I have built a cryptocurrency mining rig and the specs are as follows:
Asrock 970 extreme4 mobo
AMD Sempron 145 cpu
G.Skill Ripjaws 2x2GB DDR3-1600 ram
3x Sapphire HD 7950 gpu
Seasonc Platinum 860w psu
running headless with Xubuntu 12.10 and cgminer 3.7.2 for mining
I began this build with an Asrock 990fx extreme3 mobo, but after about a week, it sparked and burned by one of the internal usb headers that had nothing plugged into it. It was fried. I RMAed the board and while I am waiting to hear about that one I bought the Asrock 970 extreme4. Got everything installed and running fine. The next day, the machine was off. Power supply wouldn't spin up when I turned it off and on again. I disconnected everything but the CPU and it powered up just fine. Added RAM, everything fine. Added one GPU to the first slot (I am using 16x to 16x unpowered riser cables to keep the cards further apart from one another), flipped the switch and nothing. Tried a different riser cable in the same slot, same result. Moved the GPU to the second slot and it starts up, but 1 second later the GPU sparks ans releases a burning smell, and I quickly shut down the computer. At this point I figure I've found the culprit. A bad GPU. I try another GPU in the first slot and it works fine. I add the second to the second slot and everything seems to be working. I leave it running and during the night at some point it turns itself off again. I unplug the GPUs and it starts just fine. I figure I have a bad PCIe16x slot. I am afraid it might have been the reason the card went bad as well, and now I am worried that the second card that was plugged into that slot might also have been ruined. I have yet to test the cards in other slots since this last failure, but plan to do that later today. The question is, do I risk putting the third card in that first slot or should I just call it quits and RMA the board (and the fried GPUs of course)?
I have also tested the PSU with a multimeter and the readings are within the acceptable ranges for the different pins, so this doesn't seem to be a PSU problem.
FYI. I was using well accepted mining settings for these specific cards which have been tested and used by many people in the mining community. My power draw at the wall is 800w with all 3 GPUs and 550w when I only had 2 of them working. These numbers also been confirmed and used under continuous operation by many miners.