Michael Dell doesn't think the memory shortage is anywhere near over. Speaking with The Wall Street Journal, Dell's chairman and CEO said things will get worse next year, as the AI data center buildout keeps soaking up DRAM and HBM supply. According to him, the bottleneck for Dell isn't demand, since the company is sitting on roughly $95 billion worth of AI server orders it hasn't delivered yet, the largest backlog in its history. Memory is what's holding those shipments back, as AI accelerators need plenty of HBM, while the servers around them still rely on conventional DRAM.

Relief also looks quite far away. New memory fabs take around four years to build, so Dell apparently doesn't expect meaningful new supply until 2029 or 2030. That aligns with Micron's CEO, who recently said supply will be much tighter in 2027 and 2028 than it is this year. Interestingly, SK hynix's CEO said something similar, expecting the memory shortage to last through 2030. Dell is reportedly coping by buying DRAM only when final assembly is close and by locking in long-term supply agreements, though there's no word on how much of that applies to its consumer products.

AI-generated code in the realm of Linux has been responsible for more than a little upheaval in recent months, with some projects seeing heated debates surrounding the topic. Other project developers have expressed frustration with an uptick in what they call "AI slop PRs," which they say have increased maintainer workload to unreasonable levels. The Linux Foundation recently conducted a survey of Linux project maintainers, who reported that generative AI has resulted in increased productivity, but the research also finds that there has been a substantial increase in low-quality AI slop PRs, as prior anecdotes predicted.

Summary

• Linux Foundation Research published a new study examining how generative AI (GenAI) is reshaping open source software (OSS) development.
• The research uncovered significant productivity gains alongside growing maintainer pressures.
• A majority (59%) of project maintainers report growing contributions to OSS, but 25% of incoming pull requests (PRs) are estimated to be low-quality "AI slop."
• For both contributors and maintainers, GenAI boosts productivity by up to 50% across all types of OSS development tasks.
• The findings point to a need for stronger AI governance, including clear contribution policies, standardized attribution and verification and continued investment in human mentorship.

Anyone holding out for an affordable next-gen NVIDIA card might have a long wait. Hardware leaker Kepler_L2 claims on NeoGAF that the GeForce RTX 6060 will "launch much later," adding that, based on recent trends, he doubts it will even match the PlayStation 6. He was replying to a forum member who argued the PS6 would compete with the RTX 6060, in the same thread where he claimed the console's RAM and storage alone cost over $680. Kepler didn't give an actual date, and the 2029 figure going around seems to be an extrapolation. Both Kepler and kopite7kimi have placed NVIDIA's Rubin-based RTX 60 series in 2028, and the x60 cards usually trail the flagship by a few months. The RTX 5060 arrived under four months after the RTX 5090, while the RTX 4060 took over eight months to follow the RTX 4090.

With a flagship launch around mid-2028 or later, that could push the RTX 6060 into 2029. That said, the "much later" comment seems to be relative to the PS6, which is reportedly targeting late 2027, so the exact timing is still anyone's guess. Kepler didn't say why, but the ongoing memory shortage looks like the obvious suspect. NVIDIA reportedly postponed its RTX 50 SUPER refresh earlier this year to prioritize AI products that use the same GDDR7 memory, and the company recently admitted that RTX 5090 supply is tight. Of course, NVIDIA hasn't announced the RTX 60 series officially yet, so treat this as a rumor. Anyone looking to buy a new NVIDIA GPU should probably expect the current RTX 50 lineup to stick around for quite a while.

Brandon Stirrup, a senior business strategy manager at Xbox, has stirred up quite a debate after posting on X that the games industry is dragging its feet on AI. Over the weekend, Stirrup shared a rough football game he says he "vibe coded" in about three hours of active work, despite never having made a game before, with an AI agent testing and iterating on it. He went on to say that studios not planning around AI are "planning for a fake future." He also said he'd happily take the hit if it meant hundreds of people at Xbox could stop pretending they don't use AI. Here's part of his X rant:

Look if I'm fired tomorrow it will be because I told you the truth that nobody wants you to know. You can make a game with a simple prompt and yes you can make custom, tailored games. Everything you want in a game is right there for the taking, for better or worse. Gaming is the only industry slow-walking the transition because users have negative sentiment. Everyone else accepts reality, Gaming is the only industry where leaders are afraid to talk about AI. Users want the truth. We're using AI.

