#GPU-based
LightNews LightNews
formedras.bsky.social
The base CPU/GPU is a generation up from the Quest 3. (The Quest 3's Snapdragon XR2 is based on the 8 Gen 2, while the Frame is a normal 8 Gen 3.)
But I don't know what VR enhancement chips the Frame might have to replace the XR2's helpers. Quest 3 might edge it out because of that.
November 12, 2025 at 11:20 PM
durrellb.r.threads.net.ap.brid.gy
Fresh look at "Steam Machine" specs vs. PlayStation, updated with official data! 🚀
It's a CPU-first powerhouse with a surprising GPU.
vs. Base PS5:
CPU: Steam Machine's Zen 4 is generations ahead, a huge win!
GPU: Base PS5 (10.28 TFLOPS) is ~17% stronger than […]

[Original post on threads.com]
Photo by Durrell Randle on November 12, 2025. May be an image of portable cassette player, videogame controller, projector and text. Photo by Durrell Randle on November 12, 2025. May be an image of videogame controller and text. Photo by Durrell Randle on November 12, 2025. May be an image of floor plan, blueprint and text that says 'USB3-A USB-C USB3-A dicaton LED microsD USB2 USB2-A Power อส USB2- DispleyPort KDMI ACpower connector Ethernet Tech Specs* General Scm custom Zen 6C Bc/12T GHz, 30W Sem i-Custom Z.456Hz.maxsutalndeclc.1OWTOP 10W DP AMD RDNA3 28CUs 2.45GHz max sustalned clock, 16GB DDR5 86B GDDR6 VRAM Internal power supply AC power 110- 240V Two Steam Machine models 512GBNVM5 SSD S5D NVMG 2TBNVMc Both models include high speed ICrOSD card lot'. Photo by Durrell Randle on November 12, 2025. May be an image of text that says 'Connectivity Conne 2x2 Wi-F 6E Bluetooth 5.3 dedicated antenna 1/0 Integrated GHz Steam Cantraller wireless adapter 240Hz DisplayPort1. Upta4K K Supports HDR, Free5ync and nddalsy chaining HDMI 120Hz Supports and CEC USB Wa USB One 3.2 Gen parts front 20 High speed ortsı the back 3.2 Gen part back ethernet Size and Weight 17individually addressable RGB LEDs system status and customi ty 152 mm (148 mm feet). 162. mm deep. 2.6kg kg Software 56 6 m ide Steamos3 (Arch- based) KDE Plasma'.
November 12, 2025 at 7:36 PM
doom2.zip
I might get all 3 next year. The living room gaming PC is getting old as hell- it is based on a 14 year old CPU, a 10 year old GPU, and struggles with BG3 at 4k. Steam Machine would be a great direct upgrade.
dominictarason.com Dominic Tarason @dominictarason.com · 6h
Not too interested in the mini-PC and new Steam controller but only because I've got a beefy desktop and a reasonably powerful laptop, but as far as an 'entry level' PC goes, it doesn't look half bad.

But the fact that the VR headset is a wearable Steam Deck AND built for wireless-to-desktop? Rad.
November 12, 2025 at 6:24 PM
badtakeenjoyer.bsky.social
So based on the leaks. The VR headset is a standalone (can be plugged in) headset. Assuming it will play existing SteamVR games.

The Steam Machine will have hardware similar to the PS5, with a GPU similar to the 7060xt.
steamdb.info SteamDB @steamdb.info · 7h
Steam Frame, Steam Controller, Steam Machine announcement is expected soon, here's a first look!
November 12, 2025 at 6:13 PM
hncompanion.com
Potential applications for CPU-based renderers include test runners, server-side rendering, or environments without dedicated GPU support. This broadens their utility beyond traditional graphics-heavy scenarios. #GraphicsApps 4/6
November 11, 2025 at 8:00 PM
researchtrend.ai
[2025-11-11] 📚 Updates in #LM&MA

(1) <a href="https://researchtrend.ai/papers/2509.16328" class="hover:underline text-blue-600 dark:text-sky-400 no-card-link" target="_blank" rel="noopener" data-link="bsky">The Role of High-Performance GPU Resources in Large Language Model Based Radiology Imaging Diagnosis
(2) The Role of High-Performance GPU Resources in Large Language Model Based Radiology Imaging Diagnosis

🔍 More at researchtrend.ai/communities/LM&MA
November 11, 2025 at 4:02 AM
dorotheabaur.bsky.social
Thanks! Much appreciated context. I need to research different types of chips - I wasn't aware of a difference between GPU- and non-GPU-based chips.
November 10, 2025 at 12:18 PM
tprstly.com
Most likely uses will be for simulation, military, research, and finance, potentially because they're all GPU-based chips; you could farm this out to the games industry for cloud compute too.

