Homelab census: 41 containers, one 6 GB GPU, and where my AI agents run
A homelab census of 41 Docker and 9 LXC containers, and how one 6 GB GPU decides which AI agents run on local Ollama, OpenRouter or a subscription.
#AI Hardware
5338 articles tagged with AI Hardware
GPU memory arithmetic, model licences and a three-year cost comparison for an on-premises HSE platform that never sends ...
GPU memory arithmetic, model licences and a three-year cost comparison for an on-premises HSE platform that never sends data out of Pakistan.#ai #llm #selfhosted #architecture #sof...
Meta has open-sourced the implementation codebase for its Muse AI agent, allowing developers and third-party hardware bu...
Meta has open-sourced the implementation codebase for its Muse AI agent, allowing developers and third-party hardware builders to construct standalone devices.The release includes ...
AI · Google: Starship needs 1,800 flights for space data centersGoogle's first orbital AI chip test worked, but it says ...
AI · Google: Starship needs 1,800 flights for space data centersGoogle's first orbital AI chip test worked, but it says the plan needs Starship flying at a pace no rocket has ever ...
RT @bkuyper: Ich arbeite an einer neuen Inference-Engine, die sich ausschließlich auf AMD-Hardware konzentriert. Da das ...
RT @bkuyper: Ich arbeite an einer neuen Inference-Engine, die sich ausschließlich auf AMD-Hardware konzentriert. Da das Patchen von vLLM keine langfristige Lösung war, ist dies noc...
Prediction: This Artificial Intelligence (AI) Chip Stock Will Make a Big Move in October (Hint: It’s Not Micron)
Micron Technology delivered impressive results for the fourth quarter of fiscal 2026 on Sept. 30, indicating that the artificial intelligence (AI) infrastructure boom is alive and ...
Sarawak-designed, locally made chip heads to China for electric meters
By Amanda L MUKAH, Oct 3: A semiconductor chip designed by a Sarawakian and produced by a Sarawak-based foundry has been purchased for use in electric meters to be marketed in Chin...
An AI hardware advisor that walks every path — and gets its math checked by the rules in Sanity
What I built An agent that answers "what computer do I need to run this AI model?" —...
Cheapest GPU rentals today (2026-10-03), by VRAM:32GB Tesla V100 $0.124/h (Vast.ai)48GB Q RTX 8000 $0.261/h (Vast.ai)80G...
Cheapest GPU rentals today (2026-10-03), by VRAM:32GB Tesla V100 $0.124/h (Vast.ai)48GB Q RTX 8000 $0.261/h (Vast.ai)80GB A800 PCIE $0.428/h (Vast.ai)96GB RTX PRO 6000 S $1.33/h (V...
GPU 없이 소비자용 노트북에서 180억 파라미터 LLM을 구동하는 POCKET-Darwin-180B. 4비트 GGUF 양자화로 360GB→111GB 압축, 약 $1,400 하드웨어로 로컬 추론 가능.#ai #mac...
GPU 없이 소비자용 노트북에서 180억 파라미터 LLM을 구동하는 POCKET-Darwin-180B. 4비트 GGUF 양자화로 360GB→111GB 압축, 약 $1,400 하드웨어로 로컬 추론 가능.#ai #machinelearning #opensource #llm #software #coding #development...
Why Upstage Builds Mini AI Models That Run on a Single GPU
Upstage released Solar Mini 4, a 35-billion-parameter model that can run on a single GPU when quantization is applied, targeting firms short on GPU resources.
A small GRPO experiment: ISO-AdamW scored 75.8% versus AdamW's 75.4% on GSM8K, with 47% more GPU memory. The idea, the 4...
A small GRPO experiment: ISO-AdamW scored 75.8% versus AdamW's 75.4% on GSM8K, with 47% more GPU memory. The idea, the 44x kernel detour, and what four answers mean.#machinelearnin...
Amazon and Synopsys ink multi-year billion-dollar deal in multi-year IP agreement to accelerate AI chip design effortsAm...
