Why Local Inference Is 'Inefficient' — and Why That Doesn't Matter
mike64_t · x · 2026-10-30
mike64t explains the physics of local inference: loading weights from memory to the processor costs 100x more than the actual flops, so 100 users sharing one load cuts costs 50x. But he pushes back on centralization conclusions — by that logic we'd all be on time-sharing mainframes. In practice inefficiency buys quality of life and flexibility (venvs, static linking), and idle GPU capacity will be tolerated just like idle desktop CPUs.
More from Infra
- You can fingerprint your exact GPU with a single torch.index_add_ call — CatAstro_Piyush · 2026-10-31
- SGLang boosts DeepSeek V4.1 Flash throughput 67% to 1.09M tok/s per GPU in 4 days — BanghuaZ · 2026-10-31
- llama.cpp ROCm regression: prompt processing drops from ~1200 to ~460 t/s on RX 9070XT — uber-linny · 2026-10-31
- Sub-scale neocloud warning: VC-backed GPU cloud startups reportedly struggling — AccBalanced · 2026-10-31
- Qwen4 with HC architecture could widen the Q4 vs Q6 quantization gap significantly — Ok-Shower7286 · 2026-10-31
- Too many inference providers, too little demand: a developer's take — AccBalanced · 2026-10-31