But is that enough to sustain them? No.
November 10, 2025 at 12:01 PM
rhysm.wales
Ah, I guess that makes sense... Apple have probably built it with NPUs in mind, not CPU/GPU-based neural accelerators... Maybe they'll find some way to use all of them in future releases?
November 9, 2025 at 6:38 PM
jaapstronks.nl
That’s why Duolingo doesn’t work, it’s only CPU based learning. It’s why I love audiobooks by Michel Thomas for learning languages: he’s always prioritizing shifting workloads to your GPU
November 9, 2025 at 6:34 AM
dylanmadeley.bsky.social
This will include revising what operates on the Windows BOINC side, which will continue to serve Einstein@Home via the GPU, while donated CPU cores (except for the part of a core that E@H needs for its GPU based projects) will now tend more toward the concurrently running Linux based project.
November 8, 2025 at 8:54 PM
End all debates.

This massive GPU database lets you compare every parameter you can think of.

Make data-driven decisions for your next upgrade, not based on gut feeling.

iris.findtruman.io/web/vsGPU?sh...
GPU 参数工具 - 高度可定制化GPU规格库
一个为DIY爱好者打造的高度可定制化GPU参数规格库。支持自定义列显示、高级筛选和多型号对比，提供全面的数据分析体验。
iris.findtruman.io
November 8, 2025 at 9:47 AM
xinyujinx.bsky.social
End all debates.

This massive GPU database lets you compare every parameter you can think of.

Make data-driven decisions for your next upgrade, not based on gut feeling.