Amazon and Synopsys ink multi-year billion-dollar deal in multi-year IP agreement to accelerate AI chip design effortsAmazon and chip design tool maker Synopsys have inked a multi-...
juanjo-barrientos/genaiguide: Guía gratuita de IA generativa avanzada en español (311 págs.): Transformer, post-entrenamiento, inferencia, GPU, difusión, agentes, seguridad. Free advanced generative AI guide in Spanish. CC0.
Guía gratuita de IA generativa avanzada en español (311 págs.): Transformer, post-entrenamiento, inferencia, GPU, difusión, agentes, seguridad. Free advanced generative AI guide in...
DaragonTech/DLaya: Fast local AI decisions in Delphi and Free Pascal. A binding for LibLayaX that runs the Laya typed-decision model in-process: no server, no Python. One unit, Laya.pas, with a TLayaAgent class. Windows, Linux and macOS, CPU or GPU, 64-bit. An open-source alternative to Jev
Fast local AI decisions in Delphi and Free Pascal. A binding for LibLayaX that runs the Laya typed-decision model in-process: no server, no Python. One unit, Laya.pas, with a TLaya...
DaragonTech/LibLayaX: Fast local AI decisions inside your application. A shared library and C API built on laya.cpp that runs the Laya typed-decision model in-process: no server, no Python. Windows, Linux and macOS, x64 and ARM64, CPU or GPU. An open-source alternative to Jev
Fast local AI decisions inside your application. A shared library and C API built on laya.cpp that runs the Laya typed-decision model in-process: no server, no Python. Windows, Lin...
Google tar sikte på rymden. Nyligen sköt sökjätten upp fyra ai-chip ombord på en Space X-raket. Men drömmen om stora dat...
Google tar sikte på rymden. Nyligen sköt sökjätten upp fyra ai-chip ombord på en Space X-raket. Men drömmen om stora datacenter i omloppsbanan kan ligga långt bort.#tech #google #a...
Centre fortifies India’s AI mission with Sovereign Compute Pool amid rising GPU costs
The official said that the government is also preparing further bids for additional computing capacity, as it seeks to build a government-controlled pool of AI compute alongside ca...
Steam Deck 2: Is AMD Gainsborough the chip Valve’s been waiting for?The Steam Deck LCD’s “Aerith” APU. | Photo by Fritzc...
Steam Deck 2: Is AMD Gainsborough the chip Valve’s been waiting for?The Steam Deck LCD’s “Aerith” APU. | Photo by Fritzchens Fritz The Steam Deck is four and a half years old, and ...
vLLM Scheduling: throughput vs latency in shared inference.The GPU is shared; what decides how many fit is memory. Prefi...
vLLM Scheduling: throughput vs latency in shared inference.The GPU is shared; what decides how many fit is memory. Prefill, chunked prefill, KV cache.📖 https://devopstales.github.i...
🛰️ Google Invia Satellite AI Nello SpazioGoogle lancia un satellite equipaggiato con le sue TPU (Tensor Processing Units...
🛰️ Google Invia Satellite AI Nello SpazioGoogle lancia un satellite equipaggiato con le sue TPU (Tensor Processing Units) come parte del Project Suncatcher, un'iniziativa speriment...
🎮 I hate having to prove my brand loyalty just for the chance to buy hardware at full priceMy kingdom for a reasonably p...
🎮 I hate having to prove my brand loyalty just for the chance to buy hardware at full priceMy kingdom for a reasonably priced and readily available bit of kit.📰 Source: Latest from...
M6 Mac mini review: Excellent hardware badly hampered by too-small SSD
The M6 Mac mini has a faster chip and is solid hardware, but that $899 hurts and there is just not enough storage at that entry price for Apple Intelligence and all of your apps. M...
96dot/SDR-to-HDR-Video: A Chromium extension that converts ordinary (SDR) web video to HDR in real time on your GPU. It is a hand-tuned inverse tone mapper written as WebGPU shaders: no machine learning, no driver hooks, and nothing leaves your machine.
A Chromium extension that converts ordinary (SDR) web video to HDR in real time on your GPU. It is a hand-tuned inverse tone mapper written as WebGPU shaders: no machine learning, ...