that Puzhi has might be difficult.

It should be noted that lead times for XCZU47DR at various resellers like Avnet and DigiKey are 40 to 52 weeks, which indicates that the components are in high demand in the U.S. and Europe. Meanwhile, Puzhi claims it can ship its system based on the RFSoC on November 6, 2026.

AMD's existing Radeon RX 9000-series graphics processors based on the RDNA 4 architecture use GDDR6 memory, so the addition of GDDR7 identification to drivers is arguably the most explicit confirmation that AMD is setting the stage for its next-generation discrete Radeon graphics processors based on the RDNA 5 architecture.

AMD also submitted patches that enable IH 8.0, a new version of its Interrupt Handler IP block, as well as NBIF 7.10, the latest revision of the company's New Bus Interface. Several smaller patches make additional preparations for the new hardware. These changes follow earlier Linux driver work involving Display Core Next 6 (DCN6) as well as GFX 13.0.x, suggesting that AMD is gradually upstreaming support for multiple components of an upcoming GPU architecture. Yet, while enablement of IH8, NBIF 7.10, DCN6, and GFX 13.0.x clearly point to new graphics hardware, it does not necessarily point to new discrete GPUs, unlike the mention of GDDR7.

As revealed in August, 2025, Laks Pappu, Senior Fellow at AMD, is the lead architect for AMD's next generation datacenter GPU and discrete graphics platforms. Pappu was building next-generation "competitive 2.5D/3.5D chiplet-based and monolithic graphics SoCs on various packaging technologies," according to his LinkedIn profile before it was edited to remove those disclosures.

Essentially, Pappu's profile indicated that AMD's next-generation GPU architecture could support multiple physical implementations — including multi-chiplet and monolithic — depending on performance, cost, market requirements, and AMD's willingness to compete in certain market segments

AMD has already used a multi-chiplet design with its Navi 31 GPU, an implementation that kept graphics processing hardware on a large Graphics Compute Die (GCD), but disaggregated memory interfaces and caches into smaller Memory Cache Dies (MCDs). A more ambitious implementation could potentially distribute graphics processing resources between multiple dies, although Pappu's profile did not disclose how AMD intended to partition its future GPUs.

Such an approach would be considerably more complicated than separating memory and cache functionality. Multiple compute dies would require high-bandwidth, low-latency connections as well as mechanisms for synchronization and coherency, while software would ideally continue to see the set of chiplets as a single graphics processor.

Based on conventional GPU development cycles of roughly 2.5 to 3.5 years, RDNA 5 could already have been approaching tape-out or early post-tape-out stages when the information surfaced in August 2025, which means that the company is already likely to be testing the new GPUs internally.

Officially, AMD hasn't disclosed any details about its RDNA 5 architecture nor launch schedule, so take these developments with a grain of salt. For now, all we know for sure is that AMD is working on standalone RDNA 5-based GPUs, and they're likely to use GDDR7 in at least some configurations, assuming the AI-driven memory crunch doesn't choke supply of those chips even further for consumer applications.

Saša Marinković, AMD's senior marketing director, shared the milestone on X. Since posting, AMD's share has fluctuated a bit, bringing the market cap down below $1 trillion for brief periods. Undoubtedly, the buzz around the milestone will cause some turmoil in the share price throughout the day, though it should eventually stabilize.

Just a week ago, AMD was quite a ways off hitting the trillion-dollar milestone. AMD's share price has risen by around 29% over the past month, and 21% of that jump came in just the last five days. On Monday alone, AMD's shares jumped around 9% within the first few hours of the market opening.

If you're unfamiliar, market cap is a shorthand for understanding the value of a company. It looks at the number of outstanding shares multiplied by the share price. It's not necessarily tied to revenue or profit, at least not directly. Rather, market cap is a read on the value of a company in the eyes of Wall Street, fluctuating as the share price does.

AMD joins a club of fewer than two dozen companies around the world that have passed a $1 trillion market cap, including SK hynix, Amazon, Microsoft, Apple, Nvidia, Broadcom, and Micron. Although AMD has now joined the club, it's still far behind its rival in the AI accelerator space, Nvidia, which sits at a market cap of around $5.4 trillion currently.

However, AMD was able to beat Intel to the milestone. Intel's market cap currently sits at around $648 billion. Interestingly, Intel has actually seen a larger rally in its share price over the past month. Share prices have risen 37% in the past month, with 25% of that growth coming in the past five days. Similar to AMD, Intel opened on the market Monday morning with a 13% jump.

Following AMD's Q2 earnings, where it beat revenue expectations by 2%, shares traded down, creating an opportunity to hold. That's paying off now as both AMD and Intel shares jump.

Although AMD's AI accelerators have made inroads into the data center, Nvidia still largely dominates that market. AMD's unique positioning comes from its CPUs, which have become a focal point of agentic AI data centers over the past year.

• Get the AMD Ryzen 5 5500F from Amazon
• Get the AMD Ryzen 5 7500 from Amazon

AMD revealed the budget CPUs nearly two weeks ago, and despite launching on that date, the chips haven't been available for sale in the U.S. until now. The Ryzen 5 5500F is particularly interesting, as the $90 Ryzen 5 5500 has continually been among Amazon's best sellers in CPUs. Although the Ryzen 5 5500 and 5500F sound similar, there are actually quite a few differences between the two CPUs.

Both are six-core, 12-thread chips using AMD's Zen 3 architecture, but the 5500 falls under the Cezanne family, while the 5500F falls under Vermeer. The 5500F comes with PCIe 4, compared to PCIe 3 on the 5500, as well as a higher 4.4 GHz boost clock — the base 5500 tops out at 4.2 GHz. These changes apparently allow the 5500F to achieve higher performance, somewhere in the range of 5% to 10%, in games. We'll be getting the CPU in the Tom's Hardware lab to test the performance ourselves.

Although Zen 3 is aging, it has become an ideal home for budget builders as the RAM pricing crisis continues to surge. In addition to the Ryzen 5 5500F, we saw AMD re-release the Ryzen 7 5800X3D earlier this year to combat rising DDR5 prices.

We took the Ryzen 5 5500 (non-F) out for a spin earlier this year for a budget $100 CPU shootout. Generally, it underperforms the Intel competition at this price, but a 5% to 10% jump would close that gap. If performance holds up, the Ryzen 5 5500F may have a shot at our best CPUs for gaming list with its low price.

The Ryzen 5 7500 is easier to parse. It's identical to the 7500F, just with integrated graphics. It comes with six Zen 4 cores for a total of 12 threads, and unlike the 5500, it exclusively supports DDR5 memory. The 7500 comes with a boost clock of 5 GHz and supports PCIe 5. Like the 5500F, it has a 65W TDP and comes bundled with AMD's Wraith Stealth cooler.

AMD originally announced the Ryzen 5 5500F at $99 and the 7500 at $190, though the current Amazon listings are both $20 higher than that — $120 and $210, respectively. At the time of writing, the CPUs are only available at Amazon in the U.S.; Micro Center and Newegg don't have listings available. When other retailers pick up the chips, prices could come down.

This isn't the first time we've seen oddly high prices on new AMD releases at Amazon. Earlier this year, the Ryzen 9 9950X3D2 went up for sale on Amazon prior to release, selling for $1,000, $100 above MSRP. After the launch dust settled and listings went up at other retailers, the pricing dropped back down to $900. Hopefully, we'll see something similar happen here.

"Since we do not have enough capacity to even satisfy the ​demand in China, we do not have a plan to expand into the international market in a fully-fledged way," said Eric Xu, rotating chairman of Huawei, on the sidelines of the company's Huawei Connect conference, Reuters reports. He added that Huawei supplies limited volumes to 'some countries where demand is particularly strong,' though he did not elaborate.

Huawei this week unveiled its latest AI accelerator roadmap, revealing major training and inference performance gains for its next-generation Ascend 960, 970, and 980 NPUs over the existing Ascend 910C and Ascend 950-series. The Ascend 960DT and 960PR are set to increase their FP8 training performance to 2 PFLOPS and their FP4 inference performance to 4 PFLOPS and 8 PFLOPS, respectively, in 2027. Meanwhile, their successors, Ascend 970 and Ascend 980, are projected to increase their FP4 performance to 14 PFLOPS and 28 PFLOPS, respectively, in the coming years.

Huawei Ascend vs Nvidia AI GPUs

*Preliminary data
S/D - Sparse and Dense
T/I - Training and Inference

But while the upcoming Ascend NPUs will be considerably faster than their predecessors, particularly for inference, they will remain well behind Nvidia's previous- and current-generation accelerators, at least in raw compute performance. Huawei's 2027 Ascend 960DT is projected to deliver 2 FP8 TFLOPS for training, compared with Nvidia's 4 FP8 TFLOPS for the H200, released in 2023. The Ascend 960PR is expected to offer 8 FP4 PFLOPS for training, which is far behind Nvidia's B300, which delivers 15–20 NVFP4 PFLOPS. Even the Ascend 980, targeted for 2029, is projected to reach 7.2 FP8 PFLOPS and 28 FP4 PFLOPS, well below Nvidia's R200, which is on track to deliver 17.5 FP8 PFLOPS and 35/50 FP4 PFLOPS this year.

Such a massive performance difference with leading AI hardware will reinforce Huawei's reliance on massive system-level scaling rather than chip-for-chip performance to compete with Nvidia. But massive system-level scaling comes with massive power consumption, which will make Huawei's next-generation Atlas SuperPoDs and SuperClusters considerably less competitive in markets that can access hardware from AMD or Nvidia.

Huawei is in an interesting paradoxical situation. On the one hand, its integration efforts like near-package optics (NPO) clearly free up capacity on 'older' nodes that can be used for other components of AI platforms. But on the other hand, SMIC's inability to ramp production on 7nm and 6nm-class nodes limits Huawei's ability to supply its AI hardware anyway, which is why it can barely meet demand.

Then again, while Huawei's Atlas SuperPoDs with up to 15,488 Ascend 960 NPUs can deliver up to 30 FP8 EFLOPS and 120 FP4 EFLOPS performance by far exceeding the capabilities of Nvidia's NVL72 clusters with a 72-GPU scale-up world size, their performance-per-watt is poised to be dramatically lower compared to Nvidia's architectures, which means that demand for such hardware outside of China will be limited at best. That said, a global AI hardware push doesn't make much sense for Huawei right now. What perhaps does make sense is offering cloud access to its hardware to various academic and research customers to popularize its CANN software stack.

The crafty PC builder behind this new battlestation is Old-Mate-Fetus. Key to its visual appeal is the extensive use of copper cooling apparatus and wood. The images clearly show this open-build PC made with a custom liquid loop using copper pipes, elbow joints, and welds. These pipes meander around the build, feeding the Alphacool GPU block, the Heatkiller IV CPU block, and circulating around a large radiator up top. A copper design feature that really stands out, though, is the “reservoir made from an old Primus Blowtorch with sight gauge fitted to it.” Nice.

Old-Mate-Fetus sits the PC on a base/backing frame they also handmade using a dark hardwood with an antique finish, and brass screws. We can’t see it in the photos, so it's probably pretty safe to guess that the Corsair 850W SFX power supply is hidden in the wood base. The builder says that they also made custom cables from the PSU to the various components.

Custom PC component list:

• AMD Ryzen 7 9800X3D
• Gigabyte Aorus Elite B650M Wi-Fi
• Gigabyte Radeon RX 9070 XT Gaming OC
• 32GB Klevv DDR5-6000 CL28 (wanted 64GB)
• AlphaCool GPU Block
• Heatkiller IV CPU Block
• AlphaCool DDC pump with custom enclosure
• Corsair 850W SFX power supply
• Noctua Industrial 120mm Fans

If you are thinking the Old-Mate-Fetus design has some Billet Labs influence, and that it would be appreciated by them, “you are correct,” confirms the London-based copper pipe welder in the comments. The builder of this all-AMD machine warmly accepted the compliment. Redditors also joked about this new gaming PC possibly being the ultimate Steam Machine.

There are several configuration differences depending on the benchmark throughout AMD's white paper, and although we'll call out those differences here to the best of our ability, we don't have all of the details. For the Vera comparison, in particular, AMD is mixing data from different sources, and in some cases, using different major releases of the GNU Compiler Collection (GCC). That can have a substantial impact on performance, so keep your salt shaker handy.

First up are results in SPEC CPU 2026 with the intrate test, looking at total throughput. These are older numbers, gathered in July with GCC 15.2. The intrate test runs multiple copies of an application on the same CPU, and the SOP is to run one copy per thread. Presumably, that's what AMD did here, but the white paper doesn't clarify, even in the footnotes.

The 256-core 9996 is 2.37x faster than the Intel Xeon 6980P and 2.24x faster than Vera according to the slide. The white paper clarifies the mystery 9006 CPU is the 256-core flagship. Perhaps most impressive is AMD's gen-on-gen comparison. According to these results, the 9996 is around 78% faster than last-gen's 192-core EPYC 9965.

Although the high-level results bring in data from Intel and AWS, much of the white paper focused squarely on the comparison between Venice and Vera. AMD broke down the individual subtests of SPEC CPU 2026 intrate in the white paper, which you can see below.

The comparison looks good for AMD, naturally, though there are a few wrinkles in the configuration. AMD is testing a down-cored EPYC 9996, dropping from 256 cores to 96 cores. It made no mention of power budget, but when AMD originally shared SPEC numbers, the 96-core model had access to the same 600W as the 256-core model — AMD's 96-core, high-frequency Venice SKU tops out at 500W. More consequential is the compiler, however. AMD is using GCC 16.1 and comparing the results to the ones Nvidia shared in its Vera white paper. Nvidia used GCC 15.2.

Michael Larabel over at Phoronix has a nice write-up about the difference between GCC 15 and 16, but the short story is that there are performance differences, not always for the better. GCC 16 takes longer to compile due to better optimizations, hence the lower scores on the GCC and LLVM compilations above. However, that leads to faster binaries. By how much depends on the flags, software, and a whole host of other factors. Regardless, it's not best practice to compare benchmarks using two different compiler versions. It makes sense that AMD used GCC 16.1 — it includes support for Zen 6 — but ideally Vera would also be on GCC 16.1.

Speaking of Phoronix, AMD pulled some data for the publication's initial, controlled testing of Vera. Above, you can see the Stream, an industry-standard benchmark for measuring memory bandwidth. Again, AMD is using a down-cored 9996 from 256 cores to 96, and offering it a 600W power budget. Still, this is an impressive showing, as Vera absolutely clobbered the competition in the publication’s original Stream results. Here, AMD is ahead by about 18%, with per-core performance about 8% ahead.

Breaking out of Vera, AMD also showed performance in cloud workloads, including database, Java, and cryptography. Once again, the gen-on-gen comparison stands out, as AMD was already leading in these workloads with its last-gen chips. AMD ran these tests itself, rather than relying on third-party data, though the Graviton5 results came from an AWS cloud instance.

Similarly, in HPC workloads, AMD furthers its lead over Intel's flagship Granite Rapids-AP offering. Intel's next-gen data center CPUs, codenamed Diamond Rapids, are set to be released next year.

Finally, we have "agentic AI workload performance," which uses actual benchmarks for comparison, despite what the names in the chart above suggest. From left to right, AMD used NGINX, TPCx-AI kit, FAISS, TPC-H and TPC-C, and a replay of a multi-persona agent. For TPC-H and TPC-C, AMD says it derived workloads from those benchmarks, so the results here aren't comparable to published results.

Although looking at benchmark results is always interesting, it doesn't say much in the context of a server deployment, at least at the scale that AMD is targeting. Peak performance is only one of the major factors that go into server deployments, after all, and even then, performance can vary wildly depending on what software you're running and how it's built.

Still, Venice looks impressive, perhaps more so in the gen-on-gen comparison than any competitive comparison. Hopefully that bodes well for AMD's future Zen 6 rollout on consumer desktops, but we'll have to wait until Team Red has more to share before drawing any conclusions on that front.

The 25H2 update finally reintroduces a movable taskbar, just like we've had since Windows 95, with an option to make the height smaller for compact displays. Likewise, the Start menu now has multiple configurable layouts, and you can toggle each main section on or off. Microsoft also revamped Windows Search — perhaps to finally be useful — and users can now choose to only have it display local results. Explorer got a handy tweak in the form of using KB, MB, GB size indicators, as well.

The update has also surfaced some serious bugs. First off, PCs with AMD Radeon GPUs are crashing, apparently due to the update provoking driver instability. Reinstalling the driver doesn't appear to help, and the reports unfortunately cover most contemporary AMD graphics cards. Microsoft did not mention AMD GPUs specifically in its latest update.

Microsoft noted that it has fixed an issue with Remote Desktop Services where RDS might become unstable, causing connection and sign-in failures.

Microsoft is also aware of an issue with audio output and microphone input going dead. Additionally, there are seemingly reports of audio devices starting up with a botched configuration, like tweaked 3D audio settings or speaker setups. The apparent workaround seems to be putting audio in a standard 2.0 configuration, but Microsoft has now addressed this issue in the emergency patch.

Explorer may display a black screen after logging in, leaving you with a perfectly functioning operating system but no visible UI. The File History backup feature is reportedly not detecting external drives, effectively making it useless. Of particular concern to enterprises, systems administrators, and homelabs, the Remote Desktop Service is also experiencing issues. Last but not least, Microsoft says it has addressed an issue where Claude Cowork is nonfunctional due to the update breaking HC-managed Linux VMs with Plan9 shared folders

Owners of HP and Lenovo systems are also advised to consider holding off on this update. Ever since the August 2026 patch (and now, cumulatively, with the September update), there are reportedly serious issues on both camps. On the HP side, there are instances of the update triggering BIOS corruption, leading to a failed install with no ability to recover the OS — though seemingly the OEM has published a new BIOS revision, so it's hard to say which side is to blame. Meanwhile, Lenovo users might have their machines black-screen or shut down out of the blue, seemingly due to power mismanagement.

While it's heartening to see that Microsoft has significantly increased its Windows development speed and is seemingly listening to customer feedback, a string of bugs after a major feature update are still common.

According to testing shared by the developer, FSR 4.1.1 running at 2560 x 1440 with a 1706 x 960 Quality input required 11.51 ms using the original shaders, compared with 5.92 ms using the latest v4.0.0-rc9 release. At 4K, processing time reportedly dropped from 25.72 ms to 12.08 ms, while 1080p fell from 7.13 ms to 3.93 ms. The developer also reports that the optimized implementation produced byte-identical images to the original shaders in controlled testing, suggesting that the performance improvement comes from executing the same workload more efficiently, and not reducing image quality.

The BC-250 is particularly interesting in this context because it is based on much older RDNA 2 graphics hardware. FSR 4 relies heavily on machine-learning workloads and accelerated low-precision integer operations that are far better suited to newer AMD architectures. On the BC-250, those operations have to be handled through alternative instruction paths.

Earlier community work tackled the problem in Mesa, optimizing the INT8 dot-product operations used by FSR 4. Instead of relying on a much more expensive fallback sequence, the modified implementation used a more efficient mathematical path based around signed 24-bit integer operations. One representative shader reportedly fell from 64,269 instructions to 37,613.

The latest fork takes that work a step further by shifting the optimizations into a FidelityFX DLL. This removes the requirement to run a specially modified Mesa build, potentially making the optimized path easier to use with an existing OptiScaler installation or games with compatible native FidelityFX implementations.

Native Windows support, other GPUs, and frame generation have not yet been qualified, as the developer tested the BC-250 on Linux using standard Mesa and GE-Proton. The practical upper target for the board is roughly 1440p, as FSR 4 still consumes 12.08 ms at 4K even after the optimization.

