LLM-leaderboard-style atlas benchmarks silicon photonics foundries on component specs
jwt0625 · x · 2026-11-08
A WIP project called Photonics Atlas Benchmark applies the LLM-leaderboard format to silicon photonics foundries, comparing publicly available specs from imec, GlobalFoundries, Intel, Samsung, Tower, TSMC, IHP and others across waveguide loss, coupler insertion loss, detector bandwidth, modulator bandwidth and more. The author notes metrics aren't necessarily achieved simultaneously on the same chip, and most big commercial foundry data isn't public.
More from Infra
- Anthropic API users beware: cache 5-minute window can make bundled credits hilariously wasteful — teortaxesTex · 2026-11-08
- Harbor turns one: 200T tokens/week, 10M+ Terminal-Bench runs, 1B+ rollouts — BenBlaiszik · 2026-11-08
- GPU ran 16s→11s per image after cooling fix: thermal throttling unlocked — elittle1234 · 2026-11-08
- GLM-5.3-SLIM GGUF quantization trends on Hugging Face for llama.cpp — Scantist · 2026-11-08
- Nadella on datacenter water use: closed-loop cooling equals one restaurant's yearly usage — pbaylies · 2026-11-08
- Jane Street Is 'a Commune with GPUs,' Says SemiAnalysis Founder Dylan Patel — dylan522p · 2026-11-08