serving on Blackwell: what are you seeing?

[routecrate]

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

[macrotinker]

I hit almost the same issue last night.

I'm happy to test one more variant if you want a second data point.

[routecrate]

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

[cobaltlab]

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.

[batchcraft]

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

[kernelgrid]

I hit almost the same issue last night.