Im having the same issue though with grid size for the 3090, I just took #cuda cores/2 and used that. Not sure if its even close but otherwise I get the illegal memory access error and error about #'s missing.

GPU: GPU #0 NVIDIA GeForce RTX 3090 (82x0 cores) Grid(5248x128)
[3777.30 Mkey/s][GPU 3777.30 Mkey/s]

Switched over to vanitysearch, much much much faster!

Crap, posted on wrong page I guess. Using vanitygen.

So I’ve been following you guys for a while and have had excellent results with vanitygen as well as oclvantiygen-plus.

I’ve finally got my hands on a rtx 3090 and after many hours got it running in both programs but it’s only working at a rate 3 times better than my old gtx 1080.

My gtx 1080 was doing 100mkey/s
The new rtx 3090 is doing 380mkey/s

GPU utilization is only at 33%, am I missing something here or is this expected?