The achievement is the latest chapter in the BC-250's unusual story. ASRock built the board for cryptocurrency mining systems around PlayStation 5-class AMD silicon, with six Zen 2 CPU cores and 24 RDNA 2 compute units enabled by default. With the mining market's collapse, individual boards began filtering onto the second-hand market for well under $150, and enthusiasts have since assembled a mature Linux gaming ecosystem around them, complete with custom firmware, GPU governors, CU unlock tools, cases, cooling solutions, and distro-specific fixes.

We recently put the BC-250 through its paces and found that community-developed tools can unlock all 40 physical RDNA 2 compute units on compatible boards, up from 24 enabled by default, while separate firmware work can restore the two disabled Zen 2 CPU cores and take the chip from six cores to eight.

The ecosystem still comes with limitations. In our testing, the BC-250’s aging Zen 2 CPU became a substantial bottleneck at 1080p and 1440p in some games, even with the GPU fully unlocked. The board also has only 16GB of shared GDDR6 memory, which caused some titles to crash due to memory issues at 4K and made frame generation harder at higher resolutions. While FSR 4 will not erase those limitations, halving the upscaler's per-frame cost gives a once-discarded mining board significantly more headroom to spend elsewhere.

Fastest smartphone SoC

Apple's A20 Pro system-on-chip (SoC) delivers 4,006 points in single-thread and 11,460 points in the multi-thread Geekbench 7 benchmark, which represents a 23.3% higher ST performance and 27.1% higher MT performance compared to the immediate predecessor, the Apple A19 Pro, according to an early submission (which may or may not demonstrate performance of actual A20 Pro-based products, so take the results with a grain of salt).

The new SoC delivers the highest generation-over-generation performance improvement for Apple's smartphone processors in years and is currently the highest-performing mobile AP. Furthermore, the A20 Pro beats AMD’s 16-core Ryzen 9 9950X3D by 26% and Intel’s Core i9-14900KS by 32% in single-thread performance.

When it comes to the single-thread Geekbench 7 benchmark, Apple's A20 Pro outperforms its closest rivals — Qualcomm's Snapdragon 8 Elite Gen5 (SM8850) and Xiaomi's XRing O3 — by 31.5% - 33.7%. In fact, both SM8850 and XRing O3 perform on par with Apple's two-years-old A18 Pro. The six-core A20 Pro also beats the eight-core SM8850 by 12.2% in multi-thread workloads in Geekbench 7 and offers roughly similar multi-thread performance to the 10-core XRing O3.

Compared with other flagship smartphone processors, Apple's A20 Pro holds a commanding lead in Geekbench 7. It is 76% faster in single-thread and 48% faster in multi-thread performance than MediaTek's eight-core Dimensity 9400, while it nearly doubles the performance of Google's eight-core Tensor G5, with advantages of 99% and 96%, respectively. But the most striking gap of A20 Pro is with Huawei’s Kirin 9050 Pro: Apple's flagship is 290% faster in single-thread and 139% faster in multi-thread Geekbench 7 workloads.

A great laptop CPU

While Apple's A20 Pro continues to feature 'only' six cores like many generations before it, this time around the processor packs two 'super' desktop-class general-purpose cores running at up to 4.93 GHz, four efficiency cores running at lower clocks, and a memory interface that delivers +50% higher memory bandwidth compared to its predecessor (allegedly using a 96-bit memory I/O).

The architectural enhancements of advanced CPU cores running at nearly 5 GHz, along with a more capable memory subsystem, not only enable a massive generational performance uptick, but also allow the chip to offer unbeatable single-thread performance and massive multi-thread performance that is comparable to that of laptop CPUs, including previous-generation laptop CPUs from Apple.

Indeed, Apple's A20 Pro is 7% faster than M5, 20% faster than M4, and 43% faster than M3 in single-thread performance. Its six-core design cannot match its multi-thread performance, trailing the 10-core M5 by 39% and the 10-core M4 by 27%. Yet, it is only 5% behind the eight-core M3.

When compared to Intel's Panther Lake, the A20 Pro is 48.7% faster in single-thread performance than the flagship Core Ultra X9 388H, yet the 16-core Panther Lake processor is 63% faster in multi-thread workloads. Against lower-end Panther Lake parts, the A20 Pro is 74% – 88% faster in ST workloads and even leads the Core Ultra 5 325 and Ultra 5 332 by 3% and 64%, respectively, in multi-thread benchmarks.

The particularly striking results of Apple's A20 Pro are the 26% – 33% single-thread advantage over flagship AMD and Intel desktop CPUs, though the desktop processors remain dramatically faster in multi-thread workloads.

First 2nm smartphone SoC

When Apple transitioned to TSMC's N3B (3nm-class) process technology from N4 (4nm-class) with its A17 Pro SoC back in 2023, the new processor was barely 9.8% faster in ST and 6.8% faster than its predecessor A16 Bionic. By contrast, with its first 2nm smartphone SoC made on TSMC's N2 node, Apple offers a massive performance boost over the A19 Pro produced on N3P.

Indeed, Apple's A19 Pro packs two 'desktop-class' 'super cores' whose design is 'driven in part by increased front-end bandwidth, a new cache hierarchy, and enhanced branch prediction,' as Apple described its 'super cores' inside the M5 processor earlier this year. Such architectural enhancements obviously massively increase performance in single-thread workloads at the cost of increased die size, transistor count, and power. Apparently, N2 enabled Apple's designers to squeeze two desktop-grade CPU cores into a smartphone SoC.

Speaking of M5, it is noteworthy that A20 Pro delivers 7.1% higher single-thread performance than M5 while running at a clock speed that is 7.1% higher than that of M5, which is probably a good indicator that Apple's A19 Pro uses the same 'super cores' as M5.

While some may consider using PC-grade general-purpose CPU cores in a smartphone chip an overkill, Apple is known for using and supporting PC technologies in its mobile SoCs (NVMe, PCIe, DisplayPort-over-USB-C, hardware virtualization, etc.). Keeping in mind that Apple also uses A-series SoCs inside iPads and inexpensive laptops, it makes a great sense to have these technologies in its smartphone application processors. With desktop-grade cores inside the A20 Pro, the company greatly expands use cases of these CPUs while also solidifying their position in traditional segments that they will address in the coming quarters.

Without any doubts, Apple's transition to TSMC's N2 starts with a massive general-purpose performance increase, driven by 'fat' super cores and a memory subsystem featuring 50% more bandwidth compared to the A19 Pro. Over the next few weeks, we are also going to learn how Apple upgraded the GPU, NPU, and other aspects of the A20 Pro, and we are going to find out whether the upgrades are as impressive or incremental. In any case, so far, the A20 Pro looks very good.

The industry shipped 12.5 million standalone graphics cards for desktop PCs in the second quarter of 2026, up around 5.9% sequentially and 7.8% year-over-year. 12.5 million add-in boards (AIBs) is the highest number of graphics cards sold in one quarter since the first quarter of 2022, when the industry shipped 13.38 million AIBs.

It is particularly noteworthy that 2026 is shaping up to be better for unit sales of desktop graphics boards than 2025 despite raising prices. For the first half of 2026, 24.3 million desktop AIBs were shipped, up significantly from 20.8 million graphics cards supplied in the first half of 2026. JPR analysts also note that only around 14 million desktop PCs were sold during the quarter, which — given an unusually high 89% attach rate — largely means that the majority of AIBs shipped during the quarter were aimed at gamers buying in retail and not at PC makers.

"Defying common wisdom, high-end AIB sales spiked as prices increased," said Jon Peddie, president of JPR. "Our theory is consumers rushed to buy AIBs before the prices went any higher, as the war in Iran is driving prices up in all segments."

Having shipped about 11.25 million discrete GPUs for desktop computers in Q2 2026, Nvidia remained the undisputed leader of the market with around 90% market share. AMD controlled roughly 8% of the market, shipping about one million discrete desktop GPUs, while Intel's share increased to 2% on shipments of several hundred thousand units. Meanwhile, Jon Peddie Research notes that market share changes were negligible during the quarter: AMD’s overall AIB market share decreased by -0.16% from the previous quarter, Intel's market share increased by 0.3%, and Nvidia's market share decreased by -0.1%.

For Nvidia, the quarter was particularly good as it sold the highest quantity of discrete GPUs for desktop PCs in a single quarter since Q3 2017, when it sold approximately 11.72 million units. By contrast, sales of AMD's standalone graphics cards have been floating below or around one million units per quarter for nearly four years now, with only three quarters being exceptions (Q3 2023, Q4 2023, Q4 2024). Still, one million is higher than the around 700 thousand discrete desktop GPUs the company sold in Q2 2025.

In total, Nvidia shipped approximately 17.365 million discrete graphics processors in the second quarter: roughly 11.25 million units went to desktops, and around 6.115 million units were installed into notebook and compact PCs. Since both AMD and Intel have quietly quit the market for standalone GPUs for mobile PCs, their shipments to this market segment were essentially zero.

Biren reported first-half revenue of $183.9 million, up 1,998% year-over-year from around $8.665 million in the first half of 2025. The company's gross profit rose to $78.552 million, and gross margin increased to 42.7%, but it still lost $56.2 million primarily because it continued to invest in new products, including AI accelerators, optically-interconnected rack-scale solutions, and software. Biren's revenues started to climb in the second half of 2025, so for the whole year its sales reached $154.17 million as its market share of AI accelerators in the country was below 3%, according to TrendForce.

For those who follow China's AI and GPU markets, Biren Technology is certainly a familiar name as the company's products are well documented and appear to be competitive with those developed by AMD and Nvidia on paper. The company has developed at least three high-end AI GPUs — the BR106, BR110, and BR166 — and is currently working on BR20X, BR30X, and BR31X accelerators, according to JPR. Biren has also built its own Birensupa software stack meant to compete against Nvidia's CUDA and is working on a rack-scale solution.

In reality, demand for domestic AI accelerators has always been relatively low in China, as even cut-down versions of Nvidia's leading AI GPUs provided better performance and software stack than solutions developed in China. While Nvidia charged $12,000 - $15,000 per H20 AI GPU when it sold these products in the PRC, it still supplied some 2.2 million AI accelerators to the country in the first half of 2025, when it could still ship them until the Trump administration's export controls kicked off in May, according to TrendForce. By contrast, Biren shipped thousands, maybe tens of thousands of AI accelerators throughout the whole 2025. Even today, Biren's shipments are minuscule compared to Nvidia's in 2025.

Without a doubt, Biren's financial improvement is real and impressive, but it is coming from an extremely small base in the first half of 2025, so the 1,998% 1H 2026 growth figure makes Biren sound much larger than it actually is. While Biren is growing at an enormous rate, with $183.9 million in revenue, it is still a relatively small accelerator supplier in absolute terms.

What remains to be seen is whether Biren can secure enough manufacturing capacity from SMIC or other suppliers to compete with larger Chinese AI accelerator vendors, such as Huawei, Kunlunxin, and Cambricon. The company certainly has more financial resources than it did a year ago and faces less formidable competition from AMD and Nvidia amid U.S. export restrictions and China's own bans on American AI hardware. But having competitive designs is only part of the equation: Biren now must manufacture enough accelerators to satisfy customer demand and substantially increase its market share.

The increases come on the back of Intel seeking higher gross margins for its products as the PC market shrinks. This is a story we've heard directly from Intel in the past. In its most recent earnings call in July, Intel chief financial officer David Zinsner attributed a 13% YoY increase in Intel's client revenue to higher average selling price, not a higher volume of sales.

Although the Digitimes report doesn't clarify which products will see a price increase, server and mobile seem like the most likely candidates. Intel's most recent Panther Lake calls for high-speed LPDDR5X-7467 memory as a minimum, and last-gen Lunar Lake CPUs have on-package memory. Naturally, higher memory prices put more pressure on fully built systems like laptops more so than socketed, standalone desktop processors.

On the server end, there's been an unprecedented increase in demand for server CPUs on the back of agentic AI workloads. That demand led to several consecutive records for Intel's share price, even without any major product announcements. Earlier in the year, Wall Street estimated the server CPU market would rise to around $120 billion by 2030 (currently around $30 billion). Now, those projections go up to as high as $220 billion.

According to the report, Intel is set to launch a major new annual product in March 2027, followed by AMD between June and July. Last week, a leaked Intel roadmap showed the company's next-gen Nova Lake desktop CPUs entering mass production in Q4 2026 with a release in Q1 2027, lining up with DigiTimes' report.

Although the timelines line up, the rumor mill has suggested an early Q1 launch for Nova Lake. It's worth noting that the DigiTimes report doesn't make mention of which product Intel will launch in March. This year, for instance, Intel launched its Xeon 600 CPUs for HEDT in March.

Perhaps more interesting is the AMD timeline. We already know of one major AMD product launch in the second half of 2027, which is Venice-X. Those are Zen 6 server CPUs with AMD's 3D V-Cache, packing up to 1,152 MB of L3 cache on the chip. Otherwise, that timeframe seems to point to AMD's next-gen desktop CPUs with the Zen 6 architecture, codenamed Olympic Ridge.

AMD launched its Venice server CPUs earlier this year, the first sporting the Zen 6 architecture. We haven't heard anything official about Zen 6 in the desktop yet. That's strange given AMD's last several releases. There was about a two-year gap between Zen 3 and Zen 4, as well as Zen 4 and Zen 5, on desktop. We've just crossed the two-year mark for Zen 5, so assuming AMD keeps a similar launch cadence, we'd expect to hear something sooner than June or July or next year.

That same explosive demand in server CPUs could have changed AMD's launch plans, however. Given that we haven't heard anything official about Olympic Ridge at this point, a launch in June or July isn't out of the question.

Taking all of the components together, the street price should come out to over $100,000 with just the core components: memory, CPU, and dual GPUs. Configured higher, and with supporting storage, power, and cooling, the workstation could very easily climb over $150,000.

It's essentially a server tray reconfigured into a tower, with an EPYC host replaced with a 96-core Threadripper. AMD didn't share many details about the machine outside of the specs, though it appears to be a system design that AMD's OEM partners will ultimately build and ship. AMD has yet to announce any partners supporting the machine.

The Threadripper Pro 9995WX at the heart of the machine is a 96-core, 192-thread Zen 5 chip that can boost up to 5.4 GHz. It ships with 384 MB of L3 cache and has a TDP of 350W. It's hard to find Threadripper Pro standalone chips in general, but the 9995WX clocks in at around $11,000 to $12,000.

The MI350P accelerators come with 128 CDNA 4 compute units built on TSMC N3. Each accelerator packs 144GB of HBM3E memory, giving the system 288GB of HBM3E. AMD says there's a "path to four," opening up the possibility of two more accelerators bringing 576GB of HBM3E to the system. You'll need plenty of power to feed the GPUs, as each accelerator is rated for up to 600W.

Although AMD says it can support up to four accelerators, the workstation shown off at IFA only has room for two, both of which are liquid-cooled, alongside the Threadripper host. AMD doesn't sell MI350P accelerators on their own in traditional consumer channels, but the estimated price is somewhere around $20,000 per accelerator.

At a system level, the Threadripper Halo Station includes 2TB of DDR5 memory, which is the maximum capacity supported across the eight-channel memory configuration of the Threadripper Pro 9995WX. AMD supports up to DDR5-6400 on the Threadripper, though it made no mention of speed during its IFA presentation. Regardless of speed, 2TB of DDR5 costs about $50,000 right now.

AMD has yet to set a price or release date for the Threadripper Halo Station, though we'll likely hear more about the design from AMD's partners in the near future. An extremely expensive workstation isn't out of the question. The Lenovo ThinkStation P8, for instance, which uses Threadripper Pro CPUs as a host, clocks in at $334,463 right now, maxed out with 2TB of DDR5 and dual Blackwell accelerators.

Sales of graphics processing units for consumer PCs — which include integrated and standalone GPUs for desktops and laptops — totaled 75.5 million in the second quarter of 2026, up 10.4% quarter-over-quarter and 1.1% year-over-year, primarily driven by notebooks. This happened as the consumer CPU market contracted by 1.1% YoY amid a massive sequential drop in desktop CPU shipments and a significant rise in mobile CPUs shipments. Desktop GPU shipments declined by 4% quarter-over-quarter, while notebook GPU shipments surged by 16.8%, JPR claims.

But despite declining desktop PC unit shipments and modest growth in notebooks, unit shipments of discrete GPUs increased by 12.2% sequentially and 14.1% year-over-year in Q2 2026, according to JPR data. Jon Peddie Research does not publish absolute numbers of standalone graphics processors shipped in the second quarter, but our estimate is that around 20 million discrete GPUs were sold by AMD, Intel, and Nvidia in Q2, based on attach rates and Nvidia's market share and dynamics.

The results indicate that demand for PCs with discrete graphics remained remarkably resilient despite soaring component prices and slowing demand for desktop PCs. However, JPR's shipment data does not reveal whether the increase was primarily driven by gamers buying graphics cards, stronger demand for gaming notebooks, or other factors.

"The second quarter is typically down compared to the previous quarter," said Dr. Jon Peddie, president of Jon Peddie Research. "This quarter, discrete GPUs increased by 12.2%, even while a global memory crisis sent component prices soaring, driven by a mix of supply-side positioning, artificial demand shocks, and localized market dynamics."

Jon Peddie Research has yet to publish its complete desktop AIB report, which is expected later this month and will include market shares for AMD, Intel, and Nvidia; yet it is safe to say that the latter has maintained its undisputed leadership.

As for the overall consumer PC GPU market, Intel retained its leadership with a 56% market share as it increased shipments of consumer CPUs in Q2 2026. Nvidia came second with 23%, which is not bad at all considering that it only ships discrete GPUs. AMD came third with 21% share, up significantly from 14% in the same quarter a year ago, as it managed to gain seven percentage points of the consumer GPU market YoY amid growing sales of its consumer CPUs.

Broadcom CEO Hock Tan, Micron CEO Sanjay Mehrotra, SMIC co-CEO Liang Mong Song, and Huawei HiSilicon president He Tingbo all appear on the 2026 list, joined by Cerebras’ Andrew Feldman, CoreWeave’s Michael Intrator, and Nscale’s Josh Payne. Together, the roster includes Nvidia’s HBM supplier, two of its biggest GPU cloud customers, its custom-accelerator rivals, and China, which is trying — and so far struggling — to replace Nvidia altogether.

Parker’s profile, which appears under the Thinkers category, focuses on data center energy and water consumption, alongside Nvidia’s tripling of Scope 3 emissions in two years, and the Rubin platform’s shift to 100% liquid cooling. “Nvidia is at the center of the AI revolution,” Parker told TIME.

TIME editor-in-chief Sam Jacobs wrote in the list’s accompanying article that selection was led by editor Ayesha Javed, and that the 100 honorees represent the year’s “key storylines and who, in our opinion, are having the most influence in driving these developments.” Now in its fourth year, this edition of the TIME100 AI appears to have focused especially on featuring the new additions to TIME’s community of AI leaders, with Mark Zuckerberg, Demis Hassabis, Sundar Pichai, and Satya Nadella representing four other honorees, arguably deserving of a space on this year’s list, churned out.

Su’s strong TIME back catalog makes her omission all the stranger still. The magazine named her its 2024 CEO of the Year, put her on the 2024 AI list and the main TIME100 in 2025, and included her among December’s Architects of AI.

Nvidia’s second-quarter results, reported August 26, totaled $96.2 billion in revenue, including $89 billion from data centers, up 117% year-over-year. The company became the first to cross a $5 trillion market cap last October, and posted a $57 billion record quarter three weeks later.

The two chips in question aren't surprising if you know ball. Coming in at number six is the Core Ultra 7 270K Plus for $290 USD; Intel's most recently released desktop CPU, and at nine we have the workaday Core i7-14700K for $312, a steady seller since its launch in 2023. It offers less performance in gaming than many of the other chips in the top 10 (including the cheaper 270K Plus), but it has the advantage of being able to slot into both aging LGA 1700 systems as well as new builds with DDR4 memory. That's also likely why the Ryzen 5 5500 continues to stay near the top of the chart at #2, while the Ryzen 5 5600X and Ryzen 7 5800X3D (in its 10th Anniversary guise) also rate high. As for the Core i7-14700K, we actually recommended it over the Ryzen 7 5800X3D in our review of the 10th Anniversary re-release of that chip for people who are building new DDR4 machines.

