Hi All. I am a newbie and have been mining for about 3 weeks. I'm using Claymore v9.3 on Windows 10 64 Bit. My issue is that if I put any sort of stress on the machine at all, It crashes. And it is consistently "GPU 3 hangs in opencl" that causes the crash. What I mean by stress is if I try to dual mine anything, increase the difficulty setting, or overclock the GPU's.
I have 6 XFX RX 480 8gb Black Edition cards. I have modded the BIOS on all 6 cards copying the 1750 timing strap to the 2000. I am using the Crimson 16.12.2 video drivers. The most stable setup I have is when the difficulty is set to 8 on cards 0,1,2,4,5 and 7 on card 3. Using MSI Afterburner I set the voltage to 0, the core clock to 1,100 and the memory clock at 2,000. The fans on all 6 cards is set to 85%.
With all the above settings I am getting aproximately 27.7 mh/s from each card and a total of 166 mh/s. It takes approximately 14 hours to mine .1 ether using ethermine.
I would really like to dual mine and/or get my cards up to 30 mh/s but like I said, anything I try causes GPU3 to hang and restarts the machine.
As for the rig, It is open air with a HP 1200 watt power supply for the six cards and a Thermaltake 850 watt power supply for the mother board and PCI Risers. Riser 1 and 2 are powered by daisy chain 6 pin cable, as are risers 3 and 4. (2 cables, 2 plugs each). Risers 5 and 6 are powered by 6 pin to SATA adapters connected to a SATA daisy chain cable.
Any theories why I can't push this setup any further?
Thank you for reading.
Brian
I have had the same issues many a time, and it is almost always a function of memory voltage on the card. Try running -ethi 4 on that card, you won't loose much on the hash rate but it will lower the load. Do that before you do anything else because that may solve it. Also, if you are undervolting the mem clock, then you need to raise it back up. However, if you are not undervolting the memory voltage, then lower the mem clock speed until the GPU is stable. Focus on the memory voltage and clock speed and you will find a spot that will resolve this. It is fairly easy to solve but takes some trial and error. My experience is that it is almost never the PCI riser (I have had one go bad out of 200 over the past 6 months) and almost always voltage/clock speed issue.