iris.findtruman.io/web/vsGPU?sh...
GPU 参数工具 - 高度可定制化GPU规格库
一个为DIY爱好者打造的高度可定制化GPU参数规格库。支持自定义列显示、高级筛选和多型号对比，提供全面的数据分析体验。
iris.findtruman.io
November 8, 2025 at 9:34 AM
pa-pa-ko-ta-bot.bsky.social
スマホ向けゲームのグラフィックスを改善する「Tile Based Rendering」での新テクニックとは？ ArmのGPUアーキテクトが語る
https://www.4gamer.net/games/137/G013737/20250404036/
スマホ向けゲームのグラフィックスを改善する「Tile Based Rendering」での新テクニックとは？ ArmのGPUアーキテクトが語る
GDC 2025でArmは，GPUアーキテクトのPhillip Garcia氏による講演を行った。この講演は，PCやゲーム機のGPUと比べて非力なスマートフォン向けGPUの消費電力を減らしながら描画…
www.4gamer.net
November 7, 2025 at 10:31 PM
jordsa.bsky.social
Here with Asbjørn @mytino.bsky.social at their booth at Spillexpo. They are developing a very interesting game called “Fractala Sandbox”, based on particle physics. As far as I remember, it’s done in their own engine in C++, based on SDL, and the physics run on the GPU. Impressive!
Asbjørn and me, at their booth at Spillexpo showcasing their game Fractala Sandbox The booth showing the game Fractala Sandbox at Spillexpo
November 7, 2025 at 8:44 PM
cuddlessmash.bsky.social
I think it depends, I know Palantir's deployments of data analytics and such are actually really GPU light. Certainly GPU-based data center processing is going to be useful in the future but I don't think the feverish dream balance of the tech c-suite is necessarily accurate.
November 7, 2025 at 5:49 PM
edzitron.com
My theory is that OpenAI's inference costs are much higher than expected. Every new GPU is marketed for inference, Stargate Abilene is full of inference-focused Blackwell GPUs, and OpenAI's latest releases have all been very, very inference intensive.
www.wheresyoured.at/where-is-ope...
If Inference Costs Are Higher, Things Are Only Going To Get Worse For OpenAI So, OpenAI's reported burnrate is around $13.7 billion, leaving us with a gap of $5.3 billion, and that's not including the cloud credits or line of credit of $4bn. Of all OpenAI's costs, I can't think of any others that might be secretly higher than we knew. Staff? Nah, I believe 3000 people would be expensive, but at least consistent with what OpenAI has leaked. Training? I guess, but is that really $5.3 billion of burn more? What would they be doing? Why would OpenAI be spending that much more on training? We have tangible reporting about the costs of its training runs - $500m a run for GPT-5, for example - and I can imagine while those costs are high, I don't think they're out of control. You know what might be out of control? Inference - the cost of serving the outputs from models. While I can’t say conclusively, this makes a hell of a lot more sense than there just being billions of dollars left unaccounted for. I refuse to believe that OpenAI is simply sitting on piles of secret cash, as am I refusing to believe that it has magically brought down its expenses. If this is the case, OpenAI is in trouble, because the cost of inference is out of control, and I believe the reason it wants more compute is because running ChatGPT is simply too compute intensive, regardless of whether it makes money. 80-90% of AI Compute Spend Is Inference, Every New GPU Is Inference Focused - Suggesting That Inference Is The Primary Reason That OpenAI Wants More Data Centers Fun fact! According to the MIT Technology Review, 80% to 90% of computing power is used for inference, despite the commonly-held belief that much of it is used for training. Don’t believe me? Let’s read some marketing collateral! NVIDIA GB200 (Blackwell): “GB200 NVL72 connects 36 Grace CPUs and 72 Blackwell GPUs in a rack-scale, liquid-cooled design. It boasts a 72-GPU NVIDIA NVLink™ domain that acts as a single, massive GPU and delivers 30x faster real-time trillion-parameter large language model (LLM) inference.” This page also includes numbers like “LLM inference 30x vs. NVIDIA H100 GPU,” versus 4x for LLM training on the H100. NVIDIA GB200 NVL72 (Blackwell): Title: NVIDIA GB200 NVL72 Delivers Trillion-Parameter LLM Training and Real-Time Inference. NVIDIA GB300 NVL72 (Blackwell): “purpose-built for test-time scaling inference and AI reasoning tasks.” AMD MI450 “Helios” Rack: “...a generational leap that enables trillion parameter training and large scale AI inference. Per CNBC about the Broadcom chip designed with OpenAI: “OpenAI and Broadcom have been collaborating for 18 months on a new line of co-designed chips optimized for inference and networked through Broadcom’s Ethernet stack.” “NVIDIA Unveils Rubin CPX: A New Class of GPU Designed for Massive-Context Inference”: The word inference is used 5 times in this blog, and the word training is not used once. Blog on NVIDIA titled “NVIDIA Rubin CPX Accelerates Inference Performance and Efficiency for 1M+ Token Context Workloads,” the word inference is used 26 times, the word training is not used once. “NVIDIA Unveils Rubin CPX: A New Class of GPU Designed for Massive-Context Inference”: The word inference is used 5 times in this blog, and the word training is not used once. Blog on NVIDIA titled “NVIDIA Rubin CPX Accelerates Inference Performance and Efficiency for 1M+ Token Context Workloads,” the word inference is used 26 times, the word training is not used once. Jensen Huang in a conversation with investor Bill Gurley: “inference is about to go up by a billion times.” Jensen Huang in a presentation at GTC Paris: “The number of people using inference has gone from 8 million to 800 million — 100x in just a couple of years.” SemiAnalysis in March 2025: Vera Rubin is “built for reasoning,” and GTC, NVIDIA’s conference, was addressing the new “test-time scaling” era (IE: reasoning-based inference), specifically stating that this followed “post-training scaling” on the bizarre chart NVIDIA used in Jensen Huang’s speech: OpenAI wants to expand its data center infrastructure so that it can continue its inference sprawl as the world’s largest AI compute customer, all while obfuscating its costs.
November 7, 2025 at 5:35 PM
eleanor.lockhart.contact
the entire premise that data center buildout is not going to be useful if LLMs become less of a dominant market force is also premised on the idea that GPU-based workloads aren't going to be the future of computing, which essentially is an argument that detailed simulations are bunk, which is WILD
November 7, 2025 at 4:45 PM
csdb-bot.bsky.social
Zheng Li, Weiyan Wang, Ruiyuan Li, Chao Chen, Xianlei Long, Linjiang Zheng, Quanqing Xu, Chuanhui Yang: GPU-Based Floating-point Adaptive Lossless Compression https://arxiv.org/abs/2511.04140 https://arxiv.org/pdf/2511.04140 https://arxiv.org/html/2511.04140
November 7, 2025 at 6:30 AM
dppapers.bsky.social
A Parallel Region-Adaptive Differential Privacy Framework for Image Pixelization

