← All posts
🔩

Silicon

GPU selection, VRAM math, quantization tradeoffs, KV-cache tuning and cost-per-token analysis with reproducible numbers.