It seems as though AMD is readying a new mobile APU that bears a striking resemblance to the existing Ryzen Z2 Extreme. The AMD Ryzen AI 7 HX 455 was recently leaked by Golden Pig Upgrade on Weibo, with the post suggesting that the new CPU would launch with eight cores and 16 threads and a 16 CU iGPU, while the 16 CU iGPU will reportedly be identical to the existing Radeon 890M on the Ryzen AI Z2 Extreme, which has proven itself capable of very decent gaming performance in our in-depth review of the ASUS ROG Xbox Ally X.

The exact core breakdown is unknown, but in a response to a question about the CCD layout, the leaker alludes to a similar layout to the Z2 Extreme, which features three Zen 5 cores and five Zen 5c cores across two CCDs, although the new APU will have a significantly larger cache to work with—24 MB L3 instead of the Z2 Extreme's 16 MB—signalling that it will be a cut-back version of the existing Ryzen AI HX 470 APU, which features four Zen 5 cores and eight Zen 5c cores.

Sony has a new patent application that describes its own take on frame generation, and the PlayStation 6 seems like the obvious target, though no console is named. Titled "Generation Super Sampling," the filing (US 2026/0320740 A1) was published on October 15 and lists Sony engineer Denis Dupeyron as the inventor. It describes an encoder neural network that turns a newly rendered frame into a compact embedding, and a decoder network that uses it to predict the next frame. Optional inputs include motion vectors and recent controller inputs, which help the model guess where the player is heading. That frame is displayed while the next real one is still rendering.

The filing points out that existing methods insert a generated frame between two real ones, which means waiting for both and adding delay. Sony's version predicts forward from a single frame, with the goal of smoothing out hitching in heavy scenes without lowering visual quality. It can also predict several frames at once, and there's a simpler non-AI fallback that shifts blocks of pixels along their motion vectors. Apparently, the model trains on the video stream itself, comparing each guess with the real next frame, so no human labeling is needed. However, the filing includes no latency or image quality numbers, and it remains pending, so there's no guarantee any of it ships.

The memory crunch looks like it is hitting the PlayStation 6 harder than expected. Hardware leaker KeplerL2 claims on NeoGAF that the RAM and NAND flash storage in Sony's next console alone cost over $680. He posted it as a reply to a forum member who floated $680 as a possible subsidized retail price, which pretty much puts that idea to rest. Keep in mind that figure covers only memory and storage, before the APU, cooling, power supply, and the rest of the hardware are counted. This is not the first time Kepler has updated his estimate. Back in March, he put the console's bill of materials at around $760, and by July, that had climbed to roughly $960.

Current leaks suggest the PS6 uses GDDR7 memory on a 160-bit bus with support for up to 40 GB, with 30 GB considered the more likely configuration. There have also been reports that Sony looked into cutting memory capacity to keep costs down, and the culprit is the same as everywhere else: AI data centers are soaking up DRAM and NAND supply, which pushes prices up for everyone. Sony has said nothing official about PS6 pricing, so it's reasonable to treat all of this with some caution. That said, if memory prices stay where they are, a PS6 that costs significantly more than the PS5 Pro seems likely, and a $1,000 price tag now looks quite possible.

Konami Digital Entertainment president Hideki Hayakawa doesn't expect generative AI to shake up the games market much. In Konami Group's Integrated Report for the fiscal year ended March 2026, first spotted by Dengeki Online, Hayakawa responds to the popular view that AI will make game development far easier to break into. His argument is that general-purpose game engines and user-generated content have already lowered those barriers, so AI probably won't change much on that front. Hayakawa also made it clear where Konami's priorities lie.

Hideki Hayakawa (machine translated)I don't think the market environment will change drastically. Entry barriers have already been lowered by the advancements of generic game engines and the spread of user-generated content. Even if AI use does lead to improvements in production efficiency, I believe that the most important factors in producing entertaining and innovative video games are people and creators.

ASUS has listed the UGen Arc Pro B70 32 GB on its website, which makes it the company's first Intel Arc graphics card. ASUS sat out both the Alchemist and Battlemage consumer generations, and its last Intel discrete card was the passively cooled Iris Xe DG1-4G from 2021. Interestingly, ASUS files it under its UGen AI accelerator lineup rather than ProArt, putting it in the same family as the UGen300 USB AI accelerator the company showed at CES 2026. ASUS hasn't announced pricing or a release date yet. Intel launched the Arc Pro B70 at $949 back in March, though US retail listings now seem to sit closer to $1,300.

