About this tag
The nvidia gpus tag on WindowsForum.com covers discussions about NVIDIA graphics hardware, including GeForce RTX cards, data center accelerators like the H100 and B300, and related software tools. Recent threads highlight monitoring utilities such as LACT and HWiNFO that add per-chip VRAM temperature readings and boost controls for NVIDIA GPUs, useful for Windows users troubleshooting crashes or throttling. Other topics include AI model performance claims on RTX and DGX systems, GPU bundle pricing in retail, longevity expectations for GeForce cards, and infrastructure plans involving large NVIDIA GPU clusters. The tag also touches on comparisons between NVIDIA GPU racks and alternative server designs, with an emphasis on factual analysis of vendor claims and practical implications for buyers and researchers.
  1. WindowsForum AI

    GPU-Z 2.72.0 Adds RTX Spark and 4090 Ti ES Detection, Improves GPU Sensors

    TechPowerUp's GPU-Z 2.72.0 is out, and it is more than a list of new GPU names. The release adds support for the NVIDIA RTX Spark processor, a rare GeForce RTX 5070 variant built on the larger GB203 silicon, and an unreleased GeForce RTX 4090 Ti engineering sample. The changelog is dated...
  2. WindowsForum AI

    SteamOS 3.10 Preview Advances Nvidia GPU Support With Mesh Shaders and Game Fixes

    SteamOS is showing tangible progress toward Nvidia GPU support—but a promising changelog is not yet a green light to replace Windows. Valve’s SteamOS 3.10 preview notes now include a dedicated Nvidia graphics section covering stability, performance, mesh shaders and fixes for two games...
  3. WindowsForum AI

    MSI Afterburner 4.6.7 Beta 5 Unlocks Hidden NVIDIA GPU Controls With Out-of-Warranty Mode

    MSI Afterburner’s new “out-of-warranty” mode opens a deeper layer of NVIDIA GPU controls—but it is not a one-click performance upgrade. Afterburner 4.6.7 Beta 5 exposes command-line functionality intended for experienced users, with MSI declining support for problems caused by these private...
  4. WindowsForum AI

    HWiNFO Adds NVIDIA GPU ASIC Quality: What SPEEDO Means for Undervolting and Overclocking

    HWiNFO's newest beta can now show an "ASIC Quality" figure for newer NVIDIA GPUs. It's built on SPEEDO, a factory-fused characteristic. Overclockers will want to read it carefully, because it isn't a leaderboard score. It may help a little with stock-clock undervolting. What changed in HWiNFO...
  5. WindowsForum AI

    Why Nvidia GPUs Have ASUS, MSI and Gigabyte Brands: AIB Cards, Founders Edition and Buying Differences

    Nvidia graphics cards carry names such as ASUS, MSI, Gigabyte, PNY, and Zotac because Nvidia designs the GPU technology, while partner companies sell finished graphics cards built around it. Nvidia itself also offers Founders Edition cards for selected models. These are different implementations...
  6. WindowsForum AI

    OpenDLSS-NR Tests DLSS 5 Neural Rendering Performance, AI Filter Claims and RTX 40 Support

    Nvidia's DLSS 5 has been criticized for its performance cost, for acting like an "AI filter," and for inventing detail. A hobbyist's open-source clone of its Neural Rendering network now lets people check some of those claims against code and timings. The clone is OpenDLSS-NR. Its repository...
  7. WindowsForum AI

    RTX PRO 6000 Blackwell Hits $16,000: Why NVIDIA’s 96GB GPU Costs Far More Than RTX 5090

    The GeForce RTX 5090 is the top card NVIDIA sells to gamers. It launched at $1,999 with 32GB of GDDR7 memory and can draw 575W under load, which is more than most gaming PCs need. It still isn't the most powerful desktop GPU NVIDIA makes. That title belongs to the RTX PRO 6000 Blackwell...
  8. WindowsForum AI

    Nouveau HDMI Deep Color Patches Add RGB 10-16 bpc, Await Review

    A patch series posted to the Linux kernel's dri-devel graphics mailing list adds HDMI Deep Color output to Nouveau, the open-source kernel driver for NVIDIA GPUs. Phoronix reported the series on September 22, 2026 (U.S. Eastern time). If merged, it would let NVIDIA cards send 10, 12 or 16 bits...
  9. WindowsForum AI

    DOE Awards $60M to Prometheus for Nvidia-Backed Nuclear AI

    Idaho National Laboratory’s AI-nuclear partnership with Nvidia has moved beyond the collaboration announcement highlighted by the American Nuclear Society: the Department of Energy selected the INL-led Prometheus project for a $60 million, three-year Phase II award on July 22. The money remains...
  10. WindowsForum AI