The Core Ultra 7 270K Plus appearing this high is heartening for Intel's sake, though. If you haven't kept up with Intel's CPU releases, this chip is a course correction from the first-generation "Arrow Lake" processors, offering outstanding productivity performance and very solid gaming performance, just a few percentage points off of the beloved Ryzen 7 7800X3D that sits one position above it in the rankings. Put frankly, the Core Ultra 7 270K Plus beats everything else in the top 10 for both single- and multi-core performance save for the $569 Ryzen 9 9950X3D, which is in an entirely different price class from the $290 Intel chip (though discounted quite significantly at $569, 19% off the $699 list price)

It's clear that gaming continues to be a top driver for the DIY market. Five of the chips in the top 10 have been on our list of the Best CPUs for Gaming either now or at some point, including the number one seller, the Ryzen 7 9800X3D. AMD's 3D V-Cache continues to be unbeatable in gaming, even if the Arrow Lake 'Plus' parts do their best to close the gap with extremely high single-core performance. It's not single-threaded speed that's the bottleneck anymore, though, and not for some time; the limitation in gaming frames is multivariate as it has ever been, but on the CPU side, it's largely down to memory performance these days, and that's exactly why AMD's 3D V-Cache chips dominate.

Of course, if you're reading Tom's Hardware, you probably don't need me to tell you that. But it's interesting to look at the pricing of the parts that are moving the most units. The Ryzen 5 5500 is a tame processor, but if you need a gaming PC for the absolute bottom dollar, it's hard to beat six Zen 3 cores with a CPU cooler for $84. The Ryzen 7 7800X3D seems an absurd choice next to the much faster Ryzen 7 9800X3D for less than $100 more, but the gaming performance is really pretty close, and you're likely to desire a drop-in Zen 6 processor upgrade late this year or early next year, so why spend the extra $85 now?

That CPU upgrade angle warrants consideration. Any Socket AM5 processor has a path forward, and that's not the case for any extant Intel platform nor for Socket AM4, despite the fact that Intel is apparently releasing new LGA 1700 chips eventually. Ultimately, memory pricing is probably driving a lot of these choices, but in the upgrade context, it's not quite as surprising to see five and a half of the top ten CPUs on platforms that require DDR5 memory—the 'half' being the Core i7-14700K, naturally.

You absolutely can build a DDR4 machine using the Raptor Lake Refresh CPUs, but the performance penalty can be significant; in our testing, we found that it was actually worse than just running a single DDR5 DIMM. It's the PC builder's Sophie's Choice: buy a last-generation CPU to save significant cash on the RAM, or suck it up and spend out massively on a system that offers the best performance now.

Pricing is naturally also a factor; the Core Ultra 7 270K Plus launched at the comically low price of $299 in comparison to the $589 Core Ultra 9 285K, which it performs very close to, so $290 isn't much of a discount there. Meanwhile, the Core i7-14700K's $311.99 price tag is 16% off of Amazon's list price of $370.99, but it's a bit larger markdown from Intel's $419 recommended customer price. Still, if pricing is your primary motivator, it's hard to ignore the $330 Ryzen 7 7800X3D (27% off list), at least as long as gaming is your main concern. There's the 5800X3D 10th Anniversary, too, but again, the Core i7-14700K actually offers better minimum framerates and vastly superior productivity performance, especially if you're willing to do some memory tuning.

Clearly, gamers have accepted that memory pricing isn't coming back down any time soon, and that the time to buy is now before it gets even worse. The shortage isn't expected to abate between now and late 2028; if you're keen to build a new machine in the in-between, make sure you keep an eye on our Deals posts to save yourself from bleeding too much at the checkout.

The sky is blue for CPU (for now)

"In spite of a decidedly gloomy outlook on client processors from the suppliers for the second quarter, actual results for both x86 and Arm CPUs were up strongly in the second quarter of 2026, with sequential quarterly growth of the total market exceeding 10%, far in excess of normal seasonal trends which call for a slight decline in the quarter," said Dean McCarron, principal analyst at Mercury Research. "Behind this growth was a large increase in Intel's CPU supplies, especially in mobile client, after a couple of heavily supply-constrained quarters, and continued strong ramps for AMD's products. […] Arm shipments also appeared to be strongly higher in the quarter as well."

While the CPU market was up quarter-over-quarter (QoQ) mostly because AMD, Apple, and Intel increased their shipment volumes, on a year-over-year (YoY) basis, the CPU market contracted in terms of units. However, the decline was primarily driven by significantly lower IoT, SoC, and embedded shipments — largely due to AMD's shrinking game console business — as well as a substantial drop in desktop CPU volumes. At the same time, shipments of data center and notebook processors grew strongly.

Excluding IoT, SoC, and embedded products, both AMD and Intel substantially increased CPU shipments sequentially, but AMD grew faster and gained unit share in every major segment, so the company now controls 30.7% of the overall x86 processor market, its highest share ever, according to Mercury Research. With IoT, console SoCs, and embedded CPUs included, AMD now controls 34.1% of the x86 processor market. Intel continues to lead, but AMD's growth is impressive.

"AMD's total unit shipments and total market share reached new record highs in the second quarter of 2026, with a sequential share gain of 0.7% and an on-year gain of 6.5%," McCarron said.

Client CPUs: AMD gains share as Intel increases shipments

Performance of the client x86 CPU market was a mixed bag in the first quarter as sales of desktop CPUs declined badly, whereas shipments of laptop processors grew significantly. Intel remained the clear volume leader with 69.7% of client CPU shipments, but lost share both sequentially and year-over-year. AMD continued to gain ground in Q2 2026 as its unit share reached a record 30.3%, up from 29.6% in Q1 and 23.9% in Q2 2025.

AMD gained share in both client categories. Its desktop unit share rose from 33.2% to 34.9% QoQ, even though AMD itself shipped fewer desktop CPUs; Intel's shipments declined even faster. In notebooks, AMD increased its share from 28.3% to 28.9%, even as Intel expanded production capacity and shipped millions more CPUs.

In general, the results for the quarter suggest that Intel's improved client CPU supply helped to meet demand from PC makers and led to a client PC market rebound, the company could not stop AMD's share gains.

Desktop CPUs: The ugliest segment of the market

Hit by limited availability of graphics cards and high prices of components like motherboards, memory modules, and SSDs, the desktop CPU market was particularly weak in Q2 and contracted by over 20% YoY as well as quarter-over-quarter due to seasonality. Both AMD and Intel suffered significant annual shipment declines, but AMD held up better, according to Mercury Research.

Despite dropping desktop CPU unit shipments and overall market weakness, AMD increased market share in Q2 2026. AMD's desktop CPU unit share upticked to 34.9%, up from 33.2% in Q1 and 32.2% in Q2 2025, while Intel's share fell to 65.1%, from 66.8% sequentially and 67.8% a year earlier.

During the quarter, AMD gained 1.8% QoQ and 2.7% YoY in unit share because declines in its shipments were considerably smaller than Intel's. So, while normally we say that AMD is gaining desktop share because of strong sales of its latest Ryzen CPUs, in this case AMD's success was driven by slower declines than rapid gains.

Mobile CPUs: Sales of laptops are growing

The mobile CPU market segment performed completely differently from the desktop CPU market segment. Mobile x86 CPU shipments grew significantly quarter-over-quarter and were also modestly higher YoY. Mercury Research says Intel added millions of units of mobile CPU capacity during Q2, which helped close the supply-demand gap that had constrained the company in previous quarters. Yet, AMD's shipments increased at nearly the same pace as Intel's, which enabled it to capture 0.6% of the market.

AMD's mobile CPU unit share rose to 28.9%, up from 28.3% in Q1 and from 20.6% in Q2 2025. Intel remained dominant with 71.1% of shipments, down from 71.7% sequentially and 79.4% a year earlier. As a result, AMD gained a modest 0.6% sequentially, but a much more impressive 8.4% YoY increase.

Server CPUs: +20% year-over-year

Demand for server CPUs increased for both traditional data center processors — AMD EPYC and Intel Xeon — and CPUs used in networking and storage applications, so x86 server processor shipments posted nearly 20% YoY growth and moderately strong sequential growth. AMD continued to gain market share and now commands over 1/3 of the server CPU market, according to Mercury Research. There is a catch, though: due to Intel accounting peculiarities, AMD's unit share is artificially low and so is Intel's own dollar share.

In Q2, AMD’s server CPU unit share increased to 34.5%, up from 33.2% in Q1 and 27.3% in Q2 2025. Intel remained the volume leader with 65.5% of shipments, but its share fell from 66.8% sequentially and 72.7% a year earlier. While some may say that AMD's gains are slow (not too slow at +7.3% YoY), there is an explanation behind that slow growth.

Mercury Research notes that its broad server share calculation somewhat disadvantages AMD because Intel's edge and networking processors are now reported within its Data Center and AI (DCAI) business, so in reality, AMD's unit share should be higher. If the comparison is narrowed to EPYC versus Xeon processors for servers, AMD's unit share reaches 46.4%, up a substantial 2.9% QoQ and 9.2% YoY, which puts AMD very close to Intel in the core data center CPU market. Meanwhile, if CPUs for edge and networking applications are ignored in Mercury's analysis, Intel's dollar share would increase considerably.

Mercury Research indicates that server CPU supply constraints could limit gains in Q3, but in Q4 sales of data center-grade processors will increase noticeably both compared to Q2 and Q3.

Summary

After a weak Q1, the CPU market rebounded strongly in Q2 2026 as unit shipments grew more than 10% sequentially. The market growth was driven by rising notebook and server CPU sales, which more than offset a sharp decline in desktops, according to Mercury Research.

AMD's performance was stellar as it gained ground in every major segment even as Intel substantially improved processor availability. Despite improved availability of Intel client and data center processors, AMD still outgrew Intel and reached record overall and client CPU unit shares.

Desktop remained the major weak spot due to high component costs and generally lower demand as enthusiasts pulled in their purchases to 2025, greatly lowering sales of high-end hardware in 2026. Shipments of desktop x86 CPUs fell more than 20% YoY, whereas mobile and server volumes increased.

In core data center processors, AMD's EPYC share approached Intel’s Xeon share, which highlights how dramatically the competitive balance in the server market has shifted.

AMD estimates its progress at the rack level rather than on the CPU and AI accelerator levels, which obviously gives the company a lot of freedom in how to optimize performance efficiency. The company takes into account improvements in compute performance, process technology, memory bandwidth, data movement, interconnects, software, and system-level co-design. AMD says three major hardware characteristics determine AI system performance: compute capability, memory bandwidth, and interconnect bandwidth. New process technologies and architectures increase floating-point performance per watt, while improvements in memory integration and high-speed interconnects increase bandwidth available to processors. AMD expects these developments together to produce 20X higher AI performance per watt in 2030 compared to its 2024 baseline.

Memory and interconnects play particularly important roles because modern AI systems tend to move enormous amounts of data between accelerators and systems. AMD says higher memory bandwidth, greater bandwidth density, improved bandwidth per watt, larger caches, and tighter integration of memory and compute can reduce wasted energy consumption and increase performance efficiency. Meanwhile, faster scale-up interconnects can improve communication between GPUs, CPUs, and other components, which again increases performance efficiency. AMD's software optimizations are another part of the effort as higher performance achieved with optimizations ultimately means lower power consumption required to achieve an expected result.

What is a bit upsetting is that the 4X figure should be treated as an AMD estimate rather than a direct benchmark between two commercially available rack systems. AMD measures progress by comparing annual representative rack configurations with a 2024 baseline using the company's performance-per-watt methodology. In addition, its 2026 calculation combines measurements from actual products with modeled results in cases where final performance numbers were unavailable, which essentially means that AMD does not use its latest Instinct MI455X accelerators for its estimates.

If AMD reaches its 2030 targets, the company estimates that around two AMD racks in 2030 could provide the same amount of compute as 570 racks based on the Instinct MI300X from 2024. This essentially means a 20X lower power consumption or 20X higher performance at the same power consumption by 2030.

China on track for AI accelerator self-sufficiency

Nvidia commanded 66% of China's AI accelerator market in 2024, but its share dropped to 40% in 2025 and was on track to drop to 8% in 2026, according to estimates made by Bernstein investment bank earlier this year. Considering the fact that Nvidia did not officially ship any new accelerators to Chinese clients in the first half of the year, Nvidia's chief executive Jensen Huang said in May that his company's market share in the PRC was 'zero.' Of course, some Nvidia GPUs make it to China 'unofficially' as local companies are too dependent on Nvidia's CUDA and high-end AI accelerators. Still, it is safe to say that the bulk of new deployments in the PRC are based on hardware designed and produced domestically.

China's total available market of AI accelerators topped 4 million units in 2025, according to numbers published by Guancha.cn. Last year, 2.2 million Nvidia AI GPUs made it to the Chinese market, and while Nvidia's market share shrank to 55%, it still significantly outperformed its closest rival, Huawei, which shipped 812,000 AI accelerators and commanded 20.3% of the market.

Shipments by other players were by far lower: Alibaba's T-Head produced 265,000 AI accelerators, followed by AMD with 160,000. Cambricon and Kunlunxin only supplied around 116,000 AI processors each, whereas others shipped fewer than 100,000 units.

"This year, the Chinese government has actively encouraged the adoption of domestic AI chips," the report from TrendForce reads. "This policy push will likely provide priority support to high-potential domestic players, allowing them to substantially expand their market share in China's high-end AI server market. At the same time, the domestic ecosystem is maturing in key areas such as advanced foundry nodes, advanced packaging, and thermal management."

The firm now expects shipments of high-end AI processors developed by Chinese companies to increase by more than 83% year-over-year in 2026 as domestic production capacity and deployments expand. As a result, its analysts project domestic AI accelerators to capture nearly 90% of sales (up from 45% last year), which means that foreign suppliers like AMD and Nvidia will be left with roughly 10%. This latest projection represents a major revision from the research firm's December 2025 outlook, which estimated that Chinese processors would account for approximately 50% of China’s high-end AI chip market in 2026.

Dual-track strategy

As AMD and Nvidia supplied some 2.36 million AI accelerators to the Chinese market last year, commanding a 59% unit share, replacing the majority of them will take a lot of effort, assuming that the TAM will remain at around 4 million units. TrendForce claims that China is set to adopt the so-called dual-track strategy, which involves AI accelerators from merchant suppliers like Huawei and Cambricon along with custom AI ASICs from Alibaba, Baidu, ByteDance, and Tencent.

"Together, these developments are moving China's AI infrastructure away from its heavy reliance on foreign GPUs, toward a dual-track model of 'domestic GPUs + proprietary ASICs,'" the firm claims.

Hyperscale cloud service providers (CSPs) are inclined to expand usage of their own silicon because it is cheaper compared to merchant accelerators and because it is optimized for their workloads and data formats. Meanwhile, developers of merchant AI hardware — such as Huawei, Biren, and Cambricon — will also gradually expand their output of accelerators as demand is very strong.

Bottlenecks

It remains to be seen whether the Chinese chipmaking industry can indeed replace 1.96 million high-end AI accelerators in just one year. To maintain the 4 million unit TAM, China's semiconductor industry will need to increase AI accelerator output by 2.2X in just one year.

SMIC — China's largest and most advanced foundry — this week announced that its Q2 2026 revenue increased to $3.005 billion, up from $2.505 billion in Q1 2026, and from $2.209 billion in Q2 2025. This suggests that the company is both increasing the output of chips and its prices. However, it remains to be seen whether SMIC's 36% YoY revenue increase is an indicator that it can increase output of high-end AI accelerators by over 2X compared to 2025.

Another major bottleneck for the Chinese industry is the lack of domestic production of high-bandwidth memory (HBM). Huawei has reportedly acquired plenty of HBM2-class memory from Samsung, but its stock is not endless, so its Ascend 950-series AI accelerators are set to rely on proprietary HiBL 1.0 and HiZQ 2.0 types of memory, not industry-standard HBM2 or HBM3. While China's DRAM champion CXMT is gearing up for HBM3 manufacturing in late 2026, it remains to be seen how quickly the company can ramp up production to decent levels.

Nvidia's CUDA software stack is the company's biggest advantage after the performance and versatility of its AI accelerators. But while performance can be matched with brute force, the software stack cannot be reproduced quickly. Last year, Huawei opened up its CANN software stack to accelerate its development, though we do not know if the company has achieved its targeted goals with this. Yet, without a doubt, China's AI software stack is getting more mature every year, so many new AI deployments may indeed rely on domestic stacks rather than on CUDA.

A shifting market

U.S. restrictions on exports of advanced AI accelerators, combined with China's own efforts to limit the use of American AI processors domestically, have largely pushed companies like AMD and Nvidia out of the Chinese market. Analysts now expect China-based independent hardware vendors to control 90% of the domestic AI accelerator market in 2026.

Chinese AI hardware has come a long way, and Huawei's solutions can outperform Nvidia's NVL72 GB200 rack-scale system, albeit while consuming more power. Therefore, if power is not a concern, Huawei can build AI data centers with performance that matches or exceeds those based on Nvidia GPUs.

However, replacing American GPUs almost completely while maintaining AI accelerator unit TAM at 4 million units will require China's industry to product 1.96 million AI accelerators in 2026, 2.2X more than in 2025. This seems impossible not only for TSMC, but also for local memory makers that still have to start making HBM memory.

To that end, while Chinese AI accelerators may indeed capture 90% of the domestic market, without hardware from American companies, that market can shrink dramatically in terms of units.

"In addition, graphics memory costs are expected to rise further, leading to a substantial increase in VGA Card* costs in the second half of the year," a statement from PC Partner in its financial results for the first half of the year reads. "The quantity of VGA Card shipments declined substantially in the first half of FY2026 as compared with the second half of FY2025, and availability of VGA Cards is likely to decline further in the second half of FY2026."

GPU prices have increased massively in the recent quarters primarily due to high demand amid shortages of components. But just when you thought that AIB prices cannot get any higher due to the risk of a collapse in demand, GPU producers say otherwise. While the increase is not completely unexpected, what looks a bit unexpected is that entry-level graphics cards will increase the most, according to PC Partner.

"Beyond the price increases, entry-level VGA Cards are expected to face even more severe shortages, which may further drive ASP upward in the second half of FY2026," the statement by PC Partner reads.

Jon Peddie, the head of Jon Peddie Research, is a bit skeptic about whether the volume of graphics cards sold is going down, at least when it comes to consumer products.

"In Q1 2025, 9.25 million AIBs were sold, and in Q1 2026, 11.82 million; that is an increase of 28% in units, not dollars," Jon Peddie told Tom's Hardware. "AIB sales ($) will also increase, driven by higher memory costs. Our POS tracking of AIBs shows Q1 2025 at $6.18 billion and Q1 2026 at $ 10.85 billion, a 76% increase."

The question is whether price increases have benefited anyone beyond those early in the supply chain. Jon Peddie, who has been tracking the graphics industry for about 25 years, does not seem to believe that all the graphics card prices are due to increased costs.

"[This all does] not necessarily equal an increase in profits, although I suspect it does, and the AIB suppliers are taking advantage of all of the press reports on higher memory prices and shortages to raise AIB prices more than the delta in memory costs," Peddie said.

The current 2026 graphics card price surge has been very uneven by segment. The percentage increase has generally been much larger on high-end AIBs that carry a lot more memory than entry-level boards. However, it seems that the situation is going to change and even 'cheap' graphics cards are going to get more expensive. There is a more alarming trend: the drop in sales of graphics cards.

PC Partner's own brand business — principally consisting of Zotac, Inno3D and Manli that work exclusively with Nvidia — saw unit shipments fall 18.4% in the first half of the year while seeing its average sales price increase by only 10.7% YoY.

