industry move that matters for local users

[audiopoint18]

Posting this in case it saves someone else a little time.

If you were starting from scratch today for this use case, what would you choose first and what would you ignore?

Simple and boring is a valid answer.

[audiopath66]

I think the bottleneck is somewhere else. I get why that seems right, but on GB200 the bottleneck felt elsewhere. Worth testing the simpler explanation before adopting the more interesting one. The logs were somehow calmer than I was.

[phaseshed]

I think the bottleneck is somewhere else. I get why that seems right, but on GB200 the bottleneck felt elsewhere. Worth testing the simpler explanatio

That matches what I was seeing. I don't think your conclusion is far off.

[audiopoint18]

Did a small comparison this afternoon. My result was mixed: first run looked bad, second run looked normal. The main thing that moved was latency, not quality.

[bytestack]

that fixed it for me too.

[dockbrook21]

Maybe, but I had the opposite result. I get why that seems right, but on GB200 the bottleneck felt elsewhere. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.

[audiopoint18]

Tried to reproduce it just now. I couldn't reproduce the slowdown until I pushed context much higher. The main thing that moved was latency, not quality.

[southpatch]

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

[edgecoast]

My first guess would be this: Before rebuilding anything, I'd test one clean baseline with Mistral and compare from there. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[dockriver]

same here on my side.