4090 vs 5090 for 70B quants on mini PC

[fusedhall]

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?

I'm especially interested in mistakes worth avoiding.

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

[paperrunner]

Not sure I agree there. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Worth testing the simpler explanation before adopting the more interesting one.

[embersmith]

Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. The surprise was that memory pressure showed up before throughput collapsed.

[rocmriley]

Appreciate the detail. That matches what I saw on 4090. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[fusedhall]

following this because I'm in the same boat.

[sanevault]

I think the bottleneck is somewhere else. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[cudapoint]

A couple questions before I guess too hard. Do you see the same thing with a smaller batch or shorter prompt?

[vrambandit]

Ran a quick check on my box. My result was mixed: first run looked bad, second run looked normal. I expected the model choice to matter more than it did.

[cudapoint]

Not sure I agree there. I've seen people blame batch size when the real issue was much more mundane. Worth testing the simpler explanation before adopting the more interesting one.

[sanevault]

that fixed it for me too.