If you have the latest Nvidia drivers installed and there's no "Optimize for Compute Performance" option, it is possible that this option is only in driver for Windows 10.

Anyway, I'm not sure if Windows 7 users need this option as they don't have a problem that Windows 10 users faces that the speed is only ~3.5MH/s.

Hi.

I was able to launch Genoil's miner on Windows 10 and GTX 970 with the latest drivers (382.53) by setting "Optimize for Compute Performance" to On for that app in Nvidia Control Panel as described here https://www.youtube.com/watch?v=NvD57PyEJz8

This gave me 20MH/s instead of ~3MH/s without the ned to downgrade the drivers. But this only works with Genoil's miner in OpenCL mode.

Guy in the video successfully runs Claymore's miner too.

For some reason for (not only) me with Claymore's miner and "Optimize for Compute Performance" enabled I get:

CUDA error 0 #102 - 46
CUDA error 0 #103 - 46
CUDA error 0 #105 - 46
...
CUDA error - cannot allocate big buffer for DAG. Check readme.txt for possible solutions.
...
GPU 0, CUDA error 46 - cannot write buffer for DAG
...

Here's the description of what "Optimize for Compute Performance" does
http://nvidia.custhelp.com/app/answers/detail/a_id/4370/~/what-does-the-setting-optimize-for-compute-performance-do%3F

What could be the reason?