RTX 5090 local inference for Llama

[batchwire]

I'd approach it in a boring order. On A6000, the easiest win is often to back off the aggressive settings and see whether the problem survives.