Huang's compute logic: Amdahl's Law caps speedups, over-provision the slow parts
le_james94 · x · 2026-09-29
From the Stanford CS153 thread series: Jensen Huang's design logic runs through Amdahl's Law — if 5% of a step can't be sped up, no number of GPUs gets you past 20x overall. So over-provision the slow parts. Midha proposed tokens-per-watt as the key metric, and Huang agreed.
Related event: Google: Frontier AI Labs Prefer Double Capacity Over Five Nines(3 posts)→
More from Infra
- India's Sarvam launches Token Factory, hosting top open models on local infra — HarveenChadha · 2026-11-16
- Gemini 3.5 Flash now runs on air-gapped Google Distributed Cloud, offline up to 12 months — shashib · 2026-11-16
- Gemma.c: full Gemma 1/2/3 multimodal inference in a single C file, ~72 tok/s on CPU at INT8 — This-Purchase-3325 · 2026-11-16
- BIS Issues Interim Final Rule Addressing Remote Access to Advanced AI Chips — apples_jimmy · 2026-11-16
- Jailbroken PS4 runs a 9B open-source model at up to 43 tok/s by talking directly to its 2013-era GPU — AIFlow_ML · 2026-11-16
- Mobile AI Factory: agents auto-build on-device SLMs with 19x less GPU time — pmttyji · 2026-11-16