AMD

Both of the chip's CPU dies will include 64MB of extra cache stacked beneath.

AMD has released its official FSR Software Development Kit version 2.2, which includes the latest set of FSR technologies. This release features FSR Upscaling 4.1 and Ray Regeneration 1.1, showcased late last week as part of the Adrenalin Edition 26.3.1 WHQL driver. The updated FSR "Redstone" SDK v2.2 introduces a much improved FSR 4.1 upscaling technology for the RDNA 4 family of graphics cards. The transition from FSR 4.0 to FSR 4.1 demonstrates that the latest version provides much finer details of game scenery, especially when objects are in motion. In the Crimson Desert game demo, grass moved by the wind appears much more detailed with FSR 4.1 compared to FSR 4.0, which previously applied a somewhat blurry effect to the grass. This improvement brings the visuals closer to native rendering.

We take an in-depth look at AMD's 2026 and 2027 roadmaps.

Ultra Performance Mode gets a slight FPS boost, too.

AMD is spending record money on lobbying after decades of absence or near-absence from any form of political contribution and, in our opinions, is now joining the ranks of companies like Micron, NVIDIA, and Palantir in effectively bribing the US government for favorable deregulation, tax incentives, and safety bypasses for AI and data centers. We dig into AMD's millions of spending on Super PACs associated with President Trump, lobbyists, and in general, what we view as anti-consumer and anti-humanity efforts as AMD partners closely with the Federal Government to reduce safety nets around construction and development. Like NVIDIA and Micron, AMD is taking the same path of directly engaging in what we think are, effectively, bribes, while still maintaining an image of being the plucky underdog.

AMD has published a guide for running OpenClaw locally on Windows through two distinct hardware paths it calls "RyzenClaw" and "RadeonClaw," both built around its own silicon and designed to keep AI agent workloads off the cloud entirely. The push is part of AMD's broader "Agent Computer" narrative, where it argues that not every AI workload belongs in a data center, people and businesses want control over their data, affordable always-on AI without usage limits, and the confidence that their models are running locally rather than on someone else's infrastructure. The setup runs through WSL2 with LM Studio handling local LLM inference via llama.cpp, and supports Memory.md through local embeddings, no cloud dependency required. AMD says the environment can be configured in under an hour, targeting early adopters and developers experimenting with personal AI agents.

On Monday, AMD announced its latest Ryzen AI 400 Series and Ryzen AI PRO 400 Series desktop processors, based on the "Gorgon Point" silicon and powered by the "Zen 5" core configuration. This generation follows the Ryzen 8000G series, known as "Phoenix Point." However, it has been revealed that the Ryzen AI 400 series reduces the number of usable PCIe lanes compared to the previous Ryzen 8000G generation. The new top SKU offers 16 native PCIe 4.0 lanes, but only 12 are available to the rest of the system. Four of these PCIe lanes are used for the chipset link that connects the AM5 socket to the motherboard chipset, leaving fewer lanes for the end-user. Lower-tier chips may provide as few as 10 usable lanes, which is insufficient to run a discrete GPU at its full 16x lanes in the PCIe 4.0 connector on the AM5 motherboard. When a user installs an M.2 PCIe NVMe SSD, only eight lanes remain available for a discrete graphics card, meaning the GPU will operate in x8 mode instead of x16.

AMD is seeing server CPU demand it didn't fully anticipate, with CEO Lisa Su describing it as having "far exceeded" her expectations. Speaking at the Morgan Stanley Technology, Media & Telecom Conference earlier today, Su outlined how the landscape has shifted as agentic AI applications have taken off, pushing CPUs back into the spotlight alongside accelerators. The CPU-to-GPU ratio in AI compute workloads has evolved dramatically over the past few months, and top customers reportedly told Su that CPU demand sitting alongside AI was "under-forecasted." Supply is tightening as a result. The sudden spike in customer commitments left little time for the supply chain to adjust, though AMD says it's working closely with partners to ease existing bottlenecks and expects capacity to expand over the coming year. Su expressed confidence that AMD's product lineup is well positioned to address training, inference, and agentic workloads going forward. We've also seen this play out at the customer level, with hyperscalers like Meta signing CPU agreements with AMD and NVIDIA, a sign that compute is diversifying away from a pure GPU focus.