This hugely contrasts with PC Partner's financial results in its ODM/OEM graphics-card business — mostly consisting of AMD and Sapphire-branded products — that has weakened substantially. PC Partner says unit volume actually fell 38.4% year-over-year, yet its revenue increased 73.9%, from HK$853.3 million to HK$1.4843 billion. The reason: average selling price jumped 181% YoY as PC Partner received substantially more orders for high-end AIBs. Since PC Partner's ODM/OEM produces cards for AMD, Sapphire, and plenty of PC makers, this could indeed be a bad sign, but not necessarily, according to the renowned analyst.

"The ODM/OEM suppliers get closer scrutiny from the big companies (Dell, HP, etc), and those companies know the actual costs of the memory," Peddie said.

*Being an established maker of graphics cards, PC Partner still uses the 'video graphics array,' or VGA, term from the 1980s to describe graphics cards.

"Market chatter suggests [Google] is working with AMD on a TPU project in the v10 generation," a SemiAnalysis note for clients cited by Sean reads. "AMD's involvement would be the first real involvement in a custom AI ASIC project, despite having a custom silicon team. AMD has strong IP, especially in advanced packaging and SoIC. Additionally, CPU IP could also be a draw given Google and its customers are pushing for TPUs with on-package CPU cores for RL workloads."

Having developed nine generations of its proprietary AI accelerators (with Broadcom acting as actual silicon designer) and possessing extensive expertise in accelerator architecture, Google hardly needs AMD to design a conventional TPU. Hence, chances that AMD will implement Google's TPU v10i for inference or V10t for training are low. Hence, Google might need something only a CPU maker like AMD could provide, including CPU IP, programmable logic, interconnects, or certain advanced packaging know-how.

Of these, the CPU angle is particularly noteworthy. SemiAnalysis claims that Google and its customers are pushing for TPUs with on-package CPU cores for reinforcement learning and potentially other CPU-heavy workloads. While conventional LLM training remains overwhelmingly accelerator-heavy, reinforcement learning for reasoning and agentic models can require considerably more general-purpose compute around accelerator operations.

Google has already begun to increase CPU resources around its latest inference-oriented TPUs. Its TPU 8i systems, designed for inference, reasoning, and RL workloads, feature one Google Axion CPU for every two TPUs. By contrast, servers running Google's 7th Generation TPUs used one Xeon 'Emerald Rapid' processor for every four TPUs. Furthermore, we are hearing that in some cases a 1:1 ratio of CPUs to accelerators is optimal, so the future of AI may be way more CPU-heavy than we think.

Meanwhile, bringing CPU cores directly into the TPU package could be a logical next step, as reducing the distance between general-purpose and tensor compute can improve performance and reduce power consumption. This is where AMD comes into play, as it already has experience developing a data center-grade design — the Instinct MI300A — that packs both x86 and accelerator chiplets. A hypothetical Google design could therefore combine Google-developed TPU compute chiplets with AMD CPU and HBM in a tightly integrated package built by AMD. Perhaps, Intel would appear as another potential candidate given that Google and Intel have multiple strategic collaborations. Yet Intel has no experience building hybrid x86+accelerator data center designs.

Note that for now we are speculating and our analysis may be inaccurate. For now, the nature of AMD's alleged involvement remains unclear. Yet, if the report is indeed accurate, the important development may not be that AMD is helping Google build another TPU. Instead, what matters is that Google is considering a new CPU-heavy member of its TPU v10 family, optimized specifically for RL and agentic workloads, and is using AMD as a provider of some of the building blocks needed to create it.

Last month, hot off the BC-250's full 40 CU unlock patch, we thought the time was ripe and finally put together a BC-250 build of our own to see what all the fuss was about. In case you're out of the loop, the AMD BC-250 is a repurposed PS5 APU, designed for cryptocurrency mining. With mining firmly out of fashion, the chip has now found new life, thanks to community-made patches and fixes that get things up and running for gaming. After all, where else are you going to find 16GB of memory and a capable single-board computer for $200 (or thereabouts) in the current market?

The experience of reading Jake's trials and tribulations feels like harkening back to the days when everything wasn't simply plug-and-play, and if you wanted the most out of your hardware, you were going to have to get a little bit uncomfortable and more intimately familiar with the nature of the silicon you're running. If you're interested in running one of your own BC-250 systems, you don't want to miss this fantastic (and lengthy) read, which features extensive benchmarks comparing the build to the Steam Machine, alongside performance results using different CU counts and OS images.

• Benchmarking AMD's BC-250, offering Steam Machine-like performance at half the price

We also interviewed Intel's VP and GM of Enthusiast Business (and AMD's ex-Technical Marketing Director) Robert Hallock. As always, we give premium readers full access to our entire 45-minute session with Hallock, no redactions, no fluff. During the interview, Hallock addresses how Intel's entire team shifted between the release of Arrow Lake and Arrow Lake refresh platforms. He also gave us some insight into the importance of CPU software optimization, as well as dishing out some details about the highly anticipated Nova Lake lineup. (Yes, we asked if they had a V-Cache competitor, and you'll have to read the article to find out how Hallock responded.)

Catching a high-level executive in between product cycles is usually against the traditions of the press and marketing cycles that companies usually go through. This offers a fairly unique perspective, where we get to see how Intel is not only iterating upon its previous products, but also being somewhat reflective on the last few bumpy years for the company's consumer CPU segment.

• Intel VP Robert Hallock sets Nova Lake expectations, teases return to Raptor Lake for DDR4 platforms — our full 1:1 interview transcript

Elsewhere in the industry, we're starting to see real PCIe 6.0 devices that companies are actually going to be able to purchase. The standard suffered from a false start, mainly owing to the shift to PAM4 signaling, which behaves very differently from the Non-Return-to-Zero (NRZ) signaling that was deployed across prior generations. Several years on, we're now seeing commercial devices begin to arrive, including SSD's with capacities up to a staggering 2 Petabytes.

But for an ecosystem to thrive, you don't just need the devices; you need the ability for systems to communicate with them, too. So, we've offered an overview of the controllers, in addition to the drives themselves, to assess the current state of the PCIe 6.0 market.

• The current state of PCIe 6.0 SSDs and controllers — Marvell, Phison, and SMI prepare controllers as drives finally come to market following years of delays

In the news

In addition to the three features that we ran on Tom's Hardware Premium this week, we've also published a slew of News Analysis pieces, which aim to dig deeper into the biggest topics in the hardware and semiconductor industry.

Firstly, there have been dozens of headlines across the U.S. over the ongoing AI data center buildout. We've covered the protests, the impact, and the voices who are lending their ears to striking back against Big Tech's appetite for more compute by any means necessary. Our report collates the ongoing sentiment surrounding the accelerating buildout and how residents are fighting back.

• Over 70% of Americans oppose AI data centers; US protests intensify as more arrests are being made — almost 40 arrested this year in backlash to AI factory buildout

Hyperscalers are not backing down from putting money down to build more compute, which is currently constrained as AI models grow larger and more complex and as demand grows. To that end, we've seen that an eye-watering $2 trillion USD has been pledged to secure AI hardware and DRAM alone. The ongoing AI megatrend is transforming the semiconductor industry, all the way down the entire supply chain, and that's forcing companies to rethink how compute is purchased. At least for Apple and Google, that means putting up billions to secure critical components like NAND, DRAM, and more.

• Hyperscalers commit nearly $2 trillion to secure AI hardware and memory — Google leads $811 billion spending surge while Apple trails at $57 billion

Optical interconnects and photonics have become a flashpoint amid the ongoing AI saga. With the limitations of copper meeting the very real demand for faster connection speeds, optics is the answer that the industry has been developing for quite some time. However, the FCC's proposed Secure Networks Act could put the deployment of silicon photonics in jeopardy. The FCC's proposal to ban new-model optical transceivers could put an already incredibly strained supply chain at risk.

With 56% of global manufacturing for optical modules shored up in China, this would pose an immediate problem, and one that the markets have already responded to in-kind. We have broken down exactly how optical became the hottest topic in the semiconductor manufacturing industry, as well as how the U.S-China trade summit next month could weaponize the technology.

But what if there were other options? The photonics supply chain remains constrained and relatively immature. Full co-packaged optics chips might not be ready for full volume production, despite Spectrum-X CPO being rolled out across Nvidia's Vera Rubin lineup. The answer, in the short-term, may be near-packaged optics, as we explain all of the comings and goings in the wild world of silicon photonics.

• FCC proposes import ban on Chinese optical transceivers — blockade targets key AI interconnects as China holds 56% global market share
• How optical interconnects and silicon photonics emerged as AI's next hot commodity — looming US-China summit puts photonics into the crosshairs
• Near-packaged optics (NPO) gains ground as the industry hedges against CPO's growing pains

That just about wraps up everything we've published over on Tom's Hardware Premium this week. To read the above articles and to gain access to Bench, you'll need to purchase a subscription, which we've linked above.

We'll be back next week with another roundup of exclusive Tom's Hardware Premium reads that you won't find anywhere else.

"We intend to use the net proceeds from this offering for general corporate purposes, which may include the repayment of debt," an AMD statement with the Securities and Exchange Commission reads.

The offering comprises four tranches: $1.25 billion of 4.6% notes due in 2029; $1.50 billion of 5% notes due in 2031; $1 billion of 5.25% notes due in 2033; and $1 billion of 5.5% notes due in 2036. Their yields to maturity are 4.64%, 5.018%, 5.264%, and 5.532%, respectively, while spreads over comparable U.S. Treasuries range from 43 to 90 basis points, which indicates that the market is generally confident in AMD and is willing to lend it money at rates that barely exceed those of the U.S. Treasury. Moody's and S&P are expected to rate the securities A1 and A, respectively.

AMD did not disclose how it plans to spend $4.75 billion, but the additional money obtained at attractive rates gives it room to finance its increasingly capital-intensive business as well as cash for debt repayment and other corporate requirements.

AMD hardly appears desperate for additional money. At the end of Q2 2026, the company had approximately $13.1 billion in cash, cash equivalents, and short-term investments. AMD's debt totaled $3.2 billion, and only $875 million is classified as current, which means that the proceeds from the offering by far exceed AMD's current obligations. Meanwhile, AMD's business is becoming very capital intensive.

At the end of 2025, the company had around $12.2 billion in unconditional commitments, which include purchases of wafers and substrates, multi-year cloud-service agreements, software and technology licenses, and guaranteed obligations to third parties. Approximately $8.5 billion was due in 2026.

Also, AMD's working capital requirements are growing. Inventories reached approximately $8.47 billion by the end of Q2, while accounts payable climbed to $5.36 billion. AMD also spent $1.20 billion on property and equipment during the first half of 2026, compared with $494 million a year earlier.

If we were to speculate where AMD can put $4.75 billion, then long-term supply agreements for commodities like memory, logic production, or advanced packaging immediately come to mind. However, given the current market realities, $4.75 billion is 1.8x smaller than AMD's inventories as of late Q2 2026. Furthermore, an average long-term supply deal with a major memory maker now amounts to $7.14 billion (according to Micron's comments made in its recent earnings release).

That said, $4.75 billion may not be enough for AMD to make strategically important purchase commitments. Nonetheless, getting nearly $5 billion at attractive rates amid global undersupply of pretty much everything certainly gives AMD some additional flexibility to run its business.

The exploit is called Skitter Creek Bath Salts (Skitter), and was developed by prolific hacker Christopher Domas, famous for finding CPU flaws like Sandsifter and God Mode Unlocked. Only AMD chips from the 15h and 16h families are affected, roughly 2011 to 2015 vintages. Family 15 is FX-series desktop chips and some Opterons, while 16h includes low-power Jaguar- and Puma-based SoCs like those in the PlayStation 4 and Xbox One, plus a handful of Athlon, Sempron, and Opteron-X chips, among others.

To pull off this exploit, you'll need kernel-level access, meaning the ability to run your own drivers. But once you do, the entirety of DRAM is your oyster. AMD published a security bulletin on the matter, saying these chips are out of security support, plus, as mentioned, the necessary access level means an attacker already controls the machine anyway.

If you're confused as to how one instruction opens up a system, here's our attempt at a simplification. Say you have 16 GB of RAM. You'd think that Windows gets all 16 GB to play with, from address 0 to the end of memory — but as you may have noticed before, it's actually a bit less than that. The rest is reserved for system-level data.

Some parts are visible to the OS so it can interact with devices, but others include Very Important Things like PSP, SMM, microcode patches, all in sections supposed to be completely untouchable. If they were accessible, the system as a whole wasn't secure by definition anymore — just think of a malicious driver being able to freely mess with how your processor handles data.

For performance reasons, modern processors' RAM controllers don't use memory in a straight line, so to speak — they use bank interleaving, meaning that the actual bytes in the DRAM are jumbled, all while the OS sees a nice, tidy, flat surface. As it turns out, the CPU setting that controls this feature is accessible to the OS in the aforementioned chip families, and it's called BankSwizzleMode (Swizzle). It can be toggled on or off with the instruction "xor dword [0xf80c2094], 0x00400000", a simple bit twiddle. And as it turns out, this can be exploited.

First, you run a loop to figure out how the mapping normally functions. You place a canary value in memory (say, 0xDEADBEEF, according to tradition), disable Swizzle, run through memory to see where it landed, and reenable Swizzle again. Do this enough times, and you know exactly how visible memory is mapped into physical DRAM, and vice versa.

With the map now in your possession, you can now disable Swizzle and force a read or write to normally inaccessible areas of the DRAM, since you now know where it will land. With this, you can access all the previously hidden code and data, netting you hardware-level access to do anything you want, including reading fTPM signing code and any other low-level shenanigans you can think of.

Attentive readers might be wondering why the system doesn't crash during this process since you're effectively temporarily turning the RAM into a spaghetti mess. The answer is that every time you enable and disable Swizzle, you prepare the CPU by disabling interrupts, along with a number of other measures. Even still, the machine can crash during the map-collection step, but that only needs to be done once. After you have the map, the likelihood of a crash is fairly low since you'll be targeting specific locations.

Another question might be why sending a bit to a memory location somehow messes with the CPU, and the answer is that part of OS-accessible memory is actually mapped to hardware according to the Memory-Mapped Configuration Space standard (MMCONFIG) — meaning that reads or writes to that space are directed to hardware configuration settings, not actual RAM.

Edit 8/14/2026: Clarified TPM.

The Ryzen 7 9800X3D has since reclaimed the top slot — it is the best CPU for gaming, after all — but the screenshot touches on the prevalence of DDR4 platforms, nearly four years after the launch of Zen 4 and introduction of DDR5 to AMD's platform. It's not just the Ryzen 5 5500, either.

Of the 25 best sellers on Amazon, three slots are occupied by Intel, and one slot is claimed by the Thermal Grizzly AM5 contact frame. AMD holds every other slot, though it's nearly an even split between DDR4 and DDR5. There are nine Zen 3 CPUs on DDR4 platforms to 12 Zen 4/5 CPUs on DDR5 platforms.

It seems Marinković's screenshot is a bit out of date — currently, the split is even closer, with Intel picking up two extra spots and the split between DDR4 and DDR5 AMD CPUs moving to nine and 10 slots, respectively.

The irony here is that Marinković is highlighting AMD's representation on the best-sellers list while a four-year-old CPU is at the top of the charts, not one of the many performant DDR5 processors AMD has released since. It's not hard to draw a straight line between DDR4 popularity and current RAM pricing on your own. However, several motherboard vendors have previously confirmed to Tom's Hardware that they're increasing DDR4 motherboard production due to demand for older platforms.

The Ryzen 5 5500 is one of the cheapest CPUs you can buy at only $80, and it comes with a Wraith Stealth cooler in the box, saving you some money on an aftermarket CPU cooler. It's based on the Zen 3 architecture with six cores and 12 threads and boosts up to 4.2 GHz. It has a cut-down L3 cache compared to other Zen 3 CPUs at 16 MB, however, and it tops out at PCIe 3.0.

We recently published a $100 CPU shootout comparing the Ryzen 5 5500 to some other options around the same price and found its performance lacking, particularly in games; the Intel Core i3-14100F consistently offered better performance with DDR4. That conclusion only holds up if you don't already have an AM4 motherboard, however. If you're in the market for the Ryzen 5 5500, you're already on a tight budget, so a new motherboard probably isn't in the cards.

First, we have the Steam Machine. Valve’s long-awaited console-like PC is here and much more expensive than anyone expected at over $1,000. You can easily put together a BC-250 build for around $400 or $500, and even less if you have a power supply and SSD lying around. If you can get in the realm of Steam Machine performance for half the price (and the BC-250 can, based on our testing), that’s pretty compelling.

There have also been some significant developments among the BC-250 community. The board only works in the first place due to a community-developed BIOS and GPU driver, and developers have continued to experiment with what’s possible with the hardware. Now, there are tools that enable all 40 Compute Units (CUs) on the board (it defaults to 24 CUs), as well as tools to enable the two disabled Zen 2 cores, taking the six-core CPU to an eight-core CPU.

Now, the BC-250 isn’t a cut-down PS5 APU, at least from a hardware perspective. It actually patches two more graphics CUs compared to the PS5. Though, as should become clear throughout this story, relating the BC-250 to the PS5 on anything but the silicon on the board isn’t the best idea.

All in, we spent about $350 on the BC-250, with around $200 of that going toward the board (it’s currently listed for around $175 on eBay) and the additional budget going toward a 3D-printed case, fans, and a power supply. You could spend as little as the cost of the board if you can 3D print a case, bring your own fans, and repurpose an old PSU and NVMe SSD. If you’re starting from scratch, you’ll spend between $400 and $500.

Even for “full” price, the BC-250 is significantly cheaper than what you can get elsewhere. The PS5 digital edition is now $600, while the Pro will run you $900. The newly-released Steam Machine starts at $1,050, and that’s with just 512 GB of storage. The BC-250 is significantly cheaper and offers performance that can rival these platforms. What you save in money, however, you spend in time.

The BC-250 is a great project if you like tinkering, but it’s not a set-it-and-forget-it gaming device, even after the initial setup. After a few weeks of using the board, that much became clear.

A PS5 APU, not a PS5: BC-250 specifications

The BC-250 is described as a cut-down PS5 APU, and that’s exactly what it is, but the “cut-down” portion of the description carries a lot of weight. It comes with eight Zen 2 cores, but two of them are disabled, and a 40-CU RDNA 2 GPU, though only 24 of those CUs work out of the box.

The community has been able to unlock all 40 CUs with good success. Ultimately, your mileage with unlocking all 40 CUs will vary; some boards will work just fine with all 40, others with 32 or 36, and some with the base 24. As I’ll dig into more later, I was able to unlock all 40 CUs and run them stably up to 1,850 MHz, much higher than the recommended 1,500 MHz max frequency.

Out of the box, both the CPU and GPU run at locked frequencies of around 3.5 GHz and 1,500 MHz, respectively. You’ll need to separately install a GPU and CPU governor for dynamic frequency scaling (a necessity in the case of the GPU, though optional for the CPU). Achievable frequency on the CPU and GPU is entirely determined by your thermals. Both governors allow for overclocking, but the BC-250 is thermally constrained, even with active airflow, so don’t expect to push far beyond the stock frequency, particularly on the CPU.

Unfortunately, storage is a big bottleneck for the BC-250, and one of the biggest points of divergence compared to a PS5. The M.2 runs at just PCIe 2.0 x2 speeds, delivering about 1 GB/s of maximum performance with an NVMe SSD. The good news is that, due to how slow the interface is, you can save some money and get a less-performant but larger drive for storage.

As of a few days ago, at the time of writing, some initial fixes to unlock the extra two CPU cores have rolled out, which I was able to get working after a bit of trial and error. Outside of the GPU and CPU, the BC-250 has a bit less L3 cache. The APU, from a hardware perspective, is very similar to the PS5. Though that ignores the complex I/O system present in the PS5, not to mention the various layers of software that Sony runs on the APU.

