when TensorRT-LLM clearly wins: early impressions

[vectorport]

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 5090 felt surprisingly usable once I tuned around workstation thermals.

Maybe useful as a starting point for newer folks.

This is mostly for local work rather than public serving.

[rocmstack25]

A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with Qwen and compare from there.

[sysstation69]

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at INT4. Once you know the plain path works, then add the faster path back in piece by piece.

[turbosupply83]

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at INT4. Once you know the pl

Appreciate the detail. That matches what I saw on 5090. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[pixelcrate82]

My first guess would be this: If you're on llama.cpp, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[relayforge]

A couple questions before I guess too hard. Do you see the same thing with a smaller batch or shorter prompt?

[vectorport]

Thanks, that helps. I don't think your conclusion is far off. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[kernelmike]

Need a little more detail here. Are you testing this on dual 3090 or something else?

[sysstation69]

Thanks, that helps. That matches what I saw on Threadripper box. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[kernelmike]

Thanks, that helps. That matches what I saw on Threadripper box. Worth rechecking your defaults though, because one bad assumption can make the rest o

worth checking the driver first.