10-10-2026, 09:07 PM
Appreciate the detail. That matches what I saw on B200. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.
Mostly CUDA, sometimes wrong, usually measured
