RTX 5090 local inference on mini PC

[dockkite69]

This keeps coming up in conversation so I figured I'd start a thread.

One thing I keep noticing is that people underestimate KV cache pressure until they actually try to run this stuff for more than one afternoon.

Wondering if this is just where the whole space is heading now.

[cudascope]

worth checking the driver first.

[circuitwire58]

I hit almost the same issue last night.

[embersmith]

Thanks, that helps. The pattern around VRAM headroom has been pretty similar for me. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[circuitwire58]

worth checking the driver first.

[signaltrack44]

worth checking the driver first.

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

[fusedhall]

A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with Phi and compare from there. Not glamorous advice, but repeatable baselines save a lot of time.

[mangopeak72]

Can you fill in a bit more context? Are you testing this on M3 Max or something else?

[rocmriley]

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

[oxidegrove40]

Appreciate the detail. The pattern around tool calling has been pretty similar for me. The biggest change for me was simplifying the stack instead of adding one more optimization layer.