Used A6000 worth it now

[cacheline]

What worked for me was pretty simple. On 4090, the easiest win is often to back off the aggressive settings and see whether the problem survives. Once you know the plain path works, then add the faster path back in piece by piece.

[rocmriley]

Did a small comparison this afternoon. On B200 with LM Studio, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.

[signaltrack44]

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