All About BTC, LTC, ETH mining as well as other alternative crypto currencies
The developer of the Ethminer fork with Nvidia CUDA support (source) Genoil has released another update and we have compiled a new Windows binary of ethminer with CUDA support. Do note that this Windows binary release is compiled with VS2013 for windows 64-bit and is for CUDA 6.5. The latest version comes with some optimizations and a new command line option cuda-schedule to experiment with that replaces the old cuda-turbo, also note that you may need to manually specify the number of GPUs to use if you have multiple video cards using the cuda-devices command line parameter if the miner fires only on one device by default. Additionally to get better performance you can try adding the following command line parameters to the ethminer:
--cuda-grid-size 8192 --cuda-block-size 128 --cuda-schedule auto
--cuda-scheduleSet the schedule mode for CUDA threads waiting for CUDA devices to finish work. Default is sync. Possible values for mode are:
auto– Uses a heuristic based on the number of active CUDA contexts in the process C and the number of logical processors in the system P. If C > P, then yield else spin.
spin– Instruct CUDA to actively spin when waiting for results from the device.
yield– Instruct CUDA to yield its thread when waiting for results from the device.
sync– Instruct CUDA to block the CPU thread on a synchronization primitive when waiting for the results from the device.
It seems that the auto mode for cuda-schedule works best for us providing maybe a bit lower maximum hashrate, but a more stable one than the sync that may produce higher maxes, but also lower. You are free to experiment what will work best on your mining configuration however. The Ethminer CUDA fork should work on Compute 2.0 or newer GPUs, but the performance on older GPUs can be worse, also don’t forget that you can run Ethminer in OpenCL mode as well on Nvidia-based video cards and not only on AMD if you are having trouble with the CUDA support or the hashrate you get is lower as compared to OpenCL.