Glossary · Term

B200

← all terms

Definition

Plain language

A current top-end NVIDIA chip for training and running large AI models.

As stated in the literature

NVIDIA's Blackwell-architecture data-center GPU, a generation beyond the H100 and H200, used in large-scale training and serving.

Why it matters: The chip you train on sets how fast and how cheaply you can build and run big models, so newer hardware directly raises what's feasible.

For example, a lab training a large model might fill a data center with thousands of these chips running in parallel for weeks.

Heard on the show

“Twenty frames per second, live, on a single Nvidia B200 GPU — all four players' views at once.”
Episode 206 — How Four-Second Clips Become Hours of Playable AI Soccer

Mentioned in 2 episodes

  1. 206
    How Four-Second Clips Become Hours of Playable AI Soccer
  2. 033
    Echo: The Paper Arguing You Never Needed a KV Cache for Retrieval

Related terms