The ASUS card uses the full "Big Battlemage" BMG-G31 die with 32 Xe2 cores, 256 XMX engines, and 32 GB of GDDR6 ECC memory on a 256-bit bus for 608 GB/s of bandwidth. ASUS rates it at a 2,800 MHz boost clock, up to 367 INT8 TOPS, and a 290 W board power, which is the top of Intel's 160 to 290 W range for the B70 (the reference default is 230 W). Cooling comes from a dual-slot blower, and ASUS says four cards can be paired for 128 GB of combined VRAM, enough for models with up to 120 billion parameters according to the company. ASUS also joins ASRock, Sparkle, and Maxsun, which already have their own B70 models, so pricing will likely decide how well the UGen version does.

Update Oct 27th: ASUS has confirmed European pricing for its UGen Arc Pro B70 card. It is available in Germany and Austria at a recommended price of €1,799 including VAT.

Canonical has released Ubuntu 26.10 "Stonking Stingray," shipping with GNOME 51, Linux kernel 7.3, and a new optional speech-to-text feature called Myna. Following its official release in September, GNOME 51 lands in Ubuntu 26.10 and promises a more responsive desktop experience and improved accessibility. Linux kernel 7.3 brings hardware support for Intel Xe3P (Nova Lake), AMD Zen 6, and NVIDIA 615.x Linux drivers, along with improved DRM, power management hooks and out-of-the-box Lustre filesystem support. However, the headline that could easily get unnoticed is the Myna, an opt-in speech-to-text agent that runs entirely on-device using local LLMs. Canonical says it is positioning it as the first step toward a more natural AI-assisted desktop, with accessibility as the stated priority for the initial rollout. Jon Seager, VP of Ubuntu Engineering at Canonical, says "Stonking Stingray delivers the first milestone through the integration of Myna: a speech-to-text agent that utilizes on-device LLMs. We've led Ubuntu's first AI integration with accessibility to ensure that cutting-edge features are available to all users from day one."

Security and enterprise features got expanded with a full migration to Rust-based uutils coreutils, post-quantum package signing, and broader TPM-backed full-disk encryption support. Microsoft Entra ID integration brings support for password-plus-MFA and passwordless authentication flows. Ubuntu 26.10 also upgrades ROCm to version 10.0, adding support for newer AMD GPU architectures and frameworks including vLLM and SGLang. Going back to Ubuntu's new AI Myna feature, Canonical also says that it will not go away, rather it will continue to evolve and will be included by default with new Ubuntu releases.

Samsung will reportedly break ground on its second chip fab in Taylor, Texas, this November, with mass production targeted for 2030. DigiTimes, citing South Korea's The Guru, reports that Taylor Fab 2 will span roughly 2.7 million square feet. Samsung also appears to have confirmed the timing at an October 15 event near the site, where Jonathan Taylor of Samsung Austin Semiconductor said construction should start early next month and that Fab 2 will be "one and a half times bigger than Fab One." In July, Samsung admitted that its leading-edge capacity was lagging customer demand, while inquiries for its 1.4 nm process, a node it was reportedly considering shelving last year, kept climbing. The process for Fab 2 hasn't been disclosed yet.

The first Taylor fab, which has slipped well past its original 2024 target, seems to be on track this time. Installed tools have started running processes as the plant goes through staged qualification. Fab 1 uses EUV lithography for Samsung's 2 nm process, with operations planned by the end of 2026 and volume production in 2027. Its best-known customer is Tesla, which signed a $16.5 billion deal running through 2033 that covers the AI6 chip, and Samsung is also set to build Tesla's AI5 on 2 nm in Texas. How orders will be split between the two fabs is unknown. Samsung plans to invest over $37 billion across Central Texas, backed by up to $4.745 billion in CHIPS Act funding. Given how long Fab 1 took, it's reasonable to expect that the 2030 target will depend heavily on demand.

NVIDIA today announced the availability of NVIDIA RTX Spark laptops, bringing together RTX graphics and NVIDIA's full-stack AI platform in slim, premium Windows PCs for accelerated creation, immersive gaming and powerful local AI—rolling out from ASUS, Dell, HP, Lenovo, Microsoft and MSI.

