Dwarkesh Podcast

Snapshot · Dwarkesh Podcast

Reiner Pope – Chip design from the bottom up

Explore episode May 22, 2026

Where this was said

Cache vs scratchpad

At 1:08:20 · chapter starts 1:04:07

CPU cache is 100× faster than DDR but introduces non-deterministic latency. TPU scratchpad uses explicit software instructions for each memory tier, enabling deterministic latency. Groq and TPUs advertise this.

Similar snapshots