Given that the BC-250 wasn’t designed for gaming, it’s important to keep that context in mind. You could just as easily describe it as a crypto mining board that’s been repurposed for gaming as you could a PS5 APU that’s been repurposed for mining. The BC-250 is a mining board, and the only reason it works is due to a large number of enthusiasts trying to get the damn thing to work and sharing their advancements along the way.

Although Linux gaming broadly has made a ton of advancements over the past few years, the BC-250 has unique hurdles you need to overcome. It needs fixes to scale the frequency on both the CPU and GPU; otherwise, it’s locked at a fixed frequency. It needs a custom BIOS image to even boot, and a specific configuration in the BIOS to avoid immediately crashing. Those are all steps you need to take to get the BC-250 to work in the first place, let alone start optimizing it to squeeze out the best performance.

There is extensive community documentation for the BC-250, which I referenced ad nauseam throughout this process. It’s an excellent resource, but even still, I had to search out solutions outside of the documentation. This is cobbled together from community fixes, and the documentation seeks to cover as many distros as possible. But it’s nigh impossible to have complete documentation on the BC-250, as valiant as the effort linked above is.

Expect a bumpy (but rewarding) road. It’s better to think of the BC-250 as a project, rather than a commodity like the PS5. Yes, you can play games on it and get surprisingly good performance, but you’ll spend a good amount of time simply tinkering with the machine, at first to get it to work, and later to optimize it. And you will inevitably run up against games that either don’t work or have poor performance, prompting another investigation and optimization cycle. I find the process rewarding; others will find it frustrating.

Setting up and configuring the BC-250

Getting the BC-250 is only one part of the process here. At the least, you need a power supply that can deliver at least 20A on the 12V rail. A standard PSU will work, but FlexATX power supplies are a better fit. I ended up purchasing a 3D-printed case, along with a power supply, from eBay. The BC-250 documentation includes several PSU recommendations, along with case designs, if you want to go about choosing your parts piecemeal.

Here’s what I’d recommend at minimum:

• BC-250
• 300W+ PSU (at least 20A on 12V rail)
• 2x Arctic P12 high-pressure fans (if using CU unlock)

Regardless of the route you choose, double-check fan mounting points (there aren’t any mounts on the BC-250 itself) and case/PSU compatibility. Many small form factor BC-250 designs you can 3D print are designed around a FlexATX PSU. Before booting, verify your 8-pin pinout against the community documentation. If the 12V pins are in the wrong position, you can fry your board.

With the board ready, the first step is to flash a new BIOS from the shell. It’s straightforward enough, though without the modern conveniences of firmware flashback in the event of a failure. If you lose power, you’ll brick your board and need a hardware controller to reset it, so I recommend connecting to a UPS during the flashing process. You can grab the correct files from the documentation.

From there, you need to clear your CMOS by removing the battery and configure a few BIOS options to force the integrated graphics, set the UMA frame buffer size, and disable IOMMU. From there, you need to choose which Linux distro you’re going to use, which is very important. I’d recommend experimenting with a few different distros, especially if you’re new to Linux.

For gaming, the go-to options are Bazzite and CachyOS, the former of which I originally went with. There are some “optimized” images floating around for the BC-250 that supposedly rope in all of the configuration and fixes you need into a fresh OS image. I wouldn’t recommend using those. Setting up the BC-250 is getting easier by the day (I have found new fixes and scripts that automate installing several fixes just in the time I’ve been testing the board), so these optimized images are, at best, out of date, and at worst, not actually optimized at all. Tread carefully.

You don’t need an optimized image because, at least for Bazzite and CachyOS, scripts exist that can automate the entire setup process. Strictly speaking, the only thing you need is the GPU governor. Previously, you would need a kernel patch, but a GPU governor via an SMU now exists and is what I used. The governor is what allows dynamic frequency scaling on the GPU, as well as overclocking.

You’ll also need a compatible kernel, which is easier to manage on rolling distros like CachyOS. 6.18.18 LTS is the recommended kernel at the time of writing, but 6.17.11+ works, along with 6.12.x to 6.14.x LTS. Other kernels may work, though some (such as 6.15.0) are broken and will trigger a kernel panic.

With your distro chosen and a correct kernel working, here are the optimization steps I took:

• Install GPU and CPU governor
• Unlock 40 CUs
• Setup Zswap with 32GB swap page
• Install ACPI fix
• Configure PWM sensors

Again, scripts exist for these patches, the most critical of which don’t even require kernel patching any longer. The exact fixes you need and the method to install them will depend on the distro you choose, however, so make sure to keep the setup documentation handy.

The list above is in order of importance. Once you have the GPU governor, the next step is to try to enable the full GPU. Previously, you would need a kernel patch, but the bc250-cu-live-manager utility can enable the CUs via a User Mode Register (UMR) without the need for a kernel patch. Not only is this easier to do, but it also persists across updates, unlike a kernel patch, where new kernels will force you to go through the unlock process again. The end result is identical regardless of the path you choose.

I spent a good deal of time testing the 40 CUs before doing any other optimizations, and I’d recommend you do the same. Just because the board has 40 CUs doesn’t mean you can use all of them. I was able to unlock all 40, though not at the frequencies some others have reported. And some users aren’t able to use all 40, rooting out bad clusters for a 32-CU or 36-CU configuration. Overclocking-like trial and error is necessary here, though, as we’ll get to in the performance section later, experimenting is worth the hassle. Using all 40 CUs, even at suboptimal clock speeds, offers much better performance.

Although setting everything up on Bazzite should be easy, I ran into issues numerous times. It’s an immutable OS image, which can cause problems depending on the fixes you’re trying to apply. It’s a great choice if you want to quickly set up the BC-250, but it became clear that an immutable image isn’t optimal if you want to get the best performance. Further, some newer fixes simply wouldn’t work, such as unlocking the extra two CPU cores.

The CachyOS pivot

After a few days of optimizing, testing, and troubleshooting Bazzite, I made the hard decision to wipe the machine and start fresh with CachyOS. In hindsight, I should have started with CachyOS from the beginning for numerous reasons, but mainly because it’s a rolling Arch-based distro. Bazzite is a Fedora Atomic-based distro. It’s immutable, adding guardrails to make it difficult to break anything, but in the process, making it much more difficult to apply low-level changes like kernel patches.

If you’re not familiar with Linux and don’t care to become familiar, Bazzite is your best bet. It’s convenient, and it includes a ton of small community fixes right in the startup menu, such as the LSFG-VK project for Lossless Scaling. It became clear, however, that you’re trading performance for convenience with Bazzite. CachyOS is more hands-on, but it’s much easier to actually accomplish what you need to with the BC-250.

Most notably, CachyOS (and other rolling release distros) allow you to easily manage your kernel. There are alternative methods for critical setup items like the GPU governor and CU unlock that don’t require a kernel patch. However, having the ability to easily patch your kernel helps when you run up against compatibility or performance issues, allowing you to try alternative installation methods or community fixes.

Further, CachyOS has access to yay, or Yet Another Yogurt, the simple command-line call to search and install packages from the Arch User Repository (AUR). It takes some time getting used to living in the command line as often as you do on any Arch-based distro. But even knowing just a few basic commands allows you to accomplish what you want much faster than fumbling around with flatpaks, at least in my experience.

Getting set back up was simple thanks to a community script that automates nearly all of the configuration and optimization for the BC-250. The BC250-Toolkit script brings together all of the various community patches, including optional optimizations, and it just worked. Coming off of Bazzite, it was a treat not to run up against the immutable walls of the distro. I was able to accomplish in a few hours what took me more than a day with Bazzite.

CachyOS specifically has a few optimizations for gaming, as well. First, there’s the BORE, or Burst-Oriented Response Enhancer, scheduler that, as the name implies, is optimized for bursty workloads (Michael Larabel over at Phoronix has a great writeup on that). CachyOS also includes packages for x86-64 v3 and v4, the former of which is relevant to the BC-250.

These optimizations represent a substantial performance improvement. In Forza Horizon 6, for example, I went from 61 FPS at 4K with the Low preset to 72 FPS on CachyOS after it was configured properly (with 40 CUs in both cases). In Cyberpunk 2077, I went from 74.7 FPS on Bazzite to 82.2 FPS at 1080p with the Steam Deck preset, and perhaps more impressively, from 47.8 FPS to 56.5 FPS at 1440p.

I saw higher peak performance on Bazzite in Unigine Superposition, though that’s likely due to more thermal headroom for the GPU, as I wasn’t able to enable the extra CPU cores on Bazzite. I mainly wanted to show this chart to bring the “optimized” Bazzite image into the fold, which was supposedly an image that was already configured for the BC-250. Using a script to automate the installation of several tools (like the one available for CachyOS) is one thing. Using an entirely custom image is another. I’d suggest starting with a clean, official image regardless of the distro you choose.

Testing performance on the BC-250

The best touchstone for comparisons with the BC-250 is the Steam Machine. The Steam Machine is more powerful simply based on the spec sheet. I ran a truncated list of benchmarks to see how the BC-250 stacks up, as well as several additional benchmarks on the BC-250 alone. Performance here is tough to compare fairly. Even just with CachyOS and Bazzite above, we see large performance differences. And, as will become clear in this section, there were numerous tests that either showed performance issues inconsistently or simply failed to run at all.

Our best point of comparison between the Steam Machine and BC-250 comes in Cyberpunk 2077. Neither machine is capable of achieving a playable frame rate at 4K, but the fully-optimized BC-250 is just 6.7% behind at 1440p. At 1080p, however, the performance drop is 16.1%, exposing perhaps the biggest performance issue with the BC-250.

Forza Horizon 6 also highlights this issue; the CPU in the BC-250 is very weak. It’s not only weak because it’s using the aging Zen 2 architecture, but it also has only six cores (or eight, if you can enable them). It’s such a hurdle because the clock speed is severely limited to just 3.5 GHz, the same as the PS5. We are completely CPU-bound at 1080p and even 1440p in Forza Horizon 6, and then miraculously, at 4K, the larger GPU of the BC-250 is able to take over and actually beat the Steam Machine.

A system so hamstrung by the CPU makes sense in the context of a console. Sony has a performance target, and it makes sense to opt for a weaker CPU and a more powerful GPU. The CPU can reach that performance target. But in the context of a PC where you’re given more options, resolutions, and performance-enhancing features, the limitation of the CPU becomes clear. There’s an obvious performance wall you’ll run into with the BC-250 that no amount of tweaking can solve.

For a clear view of the CPU limitation here, look at the Forza Horizon 6 benchmark above. Running at 1080p with Low graphics is going to induce a CPU bottleneck in most systems, but the gap between GPU render performance and CPU render performance here is massive. According to the game’s benchmark, we were bottlenecked by the CPU entirely throughout the run.

Even at 4K, you can see that the CPU is still a significant influence on performance in this game, showcasing just how unbalanced this system is. Again, that imbalance makes sense for the PS5, but it’s a critical caveat with the BC-250.

In the gallery above, you can browse our other comparative results to get an idea where the BC-250 lands. I’m going to move past comparisons and look at the BC-250 in isolation. An apples-to-apples comparison really isn’t possible (nor fair) with the BC-250. That’s not only due to the wide performance window depending on your specific board and software stack, but also the clear performance limitations of the BC-250 that require certain workloads that wouldn’t make sense with other machines.

Chief among those limitations is VRAM, and memory in general. You have just 16 GB of GDDR6 for the whole system, which is much less of an issue on the PS5 where there’s careful memory management. As you can see in the chart above, I don’t have 4K results for Spider-Man 2 or Doom: The Dark Ages, and that’s due to a lack of VRAM.

In both cases, the games crashed at 4K, but worse, they wouldn’t start back up. I made the mistake of flipping to 4K before turning FSR on, which locked me out of starting the game until I manually edited the settings file prior to launch. 4K is possible on the BC-250, though it has a very narrow performance window considering the VRAM limitations and CPU performance.

Although FSR Performance mode is ideal for 4K (50% scaling, 1080p input resolution), I opted for Balanced to avoid the CPU bottleneck we can see present in other games (you can even see it here in some games, such as Spider-Man 2). However, turning up the input resolution also puts a greater strain on VRAM. You’ll need time and patience to experiment with these more demanding titles to find the ideal balance of settings on the BC-250; of course, if you’re just playing Silksong, none of this matters much.

The VRAM limitation is also relevant for frame generation. With such a narrow memory footprint, you’ll struggle to use frame generation at 4K if you’re already pushing the VRAM to its limits (those buffered frames need to live somewhere). Thankfully, Lossless Scaling is an option on Linux with LSFG-VK, which doesn’t strain your VRAM like in-engine tools.

An important aspect of performance with the BC-250 is cooling. Most governor profiles cap temperature at 80 degrees for multiple reasons. First, the heatsink of the BC-250 is closed on top. You can sit and rip apart the top of the heatsink for direct airflow with a pair of tweezers, but I left the heatsink on my board intact for now, using two Arctic P12 fans to assist with cooling.

Cooling is all the more important with the 40CU unlock and enabling the extra two Zen 2 cores. Stress-testing in Furmark over a 20-minute run, you can see how quickly the GPU throttles from 1,850 MHz back down to 1,500 MHz as it reaches the temperature cap of 80 degrees.

Remember, however, that Furmark is a stress test. I wasn’t thermal throttling in games, with the CPU and GPU settling around 70 degrees in Doom: The Dark Ages. I’ve included CPU and GPU power in the chart above, as well, and you can see that, combined, they rarely poked over 100W. Keep in mind that we’re exclusively looking at GPU and CPU power here, not whole system power. Whole system power can peak above 225W.

The stock heatsink and dual Arctic P12S do a good job keeping the BC-250 chugging along, even with all of the silicon unlocked and a moderate overclock on the GPU. However, Furmark makes it clear that this current iteration of the system can’t sustain its performance over long periods of time. It would take longer for that drop to show up in the most demanding games, unlike in Furmark, where we can see the drop in just 20 minutes. But I plan on revisiting the cooling solution in this BC-250 build. Maybe I can even raise the temperature limit closer to the 100-degree TJmax.

The bad and the ugly of BC-250 gaming

Although the BC-250 offers compelling performance for the price, I’d be remiss if I didn’t highlight the numerous quirks I ran into while testing. There are performance issues and limitations with the board that you can explain with hardware, and further compatibility issues that you can explain away with Linux. Then, there’s everything else: the weird bugs, quirks, and oddities that pop up when running a largely community-developed software stack on unofficial hardware that wasn’t built for this purpose.

The above image of Resident Evil 9 is a microcosm of what I’m talking about. The RE Engine is remarkably scalable, and that was on full display with my BC-250 testing. Requiem offered excellent performance, even all the way up to 4K. It didn’t feel like I was compromising much of anything with the BC-250. But then I turned on the “hair strands” setting, and the result is what you can see above.

No, it’s not due to the CU unlock, any overclock, or even the overlay you can see. It’s just something with this set of hardware, this software stack, and the hair strands setting in Resident Evil Requiem.

Similarly, in Doom: The Dark Ages, any resolution change I would apply necessitated a restart of the game. Otherwise, I’d get the ghostly streaking you can see in the image above. On one of these reboots, the game suddenly stopped rendering reflective surfaces, resulting in the strange black voids you can see below.

This isn’t a criticism of the BC-250, and certainly not of the remarkable work of the passionate developers and enthusiasts who’ve created fixes and workarounds for the hardware. The fact that this board boots at all and runs games is a minor miracle. The context of what the board is, however, is important.

You will, inevitably, run into strange issues that aren’t documented anywhere. These are two examples, but I also ran into strange problems elsewhere. The widely used MangoHUD overlay, for example, didn’t pull in my GPU’s clock speed correctly after an update, even after ensuring I applied the community fix through the GPU governor to map the reporting correctly. And once you add external devices, compatibility gets even more complex. The board doesn’t have Bluetooth or Wi-Fi, for example, and an external adapter should work (this UGreen Wi-Fi 6 adapter I purchased does), but it’s just as likely that a driver for another adapter doesn’t work or isn’t available.

I don’t have an issue with these quirks, but the BC-250 also isn’t my primary gaming machine. You should just be aware of them. And, if you’re going to invest in a BC-250 build of your own, be ready to encounter some strange issues that you might not be able to easily troubleshoot.

What’s next?

The BC-250 is a project, and like any good project, it isn’t done here. Now that I’ve messed around with the configuration, tried out different distros, got the critical mods working, and measured performance, I want to actually use the BC-250 long-term. Maybe I can find some additional optimization steps I can take, or workarounds for some of the strange issues that I noticed during testing. Regardless of what it is, I’m confident that I’ll mess around with the BC-250 more outside of just playing games on it.

My clear next step is to work on cooling. I kept things conservative here in order to get valid data in a timely manner, but I want to dig deeper into what I can do on the thermal front, not only to raise the temperature limit, but also to push the GPU overclock further. That may involve some CU tinkering, as well; if I can get similar performance at 36 CUs and better thermal headroom, I may be able to push higher overall performance.

This faceoff is particularly interesting because the two chips take very different approaches despite costing about the same price. The Core Ultra 7 270K Plus packs 24 cores and excels in productivity workloads, while also delivering Intel's strongest gaming performance in years. The Ryzen 7 7700X3D, meanwhile, relies on AMD's proven 3D V-Cache technology to deliver gaming performance within 5% of the 7800X3D, although it trails significantly in single- and multi-threaded workloads.

While the Ryzen 7 7700X3D is aimed at gamers looking for X3D performance at a lower price, it faces stiff competition from Intel's aggressively -priced Core Ultra 7 270K Plus, as well as its own sibling, the Ryzen 7 7800X3D, which is often available for only a little more.

So, is AMD's gaming-focused approach enough to beat Intel's well-rounded Arrow Lake Refresh processor, or has Intel finally found the right balance of price and performance to reclaim the mid-range crown? In this six-round faceoff, we compare these two sub-$350 CPUs to find out which one comes out on top.

Features and Specifications: Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7700X3D

Taking a look at the technical specifications, it's clear that each chip adopts a distinct strategy for the mid-range market. While Intel has optimized its architecture for a higher core and thread count along with improved interconnect speeds, AMD continues to leverage its stacked cache to maintain gaming dominance on a long-lived platform.

The Intel Core Ultra 7 270K Plus is based on the original Arrow Lake-S family, utilizing the same microarchitecture that we saw on the 265K, built using TSMC’s 3nm process. It matches the flagship 285K with a 24-core configuration (8 Lion Cove P-cores and 16 Skymont E-cores) and operates with a 125W TDP. The chip can scale to a 250W Maximum Turbo Power (MTP), which is the maximum power limit that this CPU can consume for short periods when running at maximum turbo boost frequencies

Intel has also standardized several enthusiast-level performance tweaks for Arrow Lake Refresh, most notably a 900 MHz increase in die-to-die frequency and a 400 MHz fabric speed boost. It also officially supports faster DDR5-7200, along with 20 lanes of PCIe Gen 5 and 76MB of total cache. Possibly the only major concern with the Core Ultra 7 270K Plus is the LGA 1851 platform. With Intel's next-generation Nova Lake processors expected to launch later this year, the current platform will have a relatively short lifespan, restricting any future upgrade options.

The AMD Ryzen 7 7700X3D serves as a more affordable, lower-clocked variant of the 7800X3D, featuring the same eight Zen 4 cores with a peak boost clock of 4.5 GHz. Its primary architectural strength is its 104MB of total cache including 64MB of vertically stacked L3, which is specifically tuned to eliminate memory latency bottlenecks.

The chip is rated at a 120W TDP with a 162W Package Power Tracking (PPT) limit, though it frequently operates well below these levels during gaming. A major selling point for Team Red is motherboard flexibility as the 7700X3D can be used with existing AM5 motherboards (600 and 800-series chipsets), and AMD has committed to supporting the socket through at least 2029.