Combining decades of NVIDIA innovation into a single superchip, RTX Spark delivers up to 128 GB of unified memory and 1 petaflop of FP4 AI performance to run large language models and AI agents locally. On a single device, creators, developers and gamers can move seamlessly between their everyday workflows and emerging AI agents on the familiar Windows platform.

ASUS today announced that the ProArt P16 and P14 laptops, powered by NVIDIA RTX Spark, are now available for purchase in Canada, with select configurations shipping starting today through the ASUS Store, Amazon Canada, Best Buy Canada, Memory Express and Newegg Canada. Purpose-built for creators, AI developers and workflow builders, these Windows PCs combine high-capacity unified memory, an NVIDIA Blackwell RTX GPU and an NVIDIA Grace CPU to support demanding local AI and creative workloads. The NVIDIA CUDA ecosystem and an expanding ASUS-exclusive agent-and-application solution make local AI easier to deploy, personalize and use, with the flexibility to connect cloud services when needed.

Together, Hermes Agent and ComfyUI deliver a compelling local-first, cloud-flexible AI experience. By bringing industry-validated workflows and agent-connected creative tools to users' own PCs, they enable generative AI creation and AI-assisted development, with the flexibility to integrate cloud services when needed.

Just 48 hours after the launch of the Total War Shogun 2: Complete Edition, the remake of the 2011 RTS classic, SEGA has announced that it will be pulling the game from all storefronts, at least temporarily. In the X post announcing the game's withdrawal from the market, SEGA acknowledges that the game has "fallen short of both your expectations and our own standards," citing poor reviews, comments, and player feedback following the launch of the game.

Shortly after the launch of Shogun 2 — Complete Edition, the game already has an abysmal review score of 15%. Most player complaints center on the game's exorbitant use of AI in unit cards, poor performance and graphical bugs, poorly thought out UI, and an apparent lack of care and polish in the game overall. After the game is removed from the relevant platforms, the original Total War: Shogun 2 will be reinstated and available to purchase and play, although existing owners will still be able to play the remaster. SEGA says that it is "investigating the issue," and will identify and correct the issues highlighted by Shogun 2 Complete Edition players.

Epic Games released a preview of its rebuilt, "vibe coded" Epic Games Launcher on October 14, and took a direct shot at Steam while at it. Epic's own chart shows the new launcher cold starting in 5.47 seconds against 10.07 seconds for Steam, which the company bills as 46% faster. That lead apparently didn't last long, as a clip posted by user girlglock_ on X shows the current Steam client starting in 5.23 seconds. Although it might seem like it at first, this isn't a rushed reply from Valve. The stable Steam Client Update that lists "Improved Steam Client startup performance" is dated October 13, a day before Epic's announcement, and Steam's beta client was already getting startup improvements back in September. Valve hasn't shared any numbers for it, though.

Epic's figures, on the other hand, come from testing it commissioned from Keywords Studios earlier this month, averaging five runs on 13 machines with the three lowest-end systems left out, so they were probably taken on an older Steam build. Results will vary from one PC to the next, of course. Epic's launcher is also light on features for now, and the company advises competitive Fortnite and Rocket League players to stay on the current launcher during the preview. It does run natively on Linux, with Fortnite set to follow once Epic finishes rebuilding its anti-cheat, and the vibe coding part fits Epic's recent shift to AI-first development. Readers are advised to treat both sets of startup numbers as rough guides.

Ever since the announcement that it would be laying off 1,000 workers, Epic Games has been working overtime to pull users away from its largest competitor, Steam. Following the announcement that Epic Games had, seemingly out of the blue, figured out how to get Easy Anti-Cheat working on Linux to the same standard as its kernel-level implementation on Windows, Epic Games announced a new version of the Epic Games Launcher, with some interesting performance claims.

By Epic's own admission, and in accordance with its latest proclamation of being an AI-first company with "no handwritten code," the new Epic Games Launcher is vibe coded, but Epic Games claims that the new Epic Games Launcher launches up to 46% faster than Steam—5.47 seconds vs. 10.07 seconds—and restores from the system tray 14% faster than Steam—0.36 seconds vs. 0.42 seconds. The test results are an "average of 5 tests on each of 13 machines, excluding the 3 lowest-end machines tested," and testing was conducted by Keywords Studio, an independent game studio contracted by Epic. The new Epic Games Launcher is available as a preview, and is admittedly still in testing, so there may be bugs.

