Hardware & Sizing

VRAM math, memory bandwidth, Apple Silicon and how to pick a machine that fits the models you want.

1. 1▶ Up next12 min+50 XPVRAM math: will this model fit?The three terms that decide memory use (weights, KV cache, overhead), with the formulas, effective bits per weight by…
2. 211 min+50 XPMemory bandwidth: the speed limit of decodingWhy tokens per second is predicted by memory bandwidth, how to find and calculate bandwidth for any machine, worked…
3. 311 min+50 XPApple Silicon for local LLMsUnified memory, bandwidth by chip family, the GPU wired-memory limit, how much RAM you need for which models, Metal…
4. 4Pro14 min+50 XPHardware buying guide by memory tierHow to choose hardware for local LLMs: memory capacity first, bandwidth second, compute third. Tier table, GPU vs Mac…