search

sgminer-dev / sgminer

Scrypt GPU miner
GNU General Public License v3.0
635 stars 825 forks source link

Keccak switching may need some cool time? #382

Open miningpoolhub opened 10 years ago

miningpoolhub commented 10 years ago

Hi I am running http://miningpoolhub.com which switches mining coin automatically by current profitability. x11, x13, x15, nist5 switching works smoothly but when we switch between keccak and other algo, it makes GPU hardware DEAD time to time.

As I am running a pool, I hear many reports about this error and I think I am not the one who suffers this. I am mining with 280x, and many miners say that they mine with 280x also. (Of course, there's much more GPU types) They say that it seems like to crash on under volted GPUs more frequently than normal GPUs.

I don't think it can be solved just by fixing some deallocating memory or things like that. It looks like related to driver or hardware error. When we mine keccak for a long time, we can't mine scrypt immediately with it's full speed and needed some cool time. (Turn off computer, wait a while and reboot may solve it time to time)

I want to hear whether if it's solvable, known issue or it is just hardware related problem and need to be managed by human time to time. Want to hear more deatiled info also.

Thanks.

mrbrdo commented 10 years ago

I also experience it, but I do not know exactly why it happens (it might as well be switch from Keccak as you describe). It happens from time to time, not very frequently (once or twice a week on one of 4 rigs). Reboot solves it, but I found that system may hang with normal BAMT "coldreboot", I had to use:

echo 1 > /proc/sys/kernel/sysrq 
echo b > /proc/sysrq-trigger

I never had to wait between reboot or anything, just did that reboot and it worked fine again. My cards are Sapphire R9 280X Dual-X undervolted to 1.1V. I also have a few ASUS R9 280X but I don't think they ever froze (not sure though, and forgot if I undervolted them or not). Some rigs with only Sapphire cards never had the freeze though, and some had them much more frequently (mostly those with 5-6 cards, and less or never with 4 or less).

Also I do not remember having any of these freezes on the kernels compiled on the older AMD drivers (= slower for X* algos), like 14.2 which I had before. So possibly the new drivers could have something to do with it. I am also actually still running the old drivers but using .bin files from the new drivers. I think many people have similar setup due to problems with setting frequency on windows on new drivers. So this combination, or the new drivers might be a problem. Don't know really. Just wanted to confirm and add my experience. At the moment I am not mining so I cannot test or debug more.

platinum4 commented 9 years ago

I haven't ran into this problem and I lease out time on Keccak sometimes. It usually has to do with the rental stratum proxy, and people always pick Keccak and then point it to an incorrect pool like for Jackpotcoin.