Remembering one of the jewels of the Super Socket 7 era.

cross-posted from: https://piefed.ca/c/amd/p/513386/i-m-annoyed-at-amd-s-latest-radeon-blunder

• Invidious
• Youtube

cross-posted from: https://piefed.ca/c/technology/p/484154/amd-ryzen-7-9850x3d-cpu-review-benchmarks-gaming-power-thermals-ft-ddr5-4800

• Invidious
• Youtube

The tradeoffs for the $499 9850X3D make it hard to get excited about.

The intriguing "GFX13" identifier has turned up again, after many months of silence. Late last Friday (January 23), Kepler_L2 highlighted a public-facing LLVM Project update that defines initial support of Team Red's "gfx1310 target." One team member comment points to this new property being identical—as a temporary arrangement—to the RDNA 4 generation's "GFX12" and "GFX1250" IDs. Prior to 2026, a bunch of AMD patch notes mentioned early "GFX13" commitments, alongside next-gen "RDNA 5" or "UDNA" graphics architectures.

AMD China and Micro Center have confirmed that the upcoming gaming CPU, the AMD Ryzen 7 9850X3D, will launch on January 28. Previous rumors had suggested this launch date, and now Micro Center has confirmed it. On AMD China's JD storefront, the Ryzen 7 9850X3D is already listed with a preorder option, requiring an 80 Yuan deposit, although the final price has not been disclosed. This 8-core/16-thread processor is powered by the "Zen 5" microarchitecture, enhanced with 3D V-Cache technology, and offers a speed increase over the current 9800X3D. The chip has a base frequency of 4.70 GHz and a maximum boost frequency of 5.60 GHz. Some samples have even been seen running at a 5.75 GHz boost frequency, indicating that enthusiasts might achieve even higher frequencies under regular home conditions. Our late 2024 review crowned the Ryzen 7 9800X3D as the world's best gaming processor. However, we need to determine how much of a difference the extra 400 MHz out-of-the-box overclock will make in gaming tests so we can draw more conclusions. Until third-party reviews arrive, we will have to wait.

AMD creates the top X670E and X870E chipsets by joining two AMD "Promontory 21" FCH chips at the hip. Each of these takes in a PCI-Express 4.0 x4 connection, and puts out 8 PCIe 4.0 lanes and four SATA 6 Gbps ports that can be re-configured as four additional Gen 3 lanes. The X870E/X670E chipset hence has a total of 12 PCIe 4.0 lanes and up to eight SATA 6 Gbps ports on the platform. A company decided to put one of these chips on an add-on card, so you could upgrade your single-FCH platform, such as B650, B850, or X870, with more PCIe lanes wired out as additional M.2 slots, and some SATA ports.

How long can AMD put the shield up against the pricing chaos?

Along with a British system integrator who's putting it inside a workstation.

But it may become slightly cheaper to buy AMD's fastest integrated Radeon GPUs.

In the current state of PC affairs, those who want to get the best gaming performance with the latest hardware should opt for AMD processors with stacked 3D V-Cache. Right now, the highest-spec model is the 16-core Ryzen 9 9950X3D with 128MB of level-3 cache (L3), but AMD is about to take this chip to a new level.

CUDIMM is on the way to AM5.

This mod proving that FSR4 Redstone works with RDNA 3 has got some gamers up in arms with AMD's limitation.

Is there a line of AMD CPUs which will not include AI components, or is it looking as if every new CPU is going to have AI optimized pipelines?

For several years now, I've been avoiding Intel: AMD seems to have kept an edge in multiprocessing, with more cores and threads per dollar, and I've felt Intel has been mismanaged for over a decade. AMD still feels like an engineering company run by engineers, vs a purely profit motivated company managed by sales and marketing wonks.

So I've been casually browsing around looking for a new laptop, and have noticed that all the AMD based ones are marketed as AI CPUs. I'd much rather have a CPU which doesn't waste silicon on AI customizations, so I'm wondering if AMD is a dead-end for me, or if AMD has a CPU family which doesn't pander to the LLM hype?

“Redstone” is a promising mix of old and new ideas, but Nvidia is years ahead.

Increased clock speeds at no extra wattage.

cross-posted from: https://mander.xyz/post/42900599

AMD’s next round of Radeon GPU price hikes is starting to take shape, with a new report from Chinese Board Channels claiming that board partners have already been told to prepare for higher costs. According to the post, several AMD graphics card brands have notified their channels that the “first wave” of increases will add around $20 to 8 GB models and $40 to 16 GB models, with retail prices in China expected to climb by roughly 300 RMB and 600 RMB respectively by the end of the year. The poster also claims there will be “no new products” launched through 2027, though that part is impossible to verify at this stage.