Right now working on 0.22. That one will bring a switch to bring back the 0.13 work loop as an option. This allows all for which the old version works better to switch to newer version and to profit from the other fixes.


I can not recommend doing this on CPU. Because if you use the CPU a lot the mining speed of GPU will go down. And power efficiency of CPU is much worse. Even an GTX 1030 does more sol/s then any CPU I know (well maybe except for 16+ core server CPUs ^^)


It seems we have two bugs with NVidia Windows driver here. One that makes the GPU to shut down completely (thats with zero monitor detected) and the second when there is no 3D App running lolMiner is not detected as program that requires the GPU to stay on 3D clock rates. I do not know if Nvidia workstation drivers may help here... we require a deeper investigation of this.

Can you be more specific about the error? Best way is to run the miner with --list-devices to see the order of the OpenCL platforms. It should list at least two (one Nvidia CUDA, one Advances Micro Devices... ). Maybe even three if you have an Intel IGPU installed. Then best run two instances of the miner, one for the single GTX 1050ti and one for the RX cards (here all 4 can run in one instance). In the readme are some lines how to select platform and devices.

See the answer right above. With --device and --platform you can control the OpenCL driver which cards to use.


Can not guaranty that Fury will be fastest in future versions, but I did not forget AMD for sure. 0.3 or 0.4 will come with performance tweaks, lets see how much I can do

Still it is not perfectly bug free, but definitely a large step forwards. Will see if I can do better next versions