Cloudflare, Groq and Positron chase greener inference alternatives

Chip boffins are scrambling to cut the soaring energy costs of artificial intelligence, with a new generation of startups targeting Nvidia’s grip on the market by building faster, leaner, and more efficient inference silicon.

A PetaOp/s performance on a single chip

Tensor Streaming Processor (TSP) inventor Groq announced that its new Tensor Streaming Processor (TSP) architecture is capable of 1 PetaOp/s performance on a single chip.