quant support gaps: what are you seeing?

[batchcraft]

Spent some time on this today and wanted to compare notes.

I'm comparing two approaches here: keep it simple on B200, or squeeze more out of the stack and deal with the complexity.

Interested in tradeoffs more than fan loyalty.

[syshouse]

Appreciate the detail. I don't think your conclusion is far off.

[kernelmike]

A couple questions before I guess too hard. Are you testing this on B200 or something else?

Appreciate the detail. 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.

[kernelgrid]

I think the bottleneck is somewhere else. I've seen people blame reranking when the real issue was much more mundane. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.

[kernelmike]

Can you fill in a bit more context? Are you testing this on 5090 or something else?

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. Not glamorous advice, but repeatable baselines save a lot of time. The logs were somehow calmer than I was.

[kernelmike]

Appreciate the detail. I don't think your conclusion is far off.

following this because I'm in the same boat.

[batchcraft]

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at prefill/decode split. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[northcraft90]

Need a little more detail here. Is this local-only or are you routing anything externally?

[rasterbench]

I hit almost the same issue last night.

[mangotinker73]

Can you fill in a bit more context? Are you testing this on 5090 or something else?

If you post your exact settings, people here can probably narrow it down pretty fast.