consumer GPU serving economics: anyone testing this?

[cudapeak]

Wanted to compare notes with people doing similar things.

I'm mostly working with DeepSeek in LM Studio on 5090. The part I'm trying to understand is NVFP4.

Not looking for the perfect answer, mostly what worked in real use.

[turboscope5]

Wanted to compare notes with people doing similar things.

that fixed it for me too.

[cudapeak]

A couple things I'd check first. If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious.

[vramport]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Mixtral and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.

[signalcraft14]

Need a little more detail here. Are you testing this on 4090 or something else?

[stackarch]

I'd approach it in a boring order. If you're on Ollama, 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.

[cudapeak]

Thanks, that helps. I ended up in about the same place using LM Studio.

[localpatch78]

My first guess would be this: If you're on vLLM, 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.

[southden]

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

[prompttinker44]

What exact setup are you running? Do you see the same thing with a smaller batch or shorter prompt?