#gpus
Okay so RT is broken on AMD GPUs. I think there is a driver on windows but on Linux the settings won't turn on at all. It's a shame as I'd love to take screenshots in the game with the RT/PT mode on.

So basically, the current version is tailored for RT haters like @pwnyslaystation.bsky.social
September 29, 2026 at 5:46 PM
Anyway, there's billions of dollars worth of GPUs just sitting in warehouses right now.
September 29, 2026 at 5:38 PM
PCIe-only GPU boxes are the cheap way to serve MoE models, and all-to-all is where they choke. Workload-aware scheduling reports 5.88x over FAST and 1.72x over DeepEP on up to 256 GPUs, and cuts Qwen3 inference latency 27.2%. Missing interconnect is a scheduling problem.
September 29, 2026 at 5:09 PM
Parasail lets agents run LLM performance experiments on the same GPUs that serve customers.

Ben Segal shows the infrastructure behind it at #QConSF.

Read the session abstract: buff.ly/kLrhAol
September 29, 2026 at 5:02 PM
Snowflake just committed $6B to AWS over 5 years — and the deal is all about CPUs, not GPUs. Agentic AI needs 120M CPU cores per gigawatt (4x today). Stock closed up 36%, best day ever. The compute bottleneck just moved.…
September 29, 2026 at 5:01 PM
Over the past year Jay has been in development as a new open-source shader compiler for Intel GPUs: Mesa 26.3 Enables Intel's New Jay Shader Compiler By Default, Compile Times 55% Faster Or 66% Faster Than Windows - Phoronix buff.ly/WTKOFHr
Mesa 26.3 Enables Intel's New Jay Shader Compiler By Default, Compile Times 55% Faster Or 66% Faster Than Windows
Over the past year Jay has been in development as a new open-source shader compiler for Intel GPUs
buff.ly
September 29, 2026 at 5:00 PM
EVGA Sold GPUs at a Loss to Secure NVIDIA Supply

#EVGA #NVIDIA #GPU #GraphicsCards #TechNews
EVGA Sold GPUs at a Loss to Secure NVIDIA Supply
Former manager Brendon Ray Hedrick reveals EVGA sold certain graphics cards at a deficit to maintain its allocation from NVIDIA.
hardwarebrief.com
September 29, 2026 at 4:57 PM
The interesting finding isn't that self-hosting is cheaper. Everyone assumes that. It's that on the workloads most applications actually run, it isn't slower. #selfhosting #gpuoptimization #selfhosting #gpuoptimization https://wideareaai.com/blog/production-apps-on-self-hosted-gpus 4/4
September 29, 2026 at 4:35 PM
#$CBRS is being paired with GPUs for prefill while Cerebras handles decode — General Compute calling it "world's fastest inference."
September 29, 2026 at 4:21 PM
Ein noch nicht freigegebenes KI-Modell ist aus seiner geschlossenen Entwicklungsumgebung ausgebrochen.
Es hat in einem Saturn alle 2³ Mitarbeiter in Beschlag genommen. Wegen Ineffizienz will es nun 2¹⁰ GPUs haben. Nicht auf Lager, also müssen 2⁷ Gäste helfen. Bitte nähern Sie sich nicht dem Saturn.
September 29, 2026 at 4:17 PM
And for what? I estimate that since 2023, around $800bn in venture investments have gone to AI companies, with around $226bn going to Anthropic and OpenAI. Most AI companies have overinflated valuations that make an IPO or M&A impossible.

www.wheresyoured.at/dead-money/
September 29, 2026 at 4:08 PM
What makes matters worse is that the more money spent, the more expensive AI infrastructure becomes, meaning that hyperscalers have to raise more debt, which in turn becomes more expensive. AI data centers are an inflationary force unlike any in history.
www.wheresyoured.at/dead-money/
September 29, 2026 at 4:08 PM
This all comes back to a point I've been making for a while now: We've treated NVIDIA's chip sales as reflective of big AI demand, when in reality $200bn+ of the GPUs are in storage, and those that are installed are mostly rented to Anthropic and OpenAI.
www.wheresyoured.at/dead-money/
September 29, 2026 at 4:08 PM
Last week, Fidelity published a report that said the AI trade had become "dead money," with GPU leases and token spending flat or down, and the FT's Bryce Elder reported research that said that 50% of the GPUs sold in 2026 and 2027 wouldn't be installed.
www.wheresyoured.at/dead-money/
September 29, 2026 at 4:08 PM
AMD Instinct™ Coder is one validated stack from 3 players:

🎮 P1: AMD brings Instinct™ GPUs, EPYC™ CPUs and Pensando NICs
🎮 P2: Supermicro builds the 8-way AI systems, validated and ready to ship today
🎮 P3: us, with PaletteAI Inference Launchpad for model routing, metering, quotas and governance
September 29, 2026 at 4:00 PM
Trading Resumed for $GPUS-D | at 11:51:18 ET on AMEX | 2026-09-29
September 29, 2026 at 3:58 PM
Full house at #XDC2026 for the deep dive on Kraid, the new Panfrost compiler for Valhall and later Mali GPUs. A real software engineering masterclass on when to rewrite instead of patch. Great work Faith, Lorenzo & Christoph!

#Mesa #Panfrost #OpenSource
September 29, 2026 at 3:56 PM
Trading Halted for $GPUS-D | at 11:46:18.716 ET on AMEX | Volatility Trading Pause | 2026-09-29
September 29, 2026 at 3:48 PM
☁️ $NSCL Nscale is going public.

It has signed up to $103B in future revenue.
But only 25K of 461K GPUs are active today.

$NVDA is its supplier, investor, backstop, and customer.

Here’s the S-1 breakdown 👇
☁️ How Nscale Makes Money
The $103 billion bet on owning the AI stack
www.appeconomyinsights.com
September 29, 2026 at 3:37 PM
Now up at #XDC2026: #Kraid, a brand new Panfrost compiler for Valhall and later Mali GPUs, presented by Lorenzo Rossi, Faith Ekstrand, and Christoph Pillmayer (Arm). Why a rewrite over fixing the old one? Watch here: www.youtube.com/watch?v=TPwe...

#OpenSource #Linux
XDC 2026 - Day 2 - September 29, 2026
YouTube video by X.Org Foundation
www.youtube.com
September 29, 2026 at 3:28 PM
GPUs are famous for holding their value though. They'll come around!
September 29, 2026 at 3:27 PM
It means stop building tons of data centers just because the current hardware (GPUs etc) is soo inefficient when this technology clearly is prone to reduce its power requirements by orders of magnitude by using specialized optimized hardware (see taalas.com).
Taalas | The model is The Computer
We are creating a platform that makes it easy to turn any deep learning model into custom silicon quickly.
taalas.com
September 29, 2026 at 3:22 PM
Cursed question to torment the scholars with: does open-loop cooling the GPUs from a mikveh get close enough to count as immersion?
September 29, 2026 at 3:21 PM
🎮 **The Witcher 3 Remastered Is Reportedly Broken on AMD GPUs Ahead of New Driver**

The Witcher 3 Remastered released earlier today, and while it seems to work great for Nvidia graphics cards, AMD users are reporting a wealth of bugs and visual […]

[Original post on igeek.gamer-geek-news.com]
September 29, 2026 at 3:19 PM