I have to say the same.  With my Nvidia 9800GT, poclbm OpenCL maxes at 25 MH (30 if I use -f 0), whereas rpcminer-cuda gives 18 MH if I'm lucky - and I've played with the settings a lot.

[oh and rpcminer-opencl doesn't work for me at all.  It never actually connects after initialising.  In windows task manager I just see rpc-opencl use more and more RAM, until eventually it crashes(!)]

This to me doesn't make sense - CUDA is more low-level and optimized for Nvidia, is it not? Surely it should run faster than OpenCL?


P.S. hello all - first time poster, long time lurker