Best GPUs for Local LLMs

Graphics cards, VRAM guidance, inference hardware, and local AI build advice for running models at home.

All GPUs & AI Hardware Articles

Four practical lower-cost GPUs for local AI, compared by VRAM, memory bandwidth, software compatibility, power, and buying risk.

Current inference GPUs compared by VRAM, board power, cooling, software support, physical fit, and used-market risk.

Current GPU recommendations for Ollama and llama.cpp, ranked by VRAM, memory bandwidth, power, software support, and buying risk.

Current GPUs for Ollama compared by VRAM, software path, power, used-market risk, and model fit—not unsupported tokens-per-second claims.

How to buy a used GPU for local LLM inference without getting burned. Five cards ranked by value with pricing, red flags, and testing procedures.

Step-by-step guide to GPU passthrough on Proxmox with VFIO. Covers IOMMU setup, driver blacklisting, VM configuration, and running local LLMs.

A practical guide to VRAM requirements for local LLMs. Model sizes from 7B to 120B+, quantization levels, context length, and GPU picks.

Apple unified memory and NVIDIA CUDA compared for local LLMs by memory capacity, software support, model fit, power, expansion, and risk.

A current comparison of NVIDIA CUDA and AMD ROCm for Ollama, llama.cpp, PyTorch, and home inference, including RDNA 4 support.

A lifecycle-aware comparison of two 24GB NVIDIA GPUs for used and legacy buyers, with current alternatives for new purchases.

RTX 5060 Ti 16GB versus used RTX 3090 for local inference: current specs, VRAM limits, efficiency, warranty, and used-card risk.