NVIDIA or Apple Silicon - Local

• https://obrienlabs.medium.com/running-reasoning-llms-like-the-deepseek-r1-70b-43g-locally-for-private-offline-air-gapped-259fa437da8f

Tokens/sec for various M series and NVidia Ampere, Ada and Grace Blackwell GPUs

20260120 - https://github.com/ObrienlabsDev/blog/issues/160

Google Cloud

Asus Z790 H for Intel 14900K running a dual RTX-A4500 NVLink bridge

• https://obrienlabs.medium.com/running-the-larger-google-gemma-7b-35gb-llm-for-7x-inference-performance-gain-8b63019523bb

Keep CPU Undervolt Protection off for 15% performance improvement under CPU AI Overclocking

LLM Foundation Model Training

• andrej.karpathy Nanochat on NVIDIA DGX Spark GB10 - https://github.com/ObrienlabsDev/foundation-transformer-llm/issues/7
• andrej.karpathy Nanochat on NVIDIA DGX Spark GB10 - https://github.com/ObrienlabsDev/blog/issues/170
• https://github.com/obriensystems/nanochat/blob/dev/runs/runcpu.sh#L79
• DGX Spark https://github.com/ObrienlabsDev/foundation-transformer-llm/issues/7
• Apple Silicon https://github.com/ObrienlabsDev/foundation-transformer-llm/issues/8
• Windows/WSL NVIDIA https://github.com/ObrienlabsDev/foundation-transformer-llm/issues/9 = https://github.com/karpathy/nanochat/issues/542