Apple has officially announced HomeView, its long-rumored smart home hub, priced at $299 in the US. It comes in two versions, a tabletop model and a wall-mounted one, both built around a 7-inch Liquid Retina display with True Tone that ranges from 500 nits down to just 0.3 nits. The tabletop model sits on a domed speaker base with a down-firing woofer, attached through a tilting stainless steel hinge. Inside, there's an A18 chip with Secure Exclaves, which Apple says keeps camera and microphone data separate from the rest of the system.

HomeView leans heavily on Apple's new Siri AI, which is still in beta on the device. Personal context features need a paired Apple Intelligence-capable iPhone or iPad, and some server-side features come with daily limits, with paid expanded access planned for later. The hub also recognizes up to six people by face, each with their own Home Screen. Smart home control covers Matter accessories as well as Apple's new Home Collection, designed with LG and Schneider Electric and arriving in December.

Cooler Master, a global leader in digital infrastructure cooling solutions, has outlined its end-to-end cooling portfolio for AI infrastructure, designed to support high-density computing deployments in new data centers and existing facilities expanding AI capacity. Anchored by a 2,500 kW Liquid-to-Liquid In-Row Coolant Distribution Unit (CDU), the portfolio spans facility-level heat transfer, rack and server cooling, and the equipment used to prepare liquid cooling systems for operation.

Thomas Chen, CEO at Cooler Master Corporation"Customers need choices that bring fast time-to-market, flexibility, and scalability for the deployment of their facilities. We aim to help them put liquid cooling to work, whether that means retrofitting an existing data center or building a new site. That requires comprehensive plans about how the equipment being assembled, installed, and serviced."

In the weeks leading up to publication, a number of software engineers and modders have decompiled and recompiled games, unlocked mods, and accelerated game emulation, resulting in everything from a browser-based Fallout: New York game to full PlayStation 5 game ports using AnyPS5. It seems as though Microsoft has now taken notice of the uptick in these AI-accelerated game mods and unofficial port projects and launched Xbox Labs in response.

In a recent email shared by Xbox CEO, Asha Sharma, she explains that the rapidly accelerating pace of game and mod development that has been a result of these AI "experiments challenge long-held assumptions about our industry," and that "it is easy to see the possibilities in making great games easier to access, giving old games new life and allowing more players to become creators and co-creators." The motivation behind Xbox Labs appears to be to get out ahead of these AI projects and "turn emerging ideas into things our players and developers can experience."

The second-generation Lenovo Legion Go handheld devices have already received a substantial price increase after their initial launch, with the entry-level Legion Go S jumping from $899.99 and $649.99 to $1,579.99 and $989.99, for the Z1 Extreme and Z2 Go models, respectively. This was after Lenovo already increased the price on the AMD Ryzen Z2 Extreme-powered Legion Go 2 from $1,349.99 to $1,999.99. Now, the same Legion Go 2 with 32 GB of RAM and 2 TB of storage that previously jumped to $2,849 can now be found on the Lenovo US online store for $3,081.99.

This is a $233 price jump that corresponds to around 8% of the total previous price, although the Legion Go 2 with 1 TB of storage can still be found at Best Buy for $1,999.99. The latest price increase for the Legion Go 2 comes as the gaming handheld continues to be out of stock at Lenovo, despite partner stores, like Best Buy, having stock of the handhelds. Lenovo hasn't yet announced the price hike for the Legion Go, but it seems reasonable to assume the price increase is linked to the ongoing RAMpocalypse that has repeatedly affected pricing across various PC hardware segments—but most severely RAM and DRAM.

Global 300 mm front-end fab equipment spending is projected to reach $149 billion in 2026, up 31% year-over-year, SEMI reported today in its latest 300 mm Fab Outlook. Investment is expected to continue growing at double-digit rates through 2029, reaching $227 billion. The stronger investment trajectory is accompanied by a sustained pipeline of new 300 mm fabs and production lines moving through construction and into operation. Excluding pure R&D fabs, SEMI projects 20 new 300 mm fabs or lines to begin construction worldwide in 2026 and 18 in 2027. In the United States, three are projected to begin construction in 2026 and four in 2027.

Ajit Manocha, SEMI President and CEO"Semiconductor demand is increasingly translating into physical manufacturing expansion across the global industry," "The latest outlook points to both stronger equipment investment and a substantial pipeline of fabs moving toward operation. This expansion will provide the manufacturing foundation needed to support AI, high-performance computing and a broad range of next-generation semiconductor applications."

