Epoch AI’s new index says FLOP/s badly undersells modern AI chips

Share

Epoch AI launched version 1 of its Inference Performance Index on November 9, 2026, measuring the inference capacity of agentic AI systems rather than their raw computing power alone. According to the index, Nvidia’s GB300 delivers approximately 25 times the inference capacity of the H100 while offering only about 2.5 times more eight-bit FLOP/s. Agentic inference capacity grew 4.6 times from Q2 2025 to Q2 2026, compared with 2.6 times growth in raw compute, with Nvidia’s Blackwell GPUs accounting for 74% of that increase. Normalized against the GB300 NVL72, the ranking places AMD’s MI455X first with a score of 2.89, ahead of Nvidia’s VR200 at 2.18, while the B300 HGX and MI355X are virtually tied. The index also factors in memory bandwidth, inter-chip communication and performance with long context lengths, but relies on simulations based on Nvidia’s AISimulate tool.

Source: Read the original article

Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
Telemac
Telemachttp://cryptoinfo.ch
Passionné de nouvelles technologies, j’explore l’univers de la blockchain et des cryptomonnaies pour partager l’actualité et les innovations du secteur.

Read More

Items