0001-01-01
CUDA, Vulkan, ROCm, and OpenVINO: The Local AI Backend Guide
Which compute backend should you use for consumer local AI? When to use CUDA, Vulkan via Mesa, ROCm hipBLAS, or Intel OpenVINO, explained in depth.
GUIDES
Field guides
0001-01-01
Quantization Explained: The Speed, Memory, and Quality Tradeoffs
How does LLM quantization work, and what do k-quants mean in GGUF? Real empirical benchmark data from our lab's full Q4/Q6/Q8 sweep, explained.
0001-01-01
The Local AI GPU Buyer's Guide: Tier Lists, VRAM Targets, and Value Recommendations
How do you choose a graphics card for local language models in late 2026? Budget $300 entry cards, 24GB workstation value champions, and flagships.
0001-01-01
Will It Fit? The Complete VRAM Calculation Formula for Local LLMs
Will your GPU run that model? The formula for calculating LLM memory requirements: static weights, KV cache context scaling, and runtime overhead.