non-AI thing you built recently

[localtinker]

My first guess would be this: On A6000, the easiest win is often to back off the aggressive settings and see whether the problem survives. A lot of th

That matches what I was seeing. That matches what I saw on Threadripper box. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[mlxgarden]

Appreciate the detail. The pattern around embeddings has been pretty similar for me.

[sparrowhall]

Appreciate the detail. The pattern around embeddings has been pretty similar for me.

That matches what I was seeing. That matches what I saw on dual 3090.