I know it can be done manually but some people here have dozens of GPUs of variable model, brands and bins.
The thing is, when a GPU run unstable the computer can freeze, and waiting for it rebooting take t0o much time.
A GPU can become unstable and freeze the OS after hours or days of work and, guess what, that take t0o much time !
These work conditions are not humane, programmers, help us out !
I do not expect it to be perfectly reliable, as long as it can help find the average stability threshold automaticly. (I'd downclock by 30-50 mhz and be done with it)
Features I'd like to see :
Start overclock at X
Time vs fine tuning trade-off
Continue on reboot (obviously)
Learn from previous profile, (same model)
Down voltage mode : find lowest voltage at a given speed.
It could be mining while it's stability tested and on reboot show dialog : "your computer rebooted unexpectedly, would you like to downclock your gpu by 5 / 10mhz / NO " (one GPU per machine for simplicity)
I think a USB bootable Linux version would be preferable over a windows version.
Realizable ? / time frame ? / bounty size ?
Thanks for your input.
Using what I already have setup in BAMT, this is probably about a day's work. At current prices, bounty of 75 BTC and I'll be happy to provide it.