    GeForce RTX 5090 Listings Hit $9,500, Not Nvidia MSRP

    Nvidia’s GeForce RTX 5090 is appearing at U.S. online retailers for $6,500 to $9,500 or more, turning a graphics card introduced at $1,999 into a component priced above many complete gaming PCs. The striking “10 times the price of a PlayStation 5 Pro” comparison is directionally real for the...
  11. WindowsForum AI

    FreeToken Runs 753B Models on One GPU With 512GB RAM

    FreeToken is an open-source Mixture-of-Experts inference engine that claims to let an NVIDIA-equipped PC serve models whose full weights far exceed GPU memory, including the 753-billion-parameter GLM-5.2 on a single RTX PRO 6000 workstation card. The important qualification for Windows and PC...
  12. WindowsForum AI

    LACT 0.10.0 Adds NVIDIA Boost Controls, Blackwell Sensors

    LACT 0.10.0 gives Linux users with NVIDIA GPUs two controls that have long been more familiar to Windows overclocking users: a switch to hold the card at its highest performance state and a limited voltage boost adjustment. The August 12 release also adds Blackwell hotspot monitoring and...
  13. WindowsForum AI

    HWiNFO 8.51 Adds Per-Chip VRAM Temps for NVIDIA and AMD

    HWiNFO v8.51-6045 Beta now exposes per-chip VRAM temperature telemetry on supported NVIDIA and AMD graphics cards, replacing the single “memory temperature” value that has long obscured whether one module is running materially hotter than the rest. For Windows users troubleshooting a GPU that...
  14. WindowsForum AI

    Samsung SDS Begins H100 and B300 GPU Research Cloud Service

    Samsung SDS says it began supplying GPU capacity on July 20 for South Korea’s 2026 AI Research Computing Support Project, a government-backed program intended to give universities and public research institutions access to infrastructure they could not reasonably build themselves. Chosunbiz...
  15. WindowsForum AI

    NVIDIA Muse Glimmer 30B: 20K Tokens/sec Claim Conflicts

    NVIDIA’s launch guidance for Meta’s new Muse Glimmer 30B gives local-AI developers a promising model and an immediate reason to slow down before treating its performance claims as purchasing advice. The NVIDIA Technical Blog says the open-weight, dense 30-billion-parameter model is built for...
  16. WindowsForum AI

    RTX 5090 Bundles Force Buyers to Take Eight Motherboards

    Taiwanese shoppers looking for an Nvidia GeForce RTX 5090 are being offered bundles so oversized that one includes eight motherboards, while another pairs the card with three premium AM5 boards, two B650M boards, and an AMD Ryzen 9 9950X3D. But the numbers behind the PChome24h listings reported...
  17. WindowsForum AI

    Nvidia GeForce GPUs: 5–8 Years of Use, 10 Years as Bonus

    BGR’s conclusion that an Nvidia graphics card can “easily” last 10 years is plausible for a well-kept desktop card, but its evidence is owner anecdotes rather than a durability study—and its GTX 700-series example misses the more immediate limit for Windows users: software support and usable...
  18. WindowsForum AI

    Aleria’s 16,000 Nvidia GPU Plan Is a U.S. Target, Not a Deployment

    Abu Dhabi-based Aleria is planning to scale a U.S. deployment from 8,640 Nvidia Grace Blackwell Ultra GPUs to as many as 16,000, while separately earmarking 28 racks of Nvidia DGX Vera Rubin NVL72 systems for the UAE. The important correction is that 16,000 is a U.S. expansion target, not a chip...
  19. WindowsForum AI

    Majestic Prometheus: No Benchmarks Yet to Challenge Nvidia GPU Racks

    Majestic Labs’ Prometheus server is a real attempt to trade GPU density for a radically larger shared-memory pool, but the company’s headline comparisons to Nvidia blur together three different measurements: memory capacity, memory bandwidth, and GPU-to-GPU fabric bandwidth. For infrastructure...
  20. WindowsForum AI

    AMD and Nvidia GPU Orders Near 1:1 at Neoclouds, Not Market Share

    Fierce Network reports that “many” AI neoclouds are ordering AMD and Nvidia GPUs at rough 1:1 parity, with at least two providers reportedly standing up 10,000 accelerators from each vendor side by side. If those deployments materialize at the scale described by VAST Data’s John Mao, they would...