Having one GPU going down every couple of hours

[zoufou]

I'm running Ethos with a 6 GPU rig (RX 580 and Mobo is H81 Pro BTC R2.0).

After overclocking/undervolting, i get a satisfying hashrate at ~170 MH/s for 950 Watts (from the wall), GPUs are running under ~55°C temperature.

My actual problem is am getting the 6th GPU to go off randomly after a couple of hours, in claymore i get the following error:

WATCHDOG: GPU 5 hangs in OpenCL call ...

I've gone investigating the issue, here's my results:

• Changed the riser, the cables and the power cable of the riser: still same issue
• On Win10, the system could not detect the 6th GPU ! i can only see 5 GPUs just on system startup and not after running couple of hours of mining like on Ethos
• Cleared CMOS, same result
• Undervolted and underclocked down to 24 MH/s, still same result

Any idea guys ?

[1715331149]

1715331149

[1715331149]

1715331149

[1715331149]

1715331149

[zoufou]

And PSU is a Corsair HX1000W ...

[zukko]

I had the same problem with cryptonote mining, until I set a fixed difficulty on the pool. After that the problem disappeared.
The card fell off only on a high-difficulty shares.

[zoufou]

In my case the card is off on Win10 with no mining activity at all.

Even when i have it as unique plugged card to the board, it's still not visible on Win10 and run on Ethos for some time then crashes.

When i plug it to another PCIe port the issue is fixed, can i confirm a malfunctioning PCIe port ? is it normal it run only on Linux (ethos) and for a random time ?