NVIDIA has updated its DLSS Frame Generation algorithm to be faster and more efficient. In the latest NVIDIA App beta update, rolling out today, the company introduced "Preset C," the fastest Performance Mode for DLSS Frame Generation. NVIDIA demonstrates that this mode can make frame generation up to 35% faster on the GeForce RTX 5060, with minimal trade-offs in image quality. This improvement applies to the GeForce RTX 40-Series, GeForce RTX 50-Series, and RTX Spark.

According to NVIDIA, there are two models for DLSS: Multi-Frame Generation and Frame Generation. The first model uses a standard AI approach that inputs motion vectors and depth to understand how two rendered frames are constructed. The second model employs optical flow for in-game elements like UI and ray tracing. Today, NVIDIA replaced this optical flow model with a much faster and lighter AI algorithm that boosts FPS across DLSS Frame Generation, Multi-Frame Generation, and Dynamic Multi-Frame Generation. This new model is available for download through the NVIDIA App beta update, so ensure you're running the latest beta version to try out the "Preset C" setting.

Gemtek Technology Co., Ltd. announced it is highlighting an AI-native Wi-Fi 8 broadband gateway at Network X 2026, developed in collaboration with Qualcomm Technologies, Inc. and gFiber, Inc. to bring Edge AI and next-generation connectivity into the home. Built on the Qualcomm Dragonwing NPro A8 Elite Platform, the gateway is engineered by Gemtek and powered by intelligent software from Genix Networks, a Gemtek subsidiary, with gFiber as the launch service provider. It combines Wi-Fi 8 Quad-Band connectivity, Edge AI computing, and intelligent software for a more responsive broadband experience.

Bringing Intelligence into the Home Gateway
Designed as a home internet hub and a local platform for agentic AI, the gateway supports network optimization, device awareness, security, and proactive management.

Epic Games has long been supportive of putting more generative AI into its game development processes, with the gaming giant previously revealing an ever-increasing number of generative AI features in Unreal Engine, as well as plans to create Fortnite characters with AI—which resulted in Vampire Survivors developer, Poncle, pulling out of a Fortnite collaboration with Epic Games. More recently, a report by Bloomberg's Jason Schreier claims that insiders at Epic Games say the studio will be transitioning to an AI-first game development workflow.

Reportedly, the staffer was told by a senior that there would be "no more handwritten code," and that all art would be "generated with full AI workflows." Epic's shift to AI-first coding and game art will start with an unannounced Disney shooter that was delayed but was originally planned for a March 2027 launch, according to the publication. The shooter is slated to feature characters from both Marvel and Star Wars franchises, and it will now be produced largely with Anthropic's Claude AI suite, with humans reviewing work. Jake Jones, communications director at Epic Games, told Bloomberg that "we're embracing AI-accelerated development, led by Epic developers who are the visionaries and creative drivers for everything we do."

Intel has finally unveiled what its "Crescent Island" D-Series GPUs will look like, introducing a dedicated AI accelerator in a server form factor after previous projects stalled. The newly named D-Series GPU features a total of 32 Xe3P cores, 256 XMX engines, 32 MB of unified L2 cache, and up to 480 GB of LPDDR5X memory with up to 1.5 TB of memory bandwidth. Intel rates it for up to 1.3 PetaFLOPS of FP8 performance, with support for formats ranging from FP4 to FP64. A media engine with four encoders and four decoders is also included, which could assist with multimodal workloads where video generation and understanding is a must. The card comes with 160 GB memory configuration out of the box, which is exclusive to Intel-branded "Crescent Island" graphics cards, while the company allows its ODMs to equip the card with up to 480 GB of LPDDR5X memory for maximum VRAM capacity.

What is particularly interesting about the Xe3P is its pipeline. The XMX engines have been redesigned from the four-deep systolic design on the regular Xe3 to 16-deep processing on the Xe3P. Additional details include FP4 data format support alongside FP8, as well as large register files of 1 MB per core and L1 cache of 512 KB per core. In total, there are 256 Vector engines and 256 of the new XMX engines, each divided into groups of eight per Xe3P core. Additionally, in the Xe3P, there are no graphics processing elements to output imagery, as the silicon is designed purely for AI inference workloads that require long context and agentic tasks.