19:52
2026-08-29
zartbot.github.io
ai-chips
Redesigning the Inference Chip: From Nvidia GPU's Flaws to OpenAI Jalapeño
OpenAI's new inference chip, codenamed Jalapeño, eliminates the L2 cache and adds a superscalar scheduling core to address flaws in Nvidia's Blackwell microarchitecture, according to an analysis by Za…