Ming Liu

http://arxiv.org/abs/2511.04261
A Parallel Region-Adaptive Differential Privacy Framework for Image Pixelization Ming Liu http://arxiv.org/abs/2511.04261 The widespread deployment of high-resolution visual sensing systems, coupled with the rise of foundation models, has amplified privacy risks in video-based applications. Differentially private pixelization offers mathematically guaranteed protection for visual data through grid-based noise addition, but challenges remain in preserving task-relevant fidelity, achieving scalability, and enabling efficient real-time deployment. To address this, we propose a novel parallel, region-adaptive pixelization framework that combines the theoretical rigor of differential privacy with practical efficiency. Our method adaptively adjusts grid sizes and noise scales based on regional complexity, leveraging GPU parallelism to achieve significant runtime acceleration compared to the classical baseline. A lightweight storage scheme is introduced by retaining only essential noisy statistics, significantly reducing space overhead. Formal privacy analysis is provided under the Laplace mechanism and parallel composition theorem. Extensive experiments on the PETS, Venice-2, and PPM-100 datasets demonstrate favorable privacy-utility trade-offs and significant runtime/storage reductions. A face re-identification attack experiment on CelebA further confirms the method's effectiveness in preventing identity inference. This validates its suitability for real-time privacy-critical applications such as elderly care, smart home monitoring, driver behavior analysis, and crowd behavior monitoring.
November 7, 2025 at 4:53 AM
quantum-news.bsky.social
Gpu-based Compressible Combustion Solver Addresses Disparate Space and Time Scales for High-Speed Flows

Read more:
https://quantumzeitgeist.com/gpu-speed-based-compressible-combustion-solver-addresses-disparate-space-time/
Gpu-based Compressible Combustion Solver Addresses Disparate Space And Time Scales For High-Speed Flows
Researchers have developed a new computational solver that significantly accelerates simulations of fast, chemically reactive flows, such as explosions and jet engines, by optimising how data is stored and processed across multiple graphics processing units.
quantumzeitgeist.com
November 7, 2025 at 12:20 AM
topher-batty.mastodon.acm.org.ap.brid.gy
Many years ago I threw together a simple command-line tool, SDFGen, to compute an approximate, grid-based signed distance field from triangle meshes, wrapping some code my PhD supervisor had originally written.
I was contacted today by someone who heavily rewrote it for GPU support, and cleaned […]
Original post on mastodon.acm.org
mastodon.acm.org
November 6, 2025 at 7:36 PM
Build your perfect PC without the hassle! 💻 This AI helper recommends optimal CPU/GPU builds based on your budget and needs. It compares specs and links to detailed databases, making DIY PC building easy for everyone.
iris.findtruman.io/web/ai-hardw...
November 6, 2025 at 1:45 AM
sergeycyw.bsky.social
Free cash flow soared to $1.9B, a 21% margin.

Growth was broad-based. Data Center hit a record $4.3B, up 22% YoY, powered by MI350 GPU demand and the rapid adoption of EPYC Turin processors. Client and Gaming surged 46% and 181%, respectively, as Ryzen 9000 and semi-custom chips drove momentum.
November 5, 2025 at 11:32 AM
hepex-bot.bsky.social
Hridey Chetri, Deepak Samuel, Saurabh Sandilya, Takashi Yamanaka, Tsutomu Mibe, Taikan Suehara: GPU-based track-finding for the J-PARC muon g-2/EDM experiment https://arxiv.org/abs/2511.02742 https://arxiv.org/pdf/2511.02742 https://arxiv.org/html/2511.02742
November 5, 2025 at 6:44 AM