Feature request: Hashrate throttling

I run a dual GTX 1070 setup in my desktop computer and I haven't found a way to throttle the hash rate of ethminer. It would be nice to be able to set an upper limit (around 80-90%) for allowable GPU resources so that I can also carry on with daily tasks but not suffer screen freezes.

[StefanOberhumer]

Adding some related issues:

• A few thoughts from enduser #315
• #1063

[DekInc]

@StefanOberhumer i think a sleep call will do,
if the user set the parameter --UseLazyness 500
after the search function can be a little sleep
std::this_thread::sleep_for(std::chrono::milliseconds(s_UserLazyness));

i don't know... just a suggestion (?)

[Vodochnik]

Would be nice....

[nwoeanhinnogaehr]

I was able to get significantly better system responsiveness without reducing hash rate by reducing the grid size to 256, block size to 32, and inserting a call to usleep(0); in the main search loop. For even better responsiveness you can reduce the grid size further, but this also reduces the hash rate for me.

[jonreiter]

i added flags for a tunable usleep on cuda on #2281 . that, with the grid/block size params, is enough to manually tune this in a minute or two and get a usable desktop at 90%+ the full hash rate.

[Rybec]

Disappointed to see this was never implemented. I have an AMD video card that is just barely hitting thermal shutdown, requiring a reboot, and improving the cooling isn't an option for me right now. If I could throttle it to 80-90%, it would likely run perfectly fine, without losing too much.

[jonreiter]

@Rybec possible to try my fork ref'ed above? My goal was X server responsiveness not thermals...but would be curious thermals as they are the biggest pain to test.

[Rybec]

I'll have to take a look, once I get my Linux setup going again. An update trashed it, and reinstalling the AMD and Nividia compute drivers made no difference, so I can't even mine on it anymore. I'm currently mining in Windows, where the AMD card is only doing ~6Mh/s, which isn't enough to raise the temperature more than a few degrees. The AMD card is being used both for graphics and mining, so I'm not messing with BIOS hacks, clock rates, or voltages. In Linux, running on stock settings, it gets 22Mh/s, in Windows it gets a mere 6. I might install Adrenaline and set it to compute mode, but I'm afraid I'll run into the same thermal issues. So my plan is to totally reinstall my Linux setup and go from there. If it is still having thermal issues after that, I'll take a look at your fork! (My goal is around 16-18Mh/s, at least until I am able to come up with another solution to the thermal issues. I'm afraid I might have to go for water cooling to keep temperature under control, and if that's the case, I'm not sure it will be worth it for a mixed use card.)