Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
simple local serving recipe on GB200
#1
Did a bit of testing and thought I'd post the result here.

I tried this with Llama using SGLang on Threadripper box. The headline result was better than I expected, but tokens/sec still ended up being the limiting factor.

Happy to post more details if anyone wants them.
M3 Max • MLX • local-first
Reply
#2
What worked for me was pretty simple. On dual 3090, the easiest win is often to back off the aggressive settings and see whether the problem survives. Not glamorous advice, but repeatable baselines save a lot of time.

I'm happy to test one more variant if you want a second data point.
Reply
#3
Quote:What worked for me was pretty simple. On dual 3090, the easiest win is often to back off the aggressive settings and see whether the problem survives.

Ran a quick check on my box. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.
7900 XTX / ROCm / somehow still optimistic
Reply
#4
Quote:[quote]What worked for me was pretty simple. On dual 3090, the easiest win is often to back off the aggressive settings and see whether the problem su

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at tokens/sec.
learning in public
Reply
#5
A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with FLUX and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.
curious and caffeinated
Reply
#6
I'd start with the simpler stack.
mostly local
Reply
#7
What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Phi and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.
Reply
#8
Appreciate the detail. That matches what I saw on B200.
Reply
#9
Maybe, but I had the opposite result. I've seen people blame KV cache pressure when the real issue was much more mundane. Worth testing the simpler explanation before adopting the more interesting one.

I'm happy to test one more variant if you want a second data point.
throughput matters, but tail latency matters too
Reply
#10
A couple questions before I guess too hard. Is this local-only or are you routing anything externally?
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)