AWQ vs GPTQ vs GGUF right now on A6000

[paperrunner]

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

I'm not trying to build a perfect architecture diagram. I just want a sane starting point around llama.cpp and dual 3090.

I'm especially interested in mistakes worth avoiding.

[torchshed]

worth checking the driver first.

[southtinker]

worth checking the driver first.

Did a small comparison this afternoon. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, not quality. At this point the real bottleneck is my patience.

[llmarch51]

What exact setup are you running? Is this local-only or are you routing anything externally?

[turboscope]

What exact setup are you running? Is this local-only or are you routing anything externally?

following this because I'm in the same boat.

[vramscope]

Not sure I agree there. I get why that seems right, but on 4090 the bottleneck felt elsewhere.

[marbletrack]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at router policy. Not glamorous advice, but repeatable baselines save a lot of time.

[nanobrook]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at router policy. Not glamorous

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.

[fastbench]

[quote]My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at router policy. Not gla

I'd approach it in a boring order. If you're on SGLang, I'd confirm the version first and then reduce variables until the behavior is obvious. Once you know the plain path works, then add the faster path back in piece by piece.

[llmarch51]

That matches what I was seeing. I ended up in about the same place using PyTorch.