⭐ Winner: Intel Core Ultra 7 270K Plus

While AMD offers a more stable platform and superior gaming efficiency thanks to its large cache size, Intel’s hardware package is more comprehensive for the price. The 270K Plus offers triple the core count, significantly higher frequencies, and faster memory support.

Gaming Benchmarks and Performance: Intel Core Ultra 7 270K Plus vs Ryzen 7 9700X

To evaluate the gaming performance of these two chips, we conducted our testing at 1080p resolution using an Nvidia GeForce RTX 5090. We used this setup to prevent the GPU from becoming the bottleneck, allowing each CPU's gaming performance to shine through. By removing the graphics card as a limiting factor, we can see exactly how the 7700X3D’s 3D V-Cache stacks up against the higher core count and clock speeds of the 270K Plus.

We used identical systems for testing. You can read more about our full system configuration in our Ryzen 7 7700X3D review and Core Ultra 7 270K Plus review.

Across a comprehensive 16-game geomean, the AMD Ryzen 7 7700X3D holds its position as the superior gaming chip, delivering a 174.3 FPS average. This represents a 5.3% performance lead over the 270K Plus, which produced a 165.6 FPS average. While Intel's latest flagship has significantly narrowed the gap compared to the original Arrow Lake silicon, AMD’s 3D V-cache architecture continues to provide a higher performance in the majority of tested titles.

Even when examining the 1% low frame rates, which is the perceived smoothness while gaming, the 7700X3D maintains a slight edge with a 118 FPS average compared to the 115 FPS from the 270K Plus.

AMD's stacked cache gives the 7700X3D a clear edge in games that are sensitive to memory latency. For instance, in Minecraft RT, the 7700X3D delivered 144.5 FPS average, dwarfing the 89.7 FPS produced by the 270K Plus for a massive 61.1% performance advantage. We see similarly dominant leads for AMD in F1 2024, where it leads by 30.4% (201.8 vs. 154.7 FPS), and in Final Fantasy XIV, where it holds a 21.7% lead (177.7 vs. 146.0 FPS). Even in esports titles like Counter-Strike 2, the 7700X3D maintains a comfortable 7.3% lead with 707.9 FPS over Intel’s 660.0 FPS.

Intel’s Core Ultra 7 270K Plus is no slouch, however, as it leverages its higher boost frequencies and the new iBOT runtime translation layer to take the lead in several games. Intel manages a win in Hogwarts Legacy, producing 135.3 FPS for a 9.1% advantage over AMD’s 124.0 FPS, a result directly attributed to iBOT. Team Blue also edges out victories in The Last of Us Part 1 with 175.6 FPS (6% lead) and Spider-Man 2 with 207.2 FPS (5% lead).

In many other modern titles, the two processors are essentially locked in a dead heat. For example, in DOOM: The Dark Ages, the 270K Plus delivers 200.4 FPS while the 7700X3D is right beside it at 200.0 FPS. The story is similar in Flight Simulator 24, where Intel's 120.5 FPS and AMD's 114.9 FPS result in a negligible difference during actual gameplay. These results indicate that while the 7700X3D is the more consistent gaming choice, especially in cache-heavy titles, the 270K Plus is a highly competitive gaming chip that delivers impressive results for its $300 price tag.

The biggest difference between these two architectures is noticeable when analyzing power consumption and efficiency metrics. Based on our 16-game CPU power geomean, the 7700X3D is clearly way more efficient, drawing an average of just 60.9 watts while gaming. In comparison, the 270K Plus consumed 107.3 watts on average, which is a significant 76.2% increase in power. This results in a massive gap in gaming efficiency where the 7700X3D delivers 2.86 FPS per watt, making it roughly 85.7% more efficient than the 270K Plus, which trails at 1.54 FPS per watt.

Intel does take the lead when it comes to raw frequency, with the 270K Plus maintaining an impressive 5,247 MHz average clock speed across our gaming suite. This is over 700 MHz faster than the 7700X3D, which averaged 4,505 MHz. Despite the much higher clocks and power draw of the Intel chip, its thermal management remains surprisingly competitive when paired with the right cooling solution. The 270K Plus averaged 59°C during gaming, though it still runs warmer than the 7700X3D, which stays at a cooler 55°C average.

When factoring in the cost of the silicon, the competition for the best bang for your buck is incredibly tight. The Ryzen 7 7700X3D provides a value of 0.53 FPS per dollar, narrowly edging out the Core Ultra 7 270K Plus, which sits at 0.52 FPS per dollar.

⭐ Winner: AMD Ryzen 7 7700X3D

The 7700X3D’s overall higher average frame rates and excellent efficiency makes it the more attractive option for a dedicated gaming build. Since raw gaming performance, efficiency, and thermals all favor Team Red in this category, the Ryzen 7 7700X3D takes the win for this round.

Productivity Performance: Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7700X3D

AMD's X3D processors are widely known for their gaming prowess. However, they have historically struggled to maintain the same level of dominance in productivity performance. The Ryzen 7 7700X3D is no exception, as its cache-focused architecture and lower clock speeds result in relatively weak single- and multi-threaded results.

In contrast, Intel has managed to maintain its standing in the productivity segment even when its gaming performance faced challenges. The 270K Plus is a productivity workhorse that is capable of delivering excellent productivity performance making it a standout choice for users who need a balanced system for both work and play.

In multi-threaded workloads, the performance gap between these two processors is pretty massive, primarily driven by Intel's higher core count advantage. Across our multi-threaded performance ranking geomean, the Intel chip delivers a score of 626 compared to the 272 produced by the 7700X3D, representing a staggering 130% performance lead for Team Blue. This level of parallel processing power places the 270K Plus in a completely different performance tier, making it more comparable to much more expensive flagship processors.

Individual benchmarks further highlight this lopsided victory for Intel across various professional tasks. In Cinebench 2024’s multi-core test, the 270K Plus scores 2,509 points, which is roughly 135.6% faster than the 1,065 points achieved by the 7700X3D. This trend continues in POV-Ray, where Intel leads by approximately 169.5% (15,697 vs. 5,823 PPS), and in V-Ray 6, where it maintains a massive 129.5% advantage (45,016 vs. 19,615). Even in intensive video encoding tasks via HandBrake x265, the Intel chip more than doubles the performance of its AMD rival, delivering 29.9 FPS compared to just 14.3 FPS for the 7700X3D. These results demonstrate that for heavy rendering or data-crunching workloads, Intel is the undisputed leader in this price bracket.

A similar trend can be seen when examining single-threaded performance, a crucial metric for general system responsiveness and applications that do not scale across multiple cores. In our single-threaded geomean, the 270K Plus scores 278 points, which is 42.5% faster than the 195 achieved by the 7700X3D. Specific tests like Cinebench 2024’s single-core benchmark show the 270K Plus maintaining a 39.7% lead over the 7700X3D (145 vs. 103.8 points), while the lead grows to a staggering 82.6% in POV-Ray’s single-core test (1,138 vs. 623 PPS). Even in audio encoding, the Intel chip finishes the Lame Extended task in 68.61 seconds, whereas the 7700X3D trails at 91.53 seconds.

⭐Winner: Intel Core Ultra 7 270K Plus

Ultimately, the 270K Plus simply blows the competition out of the water when it comes to productivity. While the 7700X3D is a highly efficient and specialized CPU for gaming, it cannot match the raw horsepower that Intel offers. For any user whose daily routine involves video editing, 3D rendering, or heavy multitasking, the 270K Plus is the obvious choice and should be the clear favorite for a multi-purpose system.

Overclocking: Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7700X3D

While both manufacturers provide tools to squeeze extra performance from their silicon, the Intel Core Ultra 7 270K Plus is built as an enthusiast-friendly, fully unlocked processor. The AMD Ryzen 7 7700X3D, on the other hand, is a more restricted processor designed primarily for out-of-the-box gaming efficiency.

For the Intel Core Ultra 7 270K Plus, overclocking is a centerpiece of the experience. As a fully unlocked K-series SKU, it offers users granular control over per-core voltages, power limits, and clock speeds via an unlocked multiplier. While these changes can be done by entering the BIOS, one can also download the Intel XTU (Extreme Tuning Utility) software to fine tune the CPU.

One of the most significant changes seen on Arrow Lake Refresh is that Intel has standardized several high-end performance tweaks including the 900MHz die-to-die frequency bump and the 400MHz fabric speed increase. This means users receive enthusiast-level interconnect performance without necessarily needing a premium Z-series motherboard. That said, Z890 boards are still necessary as they offer sophisticated tools for manual tuning and pushing the CPU to its limits.

In contrast, the AMD Ryzen 7 7700X3D follows the same path as previous Zen 4 X3D processors including a locked multiplier, which prevents traditional manual overclocking. Users are instead limited to automated and semi-automated features like Precision Boost Overdrive 2 (PBO2) and Curve Optimizer. PBO2 allows the CPU to dynamically adjust its frequencies based on available power and thermal headroom, while Curve Optimizer enables more advanced fine-tune voltage offsets for each of the eight Zen 4 cores.

While these tools can lead to sustained higher boost clocks, the sensitive nature of the 3D V-Cache stack leads to thermal challenges that limit frequency headroom. Thus, overclocking gains on the 7700X3D are often minimal compared to the flexibility offered by the Intel chip.

⭐Winner: Intel Core Ultra 7 270K Plus

The 270K Plus is a far more overclocking-friendly product. It offers a vast suite of features that AMD simply cannot match due to its architectural restrictions and locked multiplier.

Power Consumption, Efficiency, and Cooling: Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7700X3D

This is a crucial section of the faceoff as it highlights the most significant architectural divide between these two processors. While the 270K Plus prioritizes raw performance throughput, the 7700X3D focuses on extreme efficiency. This is immediately evident in their official power ratings where Intel specifies a 125W TDP with a massive 250W Maximum Turbo Power (MTP), while AMD utilizes a 120W TDP with a 162W Package Power Tracking (PPT) limit.

In synthethic multi-threaded workloads, the difference is quite evident. While running HandBrake x265 10-bit encode, the 270K Plus consumes an average of 226W, which is more than triple the 72W required by the Ryzen 7 7700X3D. A similar trend appears in VP9 encoding, where Intel draws 184W compared to AMD’s efficient 65W. Even in lighter tasks like single-threaded y-cruncher AVX workloads, the Intel chip requires 55W while the AMD chip stays at a modest 32W. Idle power consumption also favors Team Red where the 7700X3D idles at 19W and draws 22W during YouTube playback, whereas the 270K Plus sits higher at 29W and 38W, respectively.

When we translate these power figures into efficiency metrics, AMD’s lead tends to stay ahead in most traditional benchmarks. In Cinebench 2024, the 7700X3D produces 14.4 points per watt, significantly outperforming the 10.4 points per watt from the 270K Plus. In HandBrake x265, AMD achieves a superior efficiency rating of 5W per FPS , while Intel trails at 7.56W per FPS.

However, the efficiency scatter plots provide a more nuanced look at how performance scaling impacts total energy consumption. In the Blender Classroom scatterplot, the 270K Plus delivers a much higher performance of roughly 135 samples per minute, but it consumes over 33,000 kJ of task energy. In contrast, the 7700X3D finishes with only around 60 samples per minute but uses significantly less total energy at roughly 4,500 kJ.

The HandBrake x265 scatterplot shows a similar trade-off where Intel reaches 30 FPS but consumes over 55 kJ, while AMD provides 14 FPS at roughly 36 kJ. Interestingly, the Linpack scatterplot shows the 270K Plus reaching over 850 GFLOPS with roughly 11WHr of energy, making it slightly more efficient in terms of performance-per-energy than the 7700X3D, which delivers roughly 340 GFLOPS for 10Whr.

From a cooling perspective, the 250W MTP on the 270K Plus means that it requires a robust cooling solution, like a 360mm AIO, to avoid thermal throttling during extended all-core loads. The 7700X3D is far easier to manage, remaining remarkably efficient and manageable with a mid-range air cooler. While the 270K Plus delivers class leading performance, it does so by significantly compromising efficiency in heavy workloads compared to AMD.

⭐Winner: AMD Ryzen 7 7700X3D

Pricing: Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7700X3D

The Intel Core Ultra 7 270K Plus was initially introduced with a justified $300 price tag, but you'll now find it between $300 and $320. The AMD Ryzen 7 7700X3D made its debut just last month and carries a $330 MSRP, placing it right under the 7800X3D. Essentially, there is not a major difference when it comes to the chips themselves, however, the true financial divide becomes apparent when we look at the total platform cost.

The Core Ultra 7 270K Plus demands a more substantial investment to reach its full potential. While one can opt for a B860 motherboard, a compatible Z890 motherboard is necessary in order to access premium overclocking and tuning features. These typically start around $150-$200, and can go as high as $600 packed with premium features. To handle the processor′s 250W Maximum Turbo Power(MTP) during heavy productivity tasks, a high−end 360mm AIO liquid cooler ($100-$150) or a dual-tower air-cooler ($50-$100) is recommended.

In contrast, the AMD Ryzen 7 7700X3D offers a much more budget-friendly entry point. A solid B650 motherboard can be found for around $120-$150 and the chip can be effectively cooled with a modest $50 air cooler.

Unfortunately, both platforms are held back by the ongoing global shortage of memory, which has caused prices to shoot up significantly. Since both CPUs require DDR5 RAM, builders are stuck in a RAMpocalypse where a basic 32GB kit costs over $400. This extra cost is bad news for both sides as it cancels out any price advantage.

Overall, the lower upfront platform cost makes the 7700X3D a slightly more attractive option for users prioritizing their budget. Furthermore, the AM5 socket offers a clear upgrade path through at least 2029, ensuring that your motherboard investment remains viable for future CPU generations. That is not the case with Intel, as the LGA 1851 socket is expected to be replaced once the next-generation of Nova Lake CPUs arrive.

⭐ Winner: AMD Ryzen 7 7700X3D

Ultimately, while both the chips are available at a very similar price range, the AMD Ryzen 7 7700X3D is the smarter financial play for a majority of users. Its significantly lower platform costs, superior gaming efficiency, and guaranteed platform longevity provide a level of value that the more power-hungry Intel refresh cannot quite match.

Bottom Line: Intel Core Ultra 7 270K Plus vs AMD Ryzen 7 7700X3D

The Ryzen 7 7700X3D and Core Ultra 7 270K Plus are designed for very different users, even though they sit in roughly the same price range. Like previous X3D chips, AMD's latest processor is built with gaming in mind, using its large 3D V-Cache to deliver excellent frame rates while keeping power consumption impressively low. Intel, on the other hand, combines a 24-core design with higher clock speeds and platform improvements to create one of the strongest productivity processors in its class without giving up much gaming performance.

The final score is tied at three rounds apiece. AMD comfortably wins gaming performance, power efficiency, and platform value, three factors that matter a lot for buyers shopping in this price segment. The AM5 platform also provides a significantly longer upgrade path, allowing users to drop in future processors without replacing their motherboard.

Intel, however, deserves recognition for what it has accomplished with the 270K Plus. It delivers outstanding single- and multi-threaded performance, offers a fully unlocked overclocking experience, and narrows the gaming gap to just a few percentage points in many modern titles. If your workload includes video editing, software development, 3D rendering, or other heavily threaded applications, the 270K Plus is the obvious choice and justifies its higher power draw.

For everyone else, the Ryzen 7 7700X3D is the more compelling CPU. It offers faster gaming performance, exceptional efficiency, lower platform costs, and a future-proof AM5 ecosystem. Unless your workload regularly extends beyond gaming into demanding productivity applications, AMD's latest X3D chip is the easier processor to recommend.

⭐ Winner: Tie

More CPU Faceoffs

• AMD Ryzen 9 9950X3D2 vs Ryzen 9 9950X3D
• Intel Core i5-14400 vs AMD Ryzen 5 7600X
• AMD Ryzen 7 9850X3D vs Intel Core i9-14900K
• AMD Ryzen 7 9850X3D vs Ryzen 7 9800X3D

That’s not a new story, as AMD has continually pointed to the late stage of the console lifecycle as the driving force behind declining gaming revenue. CEO Lisa Su, however, also says revenue is declining due to rising component costs. “Gaming graphics revenue also declined year-over-year as higher industry-wide component costs contributed to higher graphics card prices and weighed on overall demand,” Su said.

Earlier this year, AMD said that it expected a 20% decline in gaming revenue in the second half of the year, which is reaffirmed in its Q2 2026 earnings call. “[We expect] a modest decline in our client gaming segment, with slight growth in client offset by strong double-digit decline in gaming,” said AMD's Jean Hu, executive vice president and Chief Financial Officer.

Despite these declines, Su says she's 'optimistic' about the future of AMD's client CPU business, particularly when it comes to AI PCs. "I think AI PCs will become more important as it relates to our client business in 2026. I think our first half performance has been very strong, and although we are expecting that the market will decline in the second half, the market's actually held up, you know, better than you know most people would have thought," said Su.

Despite a large decline in gaming revenue, AMD’s overall client business (including gaming) brought in $3.8 billion, up 6% year-over-year. Su says this is due to strong demand for Ryzen CPUs, but also increased market share for Ryzen PRO chips. “Commercial adoption continued to expand in the quarter, with Ryzen PRO sales growing more than 50% year over year as we closed new wins with large healthcare, technology, automotive, and financial services companies,” Su said.

Declining gaming revenue has been massively offset by increasing data center revenue, unsurprisingly. AMD’s data center revenue is up 107% year-over-year, representing $6.7 billion of revenue. AMD expects that the data center will continue growing the business in the next quarter, as well.

“We expect [Q3] revenue to be approximately $13 billion plus or minus $300 million, at the middle point of our guidance, revenue is expected to be up 41% year over year, driven by very strong double-digit growth in our data center segment,” said Hu.

Surprisingly, AMD also says it expected “strong” double-digit growth in its embedded business, which has traditionally represented the smallest slice of AMD’s revenue pie. In Q2, embedded revenue was $977 million, which is up 19% year-over-year, which AMD attributes to increased demand (perhaps in the realm of physical AI). That means gaming is the smallest part of AMD's revenue pie for the moment, and that position seems likely to continue until we see some kind of new discrete GPU or console hardware from the division that might spur upgrades.

It’s time to look back, not just on Tom’s Hardware’s role in the world of CPUs over the past 30 years, but on the broader CPU market. Earth-shattering releases at a given time turned out to be footnotes in hindsight, and bubbling competition that seemed like no real threat turned into seismic shifts in the AMD vs. Intel battle. Hopefully, along the way, we’ll see some reflections of what’s going on in the world of CPUs today.

This is a celebration of 30 years of CPU coverage here at Tom’s Hardware, a history lesson about how the AMD and Intel of yesteryear arrived at the positions they’re in today, and a retrospective of the CPUs that stood the test of time against the refreshes that faded into obscurity just as soon as they were released. We hope you enjoy it.

Ruffling feathers from day one (1996 - 1998)

Ruffling feathers from day one (1996 - 1998)

Like much of the early internet, the true beginnings of Tom’s Hardware (or, rather, Tom’s Hardware Guide) is difficult to pin down, but the earliest archived article concerns the SoftMenu BIOS, which allowed you to change CPU settings through software rather than physical jumper cables; it should go without saying that this article was from July 1, 1996. There were articles before this one, with the original domain of sysdoc.pair.com going online in February 1996.

Tom’s Hardware gained a lot of traction a year later with a CPU review: a look at the Intel Pentium II ‘Klamath’ processor.

Tom’s Hardware’s founder, Thomas Pabst, had discussed CPUs previously, like in a dense CPU guide published the same year, but the Intel Pentium II review marked a turning point. It was published March 1, 1997, more than two months before Intel launched the Pentium II 80522 (or Klamath). It was not positive: “...since I wouldn't eat [Mad Cow Disease] infected beef, I wouldn't be interested in risking an infection of my computer with this CPU either,” wrote Pabst.

