Tensor cores actually helping here on dual 3090

[kernelmike]

Thought I'd post a small recipe because a lot of threads ask similar things.

This isn't revolutionary, but the combination of SGLang and Threadripper box felt surprisingly usable once I tuned around long context.

Maybe useful as a starting point for newer folks.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.

[kernelmike]

I tested this tonight. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, not quality.

[phaseport36]

What exact setup are you running? What model, what runtime, and roughly what context size?

I'm happy to test one more variant if you want a second data point.

[sparrowbrook]

same here on my side.

[audioriver]

I hit almost the same issue last night.

[emberpeak]

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at NVFP4.

[phaseport36]

Interesting, I hadn't tried that. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

There are a few different conversations hiding inside this thread, so being explicit about the goal helps.

[kernelmike]

Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly. I expected the model choice to matter more than it did.

[phaseport36]

What exact setup are you running? Do you see the same thing with a smaller batch or shorter prompt?

[dockember]

Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly.