tokens/sec tuning ideas: what are you seeing?

[circuitkite]

Did a bit of testing and thought I'd post the result here.

I tried this with Phi using SGLang on 4090. The headline result was better than I expected, but workstation thermals still ended up being the limiting factor.

Curious if anyone else landed in the same place.

[stackpilot]

My first guess would be this: Before rebuilding anything, I'd test one clean baseline with Qwen and compare from there. Not glamorous advice, but repeatable baselines save a lot of time.

[audiopoint18]

What worked for me was pretty simple. On B200, the easiest win is often to back off the aggressive settings and see whether the problem survives. Not

Tried to reproduce it just now. I couldn't reproduce the slowdown until I pushed context much higher. I expected the model choice to matter more than it did.

[turbosupply83]

My first guess would be this: If you're on TensorRT-LLM, I'd confirm the version first and then reduce variables until the behavior is obvious. Once you know the plain path works, then add the faster path back in piece by piece.

[tinycontext]

[quote]What worked for me was pretty simple. On B200, the easiest win is often to back off the aggressive settings and see whether the problem survive

that fixed it for me too.

[vramport]

Appreciate the detail. That matches what I saw on Threadripper box. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[vrambandit]

A couple things I'd check first. If you're on LM Studio, I'd confirm the version first and then reduce variables until the behavior is obvious. Once you know the plain path works, then add the faster path back in piece by piece.

[torchstation]

[quote][quote]What worked for me was pretty simple. On B200, the easiest win is often to back off the aggressive settings and see whether the problem

I tested this tonight. On mini PC with SGLang, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.

[turbosupply83]

What worked for me was pretty simple. On B200, 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.