The CPU wasn’t officially available, but Pabst was able to test the chip with a pre-release unit shared by two German magazines: C'T Magazin fur Computertechnik and PC Professionell. Following the publication of all three reviews, Intel targeted the two German magazines, threatening to withhold advertising dollars and take legal action against them. Pabst publicized this fiasco and was contacted by The New York Times, which also covered the story. Intel backed down. Pabst later wrote that “...this 'David against Goliath'-incident made Tom's Hardware Guide very famous.”

In the following months, Pabst focused on CPUs quite a bit, breaking from the performance-tuning guides and general enthusiast information the website had previously published. The situation with Intel made Pabst a “secret star,” so much so that AMD not only offered Pabst a free review unit of the upcoming K6, but also apparently offered to cover legal fees should Intel pursue the situation further.

Although the unofficial review of the Pentium II predates it, the first review that looks most like the technical reviews that have been published on Tom’s Hardware for decades came in April 1997 with a review of the AMD K6, a CPU that established AMD as a true competitor to Intel, claiming Pentium II-like speeds for less money.

The performance wasn’t quite on the level of a Pentium II, most notably because the K6 originally arrived at 233 MHz, when 266 MHz was promised. But it was competitive, particularly when compared to the Pentium II 233, and much cheaper. Pabst concluded the review, “All in all I'm sure that this CPU will be very successful… Whoever is contemplating the purchase of an Intel Pentium or Pentium MMX CPU can forget about this now. The AMD K6 is faster and cheaper.”

Later that same month, Pabst published a proper review of the Pentium II, now using finalized silicon. In DOS games and the 32-bit Windows NT, Pentium IIs showed a commanding lead over the K6, though the battle on Windows 95 was closer. “The world is back to normal. Intel’s managers can sleep quietly again. The Pentium II shows that Intel is still the leader in the CPU market,” the review concluded.

A review of the Cyrix/IBM 6x86MX went live in 1997, as well, showing competitive performance with the K6, but much of the rest of the year was focused elsewhere after the Pentium II/K6 showdown. Looking back, it’s interesting to note the dynamic at play between AMD and Intel, with Pabst using a running gag of referring to Intel as the “Empire,” and competitors like AMD and Cynix as “Rebels.” Although Intel backed away from further action on Pentium II, it wouldn’t be the last time Pabst and Intel butted heads.

The Pentium III incident (1999 - 2001)

The Pentium III incident (1999 - 2001)

The competitive performance of K6 brought AMD into focus as the rival to Intel, which Pabst often described as “almost-monopolistic” at the time. However, AMD really put itself on the map with the release of K7, or as it’s better known, Athlon, in June 1999. In our review, we described it as “a milestone in the whole processor scene,” due in part to AMD’s unique three-way instruction decoder, which allows instructions of variable complexity through all three lanes. Intel’s P6 architecture in Pentium III also had a three-way instruction decoder, but they were segmented based on the complexity of the instruction.

We’re still firmly in the single-core era of CPUs here, so an architectural divergence like this, even before testing, represented a goldmine of speed gains. Rather than releasing several variants of silicon sliced up in different ways, as we see with modern CPUs, Intel and AMD refined their chips and released new versions with faster clock speeds. AMD released the Athlon 600 as the fastest chip in the range at 600 MHz first, but less than two months later, it introduced the Athlon 650, and two months after that, the Athlon 700.

Intel was caught off guard and quickly introduced the Pentium III 600 to counter AMD. The original Pentium III range, codenamed Katmai, was largely a refinement of Pentium II ‘Deschutes’ chips; both were built using a 250nm node, in fact. In our original Athlon review, we noted instability with the Pentium III 600, suggesting the architecture couldn’t handle such high clock speeds.

The original ‘Katmai’ range of Pentium IIIs was short-lived, and later in 1999, Intel introduced a new revision called Coppermine. Coppermine introduced an on-chip L2 cache for Intel, and it was built on a 180nm node, allowing Intel to go from 9.5 million transistors on Katmai to 28 million on Coppermine, as well as push clock speeds up to 733 MHz with the initial range, squeezing out a minor increase over the Athlon 700. Less than two months later, AMD introduced Orion, or Athlon Model 2, which also used a 180nm node and clocked up to 750 MHz. This back and forth of minor clock speed improvements is going to stick with us for at least another half of a decade, so strap in.

The goal, of course, was the first 1 GHz CPU, a milestone that AMD claimed for itself in March 2000 with the introduction of the Athlon 1000. Intel had shown a 1 GHz Pentium III, but AMD released a 1 GHz Athlon first. Intel followed shortly after with its first CPU to hit the gigahertz milestone. Intel didn’t want to place second again, so it went to work on the Pentium III 1.13 GHz, which it introduced in July.

Pentium II made Tom’s Hardware a name in the PC industry, but the Pentium III 1.13 GHz gave it a name among enthusiasts. Our review found that the processor wasn’t stable at 1.13 GHz, and less than a month after introducing the chip, Intel recalled it. The Pentium 1.13 GHz would eventually come back in 2001, but at the pace of CPU innovation at the time, even a six-month delay was detrimental.

In June 2000, AMD introduced a refinement of Athlon, codenamed Thunderbird, and days before Intel’s recall, introduced the Athlon 1100. AMD remained uncontested throughout the rest of the year, pushing Thunderbird up to 1.2 GHz. Intel was closing the curtain on Pentium III and trying to move attention away from Athlon toward its upcoming Pentium 4 range.

Going for gigahertz (2001 - 2004)

Going for gigahertz (2001 - 2004)

Intel was struggling to keep pace with Athlon, but work was going on behind the scenes on the new NetBurst microarchitecture, which was set to become the successor to P6. It was introduced to the world with the Willamette core inside the first Pentium 4. Like most major microarchitecture shifts we’ve seen from Intel and AMD, NetBurst wasn’t an immediate success. However, Pabst noted in our review: “I am certain that Intel will deliver very fast Pentium 4 processors very soon. Intel has finally won back the ability to make AMD's life a lot harder.”

Intel pushed P6 down to a 130nm node with the release of Tualatin Pentium III chips, but the focus was on Pentium 4 and optimizing NetBurst. In March, AMD introduced the Athlon 1333, but Team Red was losing steam. Intel had already released a Pentium 4 1.5 GHz, and it introduced the Pentium 4 1.7 GHz (along with a price cut to the range) in April 2001. With a new microarchitecture seemingly bursting with potential, it was only a matter of time before Intel made its way back to the top.

That came in August, when Intel planted its flag on the 2 GHz milestone with Pentium 4, and beat out AMD’s fastest Athlon chip. AMD didn’t like that. Later in the year, in October, AMD introduced its Athlon XP range, and with it came a sneaky switch in marketing strategy. Rather than include the clock speed as part of the processor name (i.e., Athlon 1333), AMD started using model names. No, the Athlon XP 1500+ wasn’t clocked at 1.5 GHz; it was clocked at 1.3 GHz. This nomenclature climbed all the way to the top, with the Athlon XP 2100+, which was not 2.1 GHz, but rather 1.7 GHz.

At the time, AMD described this shift as an alignment with what it was able to deliver with Athlon. During this era, we start to see more significant architecture divergences between Intel and AMD, so much so that like-for-like clocks could result in vastly different performance. That’s what we found in our Athlon XP review, in fact, with the Athlon XP 1800+ (clocked at 1,533 MHz) contesting the Pentium 4 2 GHz. Still, it’s not hard to see what AMD was trying to do. In January 2002, Intel introduced its Northwood core for Pentium 4, which could clock up to 2.2 GHz. AMD wasn’t able to break the 2 GHz barrier yet, but it used product names to suggest it had.

Names weren’t enough. By the middle of 2002, Northwood had picked up steam and could clock as high as 2.8 GHz; even the 2.4 GHz Pentium 4 was able to beat AMD’s fastest Athlon XP 2100+ across our benchmarks. AMD was working on the Thoroughbred revision of Athlon XP, including a node shrink down to 130nm to match Intel’s new Northwood core, which it trickled out through 2002. By the end of the year, AMD had shown the Athlon XP 2800+ as a rival to the Pentium 4 2.8 GHz.

Despite lagging in clocks, AMD had become quite popular during this time due to the competitive performance of Athlon XP and (most importantly) a lower price point than the expensive Pentium 4s. Still, there was a stalemate in technology between Intel and AMD, and Intel would break it in November 2002 with the Pentium 4 3.06 GHz.

It was the first consumer CPU to clock to 3 GHz out of the box, adding another notch to Intel’s belt, but it was also the first CPU to bring Intel’s long-standing simultaneous multithreading implementation, called Hyper-Threading, to market. With higher clocks and two threads to play with, the Pentium 4 3.06 GHz cemented Intel at the top of the performance charts, beating out AMD’s fastest Athlon XP and even a 3.6 GHz Pentium 4 (no Hyper-Threading) in some benchmarks. Tom’s Hardware actually benchmarked the Pentium 4 3.6 GHz nearly a year before it was available, and had data in time to compare to the Pentium 4 3.06 GHz with Hyper-Threading.

More cores, more fun (2004 - 2007)

More cores, more fun (2004 - 2007)

The clock speed battle from the late 90s and early 2000s was starting to fall apart. AMD had demonstrated that performance was more than peak clocks with Athlon XP, and Intel was pushing ahead with Hyper-Threading to get more work done simultaneously each clock cycle. The first half of 2003 was dull in the world of CPUs as Intel worked on its Prescott core and AMD mulled over “ClawHammer,” which would eventually become Athlon 64.

Athlon 64 would be the first to bring AMD’s x86-64 ISA extension (called AMD64) to the desktop market (it previously showed up in Opteron). It rolled out in September 2003, and just a week before release, Intel launched the Pentium 4 Extreme Edition, which was widely considered a panic switch in response to Athlon 64 while Prescott was still under wraps. Although Intel maintained the performance crown with P4 Extreme against the Athlon 64 FX-51, it did so at a much higher price.

Still, this late-stage battle as Intel and AMD moved toward a 90nm node was important. Intel started marketing expensive, high-performance processors directly to enthusiasts in a market that would eventually get the HEDT, or high-end desktop, name. Meanwhile, AMD designated some Athlon XP chips with the FX name, which signaled high-performance chips with unlocked multipliers.

Intel was first to get down to 90nm with its highly anticipated Prescott core, though it arrived with a whimper. It clocked slower, allowing the FX-51 to remain competitive and Northwood chips to remain at the top of the charts in our review. There, our reviewer Patrick Schmid wrote: “In our opinion, Intel today does not care about Prescott as a processor, but as a marketing instrument. It is fast enough, which is mainly what counts, and since the 90 nm production process yields cheap processors in vast quantities, the Santa Clara-based company gains new flexibility.”

AMD made it down to 90nm later in the year with the FX-55. The FX-55 allowed AMD to close the gap with higher-clocked Pentium 4s, as we found in our review. However, it was becoming clear that the chase for higher clocks wouldn’t be enough. In our review, Schmid wrote in 2004: “Today, the performance gap between the fastest and the slowest processors in our benchmark charts is rather small.”

Intel fired back at the beginning of 2005 with Prescott 2M, a minor revision to the Prescott core with 64-bit ISA extensions, and released the Pentium 4 Extreme Edition 3.73. The tide shift came in the summer of 2005 when AMD launched the Athlon 64 X2, built on a 90nm node, and featuring two cores on the same package.

Intel had released its double-core Pentium D just weeks earlier, which was also a dual-core chip, but the design forced Intel to compromise clock speeds — an important spec given how few applications could actually leverage a dual-core chip in consumer software. AMD ultimately won the battle with the Athlon 64 X2, largely due to the fact that it could keep pace with single-core Athlons in most benchmarks, as you can see in our early preview of the chip.

AMD released more Athlon 64 X2 chips throughout 2005, while Intel rolled out various chips under different Pentium brands, including the Smithfield core on Pentium D. Intel built out the range into 2006, using another node shrink down to 65nm with the Presler core, which was the final revision under the Pentium brand. In January 2006, Tom’s Hardware first reported that Intel planned to drop the Pentium brand, a name that it had kept for over a decade.

In its place? The new Core microarchitecture, finally moving on from NetBurst, which had been plagued with thermal issues as clocks climbed. Built out of Intel’s work in mobile chips, the Core 2 Duo was Intel’s first proper dual-core processor — the “double-core” Pentium D was just two Pentium dies fused together. It was a tide shift.

As we found in our Core 2 Duo review, the base E6600 and E6700 often beat, or at least matched, AMD’s Athlon 64 FX-62, while the supercharged Core 2 Extreme X6800 established a new performance tier at the high-end. Further, it did so at reasonable power levels, finally taking the efficiency fight back to AMD.

Intel doubled down, literally, with the Core 2 Extreme QX6700 at the end of 2006. Although Intel had introduced a four-thread processor previously, the Core 2 Extreme QX6700 was the first CPU with four cores to hit the market. It took two dual-core dies from Core 2 Duo and put them together on a single package. AMD brute-forced a quad-core with the 4x4 platform and dual Athlon 64 FX-70 chips, but there was a major tradeoff in cost, thermals, and power demands.

The coast of Nehalem (2007 - 2010)

The coast of Nehalem (2007 - 2010)

For the first time in the early aughts, Intel was firmly in the driver’s seat for enthusiasts. AMD drummed up some interest with its dual-chip Athlon 64 systems, and it finally moved down to a 65nm node at the start of 2007. But Intel was on a tear, and it would continue its momentum for years to come.

That started by formalizing the success of the Core microarchitecture. Instead of squeezing everything out of a microarchitecture for several generations, as it had done with P6 and NetBurst, Intel transitioned to its well-known tick-tock cycle. First, there’s a node shrink on its existing architecture, then there’s a new microarchitecture on that node, and the cycle continues. Core was the tock at 65nm, and it would move down to Penryn revision at 45nm later in 2007.

AMD struggled to keep pace. After introducing its AM2 socket in 2006, it continued refining the Athlon 64 X2 lineup with a new 65nm node, but Intel was firmly in the lead, forcing AMD to slash prices and settle into the market as a budget-focused alternative to the shiny Core 2 Duos. AMD doubled down in this area with its K8 microarchitecture, releasing a series of efficient “BE” series chips throughout the back half of 2007, which we found were excellent for efficiency and price-to-performance in our review.

Team Red had its chips on K10, and specifically, the new Phenom brand that it made noise about throughout 2006 and 2007. The Phenom X4 series launched in November, featuring the first true quad-core design; that is, using a monolithic die as opposed to MCM like Intel’s Core 2 Quad. Unlike Intel’s Core 2 Extreme, which was reserved for only the most entrenched enthusiasts, AMD targeted midrange builders.

In our Phenom 9700 review, we found that Intel still maintained the performance crown, but AMD offered a cheaper quad-core and offered identical price-to-performance at release. Further, AMD offered support for the chips on both the AM2 socket and the new AM2+ socket, starting a trend of socket longevity that we can still see in action today. With cheaper quad-core CPUs and less upgrade cost, AMD focused less on battling Intel at the high-end and more on delivering in the midrange.

Intel continued to release more Core 2 Duo and Core 2 Quad models throughout 2008, but behind the scenes, it was working on its next major architectural shift: Nehalem. AMD built out its Phenom line, meanwhile, introducing a tri-core variant, as well as several “Black Edition” models that featured peak clocks and unlocked multipliers for enthusiasts, likely in a bid to grab some attention from Intel’s Extreme lineup.

Nealem came onto the scene in late 2008 in the form of Bloomfield chips. They required an entirely new platform and DDR3 memory, but also promised entirely new performance benefits. Intel managed to create a true quad-core chip with Bloomfield, and one that enthusiasts could actually afford, with the range going down as low as $280. Further, it reintroduced Hyper-Threading after ditching the technology during the early dual-core days, giving enthusiasts eight threads to play with.

The spread between Intel and AMD grew wider. In our review of the Core i7-965 Extreme, we found it was 64% faster than AMD’s fastest chip at the time. Despite offering compelling products at competitive prices, AMD was falling further behind. Intel was far ahead; it had a die shrink ahead, and the legendary Sandy Bridge microarchitecture was waiting in the wings to pick up the tick-tock cycle once again.

AMD bit back in early 2009 with Phenom II (along with Athlon II), finally moving down to 45nm. It brought Team Red back up in the rankings, but Bloomfield still held onto top-end performance. In our original Phemon II review, we found that Intel was about 22% ahead of AMD at the high-end, but AMD offered compelling performance given the platform costs. Sounds familiar.

With tight integration of design and manufacturing under one roof, Intel was moving faster than AMD on process improvements, allowing it to keep a consistent lead in performance, particularly at the high-end. Intel had its Lynnfield chips at 45nm, as well as Clarkdale chips at 32nm, but the 32nm successor to Bloomfield came in early 2010, known as Gulftown. And with Gulftown, Intel could claim a multi-core milestone on a monolithic die for the first time with the six-core Core i7-980X.

At $1,000, the Core i7-980X still only appealed to a small number of enthusiasts. AMD helped fill the gap later in the year with the Phenom X6 launch, introducing six-core models of its own. AMD couldn’t match the 980X at the high-end, but Phenom represented an affordable entry-point to six-core chips if you ran heavily-threaded workloads, as we noted in our review. Still, Phenom X6 was AMD trying to keep pace with Intel. Behind the scenes, AMD was working on a new microarchitecture called Bulldozer, which was built from the ground up for a new generation of chips and finally allowed AMD to move down to 32nm.

Bulldozer bulldozes world records, thermals (2010 - 2013)

Bulldozer bulldozes world records, thermals (2010 - 2013)

It’s easy to pick on Bulldozer in hindsight, but leading up to release, the anticipation was palpable. K8 was a smash success, and K10 built on that success, but Bulldozer was built from the ground up, presumably for the position AMD had found itself in the market. Nearly a year before Bulldozer showed up on the market, however, Intel introduced an architecture that still resonates among enthusiasts today — Sandy Bridge.

Intel unified its product stack with Sandy Bridge and set much of the foundation for the company’s releases through Raptor Lake Refresh. Instead of two broad ranges, Intel placed most of the Sandy Bridge lineup on a single socket. It also brought integrated graphics to all the chips in the lineup, at least in the initial range (some later releases cut the iGPU), as well as unlocked the multiplier on some non-Extreme SKUs to compete with AMD’s unlocked Black Edition chips.

It was a smash success. Reviewing four Sandy Bridge chips, our reviewer Chris Angelini wrote: “Existing Lynnfield- and Clarkdale-based processors already offer strong performance compared to AMD’s lineup. Significant gains, clock-for-clock, compound in the face of notable frequency increases across the board (thanks to a mature 32 nm process), giving Sandy Bridge an even more commanding position.”

Raw performance improvements are one thing, but Sandy Bridge was attractive for several other reasons, as well. Overclocking still represented a solid performance boost in this era, and Intel was offering overclocking capabilities at mainstream price points alongside chart-topping performance out of the box. They also introduced Quick Sync to accelerate video encode/decode, and although we have video encode/decode acceleration in modern GPUs, Quick Sync still serves as a fundamental feature for video editing and media servers.

AMD’s response was Bulldozer, which was rumored ahead of release to outperform Intel’s Core i7-950 by upwards of 50%, as well as introduce a true eight-core chip to the consumer market for the first time. If that wasn’t enough, prior to release, the flagship FX-8150 set a world record for clock speed, peaking at 8.429 GHz. AMD would be the first to release a consumer CPU with eight cores, but just about every other aspect of Bulldozer was problematic.

Achieving eight cores in a single package came with significant trade-offs. Up to this point, AMD hadn’t used any form of simultaneous multithreading, but it implemented a version of SMT in Bulldozer. Unlike Intel’s traditional SMT implementation, Bulldozer used two integer units on a core, but shared floating point resources. The major trade-off was how small the integer execution clusters were in order to save space. AMD designed an architecture for a world of heavily-threaded software that just didn’t exist at the time, and it traded very important single-core speeds to achieve that design.

