Blackwell deployment chatter on A6000

[southpatch]

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

This isn't revolutionary, but the combination of MLX and B200 felt surprisingly usable once I tuned around prefill/decode split.

Maybe useful as a starting point for newer folks.

I care more about stability than squeezing the last possible tokens/sec out of it.

[routecrate]

What worked for me was pretty simple. If you're on TensorRT-LLM, I'd confirm the version first and then reduce variables until the behavior is obvious.

[thetapath]

Can you fill in a bit more context? What model, what runtime, and roughly what context size?

[edgecoast]

Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, not quality.

[cobaltspark]

Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, no

I'd approach it in a boring order. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at NVFP4. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[latentgarden]

Did a small comparison this afternoon. My result was mixed: first run looked bad, second run looked normal. I expected the model choice to matter more than it did.

[amberpatch25]

Interesting, I hadn't tried that. I ended up in about the same place using TensorRT-LLM. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.