Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
PCIe lane weirdness with inference loads
#11
Tried to reproduce it just now. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.
Reply
#12
Quote:Tried to reproduce it just now. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up b

Need a little more detail here. Do you see the same thing with a smaller batch or shorter prompt?
Mostly CUDA, sometimes wrong, usually measured
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)