The flagship FX-8150 was marketed as an eight-core chip, with the eight-core count coming from AMD’s odd SMT implementation. Although there were two integer execution units per “module,” as AMD calls them, the floating point unit was shared. This discrepancy was actually the focal point of a 2015 lawsuit, which AMD settled in 2019 to the tune of over $12 million.

It was a flop. Angelini sums up the issue nicely in his October 2011 review of the FX-8150: “[Intel doesn’t] have to do anything at all. Its nearly year-old 95 W parts fend for themselves without even a price adjustment.” Intel had new chips of its own, as well. Just six months later, it introduced Ivy Bridge, taking the solid foundation of Sandy Bridge and moving it down to a 22nm node.

Ivy Bridge wasn’t a big hit on desktop, but it didn’t need to be, given the advantage Intel had already established with Sandy Bridge. Improvements were in the single digits, as we noted in our Core i7-3770K review, but Ivy Bridge brought improvements to integrated graphics to fight against AMD’s burgeoning lineup of APUs and thermal improvements targeting small form factor devices — especially laptops in a new category of “ultrabooks” that Intel was targeting.

12 months after the FX-8150 was introduced, AMD released a revision of the Bulldozer architecture named Piledriver. It promised better IPC and higher clock speeds, which was compelling, as AMD’s FX-4170 released earlier in the year as the first CPU to hit 4 GHz out of the box. Piledriver had a test run earlier in the year through AMD’s Trinity APUs, as well, showing around a 15% improvement compared to Bulldozer.

The flagship Piledriver, the FX-8350, was indeed better than its Bulldozer predecessor, but it was clear the underlying architecture had issues that wouldn’t allow AMD to scale up. AMD’s flagship was only competitive with Intel’s Core i5 options, and power use, although tamed in Piledriver, still meant the chip ran hot. It was immediately forced into a price cut upon release. After two failures to launch flagships, Intel effectively owned the high-end, which it continued to dominate with CPUs like the Core i7-3970X.

Around six months after Piledriver chips shipped, AMD released another two chips, both running at an insane 220W TDP and shipping with their own liquid cooling system. The highest-end offering, the FX-9590, was the first CPU to hit 5 GHz out of the box. AMD had learned the hard way what it had preached back in the Athlon XP days — clock speed isn’t everything.

The dawn of 14nm (2014 - 2016)

The dawn of 14nm (2014 - 2016)

With a 22nm product shipped, Intel went back to a tock and came out the other side with Haswell. Today, Haswell is heralded as a legendary architecture; we’ve all seen forum comments about gaming with a Core i7-4770K more than a decade after it was released. At the time, however, it established a narrative that would follow Intel for the next several years. That narrative being that Intel ships a new generation of quad-cores, each year, with minor IPC improvements and not much more. That’s certainly the impression Angelini came away with after reviewing the Core i7-4770K.

Intel didn’t have to move the needle beyond that point. Although AMD had shipped Bulldozer and Piledriver, a lot of turmoil was going on behind the scenes. Two further revisions of Bulldozer, Steamroller, and Excavator were planned, but AMD largely canceled the two revisions outside of a few low-end products. During this time, AMD underwent a series of sweeping layoffs and executive changes, cutting thousands of employees. Looking back at the time through a modern lens, some long-time AMD employees say that the company would’ve faced bankruptcy had it not been for semi-custom partnerships with Microsoft and Sony for their game consoles.

From the release of Piledriver in late 2012 through 2017, AMD didn’t release a ton of new CPUs. It released some revisions of Piledriver chips like the FX-8370, but we didn’t see any new microarchitecture. And planned node shrinks with Steamroller and Excavator were canned, with only a handful of desktop CPUs surviving, which were repurposed as low-end Athlon X4 CPUs years later. Intel had won.

It didn’t immediately rest on its laurels, however. Hearing the criticism of Haswell for desktop enthusiasts, Intel introduced the Core i7-4790K alongside other ‘Devil’s Canyon’ chips in mid-2014 as it worked on another tick behind the scenes down to 14nm. The result, which arrived almost a year to the day after Devil’s Canyon, was Broadwell. The Broadwell-H range — not Broadwell-E, which shows up later — isn’t very big and didn’t have much for enthusiasts. But Broadwell got Intel down to 14nm, and it would stay there until the release of Alder Lake CPUs in 2021.

We didn’t know that at the time, though. Broadwell laid the 14nm foundation for Skylake, which came hot on the heels of Broadwell-H after that architecture experienced a series of delays. Intel’s first true eight-core chip, the Core i7-5960X from the Haswell-E range, still gave the HEDT market what they were looking for, but Skylake was pushing ahead in the mainstream. In early performance testing, we said Skylake was the first architecture “to really get enthusiasts excited since Sandy Bridge.”

Intel didn’t need to rush at the time, so it didn’t. There wasn’t an Athlon 64 breathing down Team Blue’s neck. A year later, in 2016, Intel launched Broadwell-E for HEDT, marking the first-ever 10-core desktop CPU with the Core i7-6950X. It wasn’t a massive leap forward over Haswell-E, but Intel was competing with itself. “Intel’s clearly the prettiest girl in the room, is well aware of this fact and, based on Broadwell-E's pricing, doesn't need to beat the ‘value’ of last generation's -Es by much,” our reviewer Igor Wallossek wrote in his Broadwell-E review.

Although Intel was a clear market leader, it was slowing down significantly. Earlier in 2016, it quietly revised its tick-tock cycle, moving to a tick-tock-tock cadence where we’d see a new process followed by a new microarchitecture followed by an optimization of that architecture. Intel wasn’t juiced up with Moore’s Law like it was in the early aughts, but it competed in a category of one. What were you going to do? Buy AMD?

Feeling Zen (2017 - 2020)

Feeling Zen (2017 - 2020)

Shortly after the ball dropped into 2017, Intel released its Kaby Lake range of CPUs, now sporting the “14nm+” process and serving as the first optimization pass in Intel’s new release cadence. It was fine. The range came with a clock speed bump over Skylake, but otherwise, Intel released the same architecture sporting nearly identical specs, from core counts to cache sizes.

Behind the scenes, trouble was brewing. Six months before Kaby Lake made its way to market, AMD detailed its first entirely new microarchitecture since Bulldozer, named Zen. In addition to promising a 40% improvement in IPC over Excavator, the Zen platform would come with support for DDR4 and finally move AMD down to a 14nm node. For the architecture itself, AMD implemented SMT, completely redesigned its cache hierarchy, and added a micro-op cache to aid an updated branch predictor.

Two months after Kaby Lake rolled out, AMD launched the Ryzen 7 1800X. The revolution didn’t happen in a day. In our Ryzen 7 1800X review, reviewer Paul Alcorn (now Tom’s Hardware editor-in-chief) wrote: “AMD's Ryzen 7 launch represents more than just a new CPU family. For most of our readers, it signals the return of competition to the enthusiast-oriented processor market. And considering the flagship 1800X’s potent cost advantage compared to Intel's Core i7-6900K… Ryzen 7 does deliver. It's just not as universally superior as the company wanted everyone to believe.”

AMD still had quite the year ahead. A month later, the Ryzen 5 1600X launched with performance that could rival Broadwell-E, just for a much cheaper price. And by Summer, the Ryzen 3 1300X proved you didn’t need an expensive CPU and motherboard to get into overclocking. AMD capped its Ryzen rollout with Threadripper, scaling up the Zen microarchitecture to massive core arrays and finally bringing something to the HEDT market — a market that Intel had almost wholly owned since the Pentium 4 days.

Still, Zen had shortcomings, particularly in games, where just about any quad-core from Intel still ruled the roost. AMD was competitive, but Intel was still firmly in the driver’s seat. It barely reacted to the Ryzen onslaught over the summer, rolling out its high-end Skylake-X and Kaby Lake-X HEDT offerings throughout the back half of 2017. Even then, however, problems started emerging.

Kaby Lake-X was effectively a rerelease of Kaby Lake with a bit of extra headroom, but restricted to the expensive X299 platform. It was discontinued less than a year after release. Skylake-X was Intel’s true next-gen HEDT offering, signaled by the first use of “Core i9” in front of its flagship SKU. It performed like an Extreme Edition, and it was priced like one too, despite an issue in thermal dissipation that we uncovered in our Core i9-7900X review. Meanwhile, AMD was rapid-firing firmware and chipset updates for its small Ryzen range, and fixing several issues that came up in reviews in the process.

Less than a year after Kaby Lake launched, Intel released Coffee Lake, which was yet another Skylake revision built on 14nm, but this time with extra cores in tow. As you can read in our Core i7-8700K review, Coffee Lake did what Intel wanted it to do, shoring up the fight in heavily-threaded productivity applications against AMD while maintaining leadership in games. Still, AMD was making headway. By the end of 2017, estimates suggest AMD took back anywhere from 2% to 12% market share from Intel, with the higher end of the spectrum coming mainly from the DIY PC market. That’s no small feat for a company that was dead in the water with CPUs 12 months earlier.

Back on more even footing, the next goal post was a node shrink. Intel was gunning for 10nm, which is a milestone it failed to meet with both Kaby Lake and Coffee Lake. AMD, as a fabless designer, was at the mercy of its then-partner GlobalFoundries for the next node shrink. AMD struck first with Ryzen 2000 in early 2018, built on GlobalFoundries 12LP node, which was a revision of the 14LP (14nm) node used in the original Zen. Fittingly, AMD called it Zen+.

Debuting the architecture was the Ryzen 7 2700X, which was an iterative update. However, it helped reacquaint the market with the progress AMD had made. Zen+ came with higher frequencies and reduced memory latency, and all of the software adjustments AMD had made after the original Zen launch. And the range seemed specifically designed to undermine Intel, offering overclocking support across the full stack (and with B-series chipsets), and bundling a surprisingly decent cooler in the box.

Intel still held the edge in gaming, but the margins were narrowing, especially with a bit of overclocking thrown into the mix. Intel was feeling the heat, due in no small part to its continued issues moving down to 10nm. It responded in late 2018 with Coffee Lake Refresh, bringing the Core i9 branding into its main lineup for the first time with the Core i9-9900K and offering an eight-core, 16-thread chip. The strategy, it seems, was to push out AMD at the high-end, as Ryzen 7 was closing in on Core i7.

It worked. Intel had the fastest gaming processor on the market, and even the Core i7-9700K managed to push Intel’s lead in the Ryzen 7 battle higher. These marginal updates were buying time for AMD and Intel. Both companies clearly understood that whoever could go below 14nm first would have a massive advantage, and likely define an entirely new market dynamic.

AMD claimed that advantage for itself with the introduction of Zen 2 in mid-2019. Bolstered by TSMC’s 7nm node, AMD pushed out the Ryzen 9 3900X, moving beyond eight cores to AMD’s first 12-core consumer design. Intel held a slight edge in gaming through Coffee Lake Refresh, as you can read in our Ryzen 7 3800X review, but that delta was becoming less important as AMD took the lead in heavily-threaded workloads.

Later in the year, AMD rolled out the Ryzen 9 3950X, the first 16-core desktop processor ever. It was a bloodbath. Less than a year earlier, Intel had introduced its Skylake-X HEDT platform, including the Core i9-9980XE priced at $2,000. Now, at stock settings, the $750 Ryzen 9 3950X offered better multithreaded performance, along with competitive single-threaded and gaming performance. And you didn’t need to shell out for Intel’s expensive HEDT platform. And you could unlock PCIe 4.0, whereas Skylake-X (and even the following Cascade Lake-X) were locked to PCIe 3.0. You don’t spend top dollar on an HEDT platform for last-gen connectivity.

Intel was getting pushed into a corner, and it followed up less than six months later with Comet Lake to stave off AMD’s Ryzen onslaught. The flagship Core i9-10900K allowed Intel to maintain the lead in gaming, but now, AMD was in a clear lead in applications with the Ryzen 9 3950X.

Reviewing the Core i9-10900K, Alcorn wrote: “The Core i9-10900K is exactly what we would expect from an overclocked 10-core 14nm Skylake derivative: Exceptional performance in gaming and lightly-threaded workloads, competitive performance in multi-threaded work, and downright ugly power consumption and thermal output. And that's pretty much what you get with the Core i9-10900K – an overclocked 14nm processor right out of the box.”

AMD didn’t let up. It moved onto Zen 3 later in the year, launching its 16-core Ryzen 9 5950X alongside the main range in late 2020. And with its fourth Ryzen salvo launched, the battle was over. Opening our Ryzen 9 5950X review, Alcorn wrote, “With the Ryzen 5000 series, it's fair to say that AMD has finally, and fully, eclipsed Intel's performance dominance in desktop PCs.” It was a clean sweep, with AMD taking the lead in gaming, multithreaded, and single-threaded performance. Three years and four CPU generations later, AMD was back on top.

Forging a new path (2021 - 2024)

Forging a new path (2021 - 2024)

In the years leading up to 2021, it had become clear that a tick-tock, or even a tick-tock-tock, wasn’t possible any longer. Process shrinks were arriving later, and a pesky little pandemic threw the tight supply chain required for chip manufacturing into a frenzy. Both AMD and Intel knew they needed a different approach, but that would manifest in wildly different ways.

Intel was all-in on a hybrid architecture, using a mixture of microarchitectures on a single package to bolster core counts, similar to Arm-based designs. Intel talked a lot about Alder Lake leading into 2021, overshadowing its own launch of 11th-Gen Rocket Lake chips. The flagship Core i9-11900K was a massive disappointment, carrying all of the issues of the previous-gen Core i9-10900K while packing two fewer cores. Yes, Intel actually cut two cores from its flagship.

It seems Intel knew the issues with Rocket Lake. The chips launched with little to no fanfare, and as opposed to a gradual rollout like we see with most CPU generations, Intel blasted every model of Rocket Lake onto the market, knowing full well that Alder Lake chips would take their place eight months later. AMD, with renewed confidence, slowly built out the Zen 3 lineup with new APUs and variations of Ryzen 5000 as it worked on its next-gen Zen 4 architecture.

In late 2021, Intel swept Rocket Lake under the rug with the release of Alder Lake. Intel had finally moved on from 14nm with Intel 7 (10nm), and it was mostly successful. Intel reclaimed top placements in gaming, multithreaded, and single-threaded performance, and although the margins were thin, Alder Lake made it clear that Intel wouldn’t go quietly. “The Alder Lake processors mark a massive generational leap forward for Intel in nearly all facets, including gaming, performance in lightly- and heavily-threaded work, power consumption, overclocking, and platform connectivity options,” wrote Alcorn in our Core i9-12900K review.

AMD was working on something unique of its own, however. Zen 4 was in the oven, and it was clear there would be a competitive battle with Alder Lake. But before Zen 4 arrived, AMD introduced the Ryzen 7 5800X3D. It was the first processor with AMD’s 3D V-Cache packaging, and at the time, it looked like a slightly-tuned processor targeting gamers, with somewhere in the range of a 10% to 15% uplift in gaming performance specifically. In a surprising turn, the speculation actually undersold just how big of a deal the Ryzen 7 5800X3D would become.

Six months after the release of the Core i9-12900K, AMD was back on top of the gaming charts with the Ryzen 7 5800X3D, no less sporting a last-gen architecture and an SRAM stacking technique that limited boost clocks and locked the multiplier down. It outran the Core i9-12900K by nearly 10% in games while costing hundreds less, and it was nearly 30% faster than a stock Ryzen 7 5800X, as you can see in our Ryzen 7 5800X3D review.

Intel would come back with Raptor Lake in late 2022, but the Ryzen 7 5800X3D established a new category of true gaming CPUs that traded some application performance for peak frame rates. And that’s a category of chips that even today Intel hasn’t managed to crack.

AMD came first, however, launching Zen 4 in September 2022. The flagship Ryzen 9 7950X managed to leapfrog the Core i9-12900K, as you can see in our Ryzen 9 7950X review, but not by much, and the Ryzen 7 5800X3D remained at the top of the gaming charts. Immediately, speculation around 3D V-Cache chips for Zen 4 went into motion. Also tampering the Zen 4 release was an entirely new platform, which required costly DDR5 memory.

Intel capitalized with Raptor Lake mere weeks later. The flagship Core i9-13900K was back on top across tests, even managing to outclass the Ryzen 7 5800X3D in games. For the first time since the heyday of Athlon, we had a hotly competitive CPU market with AMD and Intel leapfrogging each other with each new release. Still, there was a niche that wasn’t being filled. 3D V-Cache disrupted the status quo for gaming processors, but it came with a significant trade-off to application performance. The stage was set for a CPU that could offer the best of both worlds.

AMD delivered that in early 2023 with the Ryzen 9 7950X3D and (to a much lesser degree) Ryzen 9 7900X3D. A couple months before, Intel cracked the 6 GHz barrier out of the box with the Core i9-13900KS, but AMD was offering something more compelling than peak clocks. The Ryzen 9 7950X3D managed to outclass Intel in multithreaded and single-threaded performance, all while offering a double-digit jump in gaming performance thanks to 3D V-Cache.

Raptor Lake saw a refresh later in 2023, and although the flagship was able to close the application performance gap in our Core i9-14900K review, AMD still held a firm grip on gaming performance, especially with the trimmed-down and relatively affordable Ryzen 7 7800X3D. AMD had taken the lead, but Intel, finally, executed its tick-tock-tock strategy and set its eyes on a radically new architecture in the form of Arrow Lake.

Reckoning with the real world (2024 - today)

Reckoning with the real world (2024 - today)

Under AMD’s thumb and clearly behind in pace, Intel needed to innovate. The result was Arrow Lake. Like Bulldozer, it’s easy to write Arrow Lake off in hindsight, especially given how recent it is. As you can read in our Core Ultra 9 285K review, Arrow Lake chips only marginally improved in application performance over their 14th-Gen counterparts, and they were actually slower across most games. But, architecturally, Arrow Lake is as big a swing as Bulldozer was.

For the first time, Intel outsourced manufacturing to TSMC, clearly behind the Taiwanese manufacturer for cutting-edge nodes. It disabled Hyper-Threading, killing off a staple of Intel CPUs for decades, and it doubled down on Intel’s hybrid architecture. Those bets didn’t pay off, but they were big bets for a company struggling to reckon with a reinvigorated AMD.

AMD followed up Zen 4, predictably, with Zen 5 in mid-2024, shortly before the Arrow Lake release. With Arrow Lakes' struggles, it’s easy to forget the problems Zen 5 had at launch, and the relatively small generational uplift it offers even today. AMD has continued to build out this lineup with X3D chips, and it finally delivered 3D V-Cache on both CCDs with the Ryzen 9 9950X3D2. But going back to our Ryzen 9 9950X review, Zen 5, at its core, isn’t the massive uplift we had become accustomed to in the early days of Zen.

Intel tried to give Arrow Lake a bit more life with a small refresh earlier this year in the form of the Core Ultra 7 270K Plus and Core Ultra 5 250K Plus, and those CPUs set the stage for the next era of CPUs. They put Intel into the position AMD found itself during the Bulldozer/Steamroller days, clearancing off silicon to maintain a competitive position in the market.

That’s where we are today, with our sights set on Zen 6 and Nova Lake. But there are some realities in the PC enthusiast space that we have to contend with today. DRAM pricing is out of control, and showing no signs of slowing down, and a sudden boost in demand for CPUs for agentic AI means consumer chips have taken a backseat. Zen 6 and Nova Lake were both expected by the end of the year; it’s looking more likely that they’ll slip into 2027.

History doesn’t repeat, but it often rhymes, and we can see traces of days past over the last 30 years start to creep into the dynamics today. Today, we see a defiant AMD and an Intel that seems ready to get scrappy in order to earn back market share. Will it pay off? We don’t know, but Tom’s Hardware will be here to cover whatever comes next in the world of CPUs, just as we’ve been for the past 30 years.