Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Vector Databases discussion on Threadripper box
#11
I don't think that's the whole story. The reported numbers can look compelling, but in ordinary usage I didn't see the same benefit. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.
2x3090 • llama.cpp • too much VRAM is still not enough
Reply
#12
I tested this tonight. On M3 Max with SGLang, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)