Can anyone explain really what is the behavior with the callbacks/events and the host thread? (best practices, how to avoid deadlocks) I need fine grained control and the fastest performance, and it looks like is callbacks, but they have weird behaviors... Well, I made the set up (cards, risers powered by sata/molex), I put 3 cards by PSU, installed win 10 and started mining on claymore. I get a 1200W power consumption, and a 145mh/s speed which is good for me. Well, it mined for 40min and I got a blue screen "thread stuck in device driver" .. Well I said to myself, it perhaps about windows so I installed Ubuntu and amdgpu-pro.. Well, same thing, 6 cards detected but I get a watchdog error "GPU2 Hangs in OpenCL call" or "GPU4 Hangs in OpenCL call" the miner try to reload but I get a frozen screen again.. Today I tried some switch between risers and PCIE slots I got the same error on same card (I identified the card with a 100% fanspeed settings on claymore).. I tried to remove completely the GPU2 so 5 gpu mining but i get the "GPU4 Hangs in OpenCL call" error when I try this. It work good with 4 cards only (even if the gpu4 is one of the 4 cards used). So I don't understand.. Could it be a PSU problem ? Not enough power ? It was a little bit more stable on windows I think. In Windows, Claymore did not go down, so when I looked up the log, Windows sometimes restarted the program. But I found a difference with Linux. < Claymore Restart Log - Windows > 17:59:47:926 f80 WATCHDOG: GPU 4 hangs in OpenCL call, exit 17:59:47:942 f80 watchdog - thread 10 (gpu5), hb time 297 17:59:47:957 f80 watchdog - thread 11 (gpu5), hb time 156 17:59:47:957 f80 watchdog - thread 12 (gpu6), hb time 296 17:59:47:973 f80 watchdog - thread 13 (gpu6), hb time 172 17:59:47:988 f80 watchdog - thread 14 (gpu7), hb time 219 17:59:48:004 f80 watchdog - thread 15 (gpu7), hb time 78 17:59:48:254 f80 OC v7, Reset control for GPU 0, close miner right now if you want to use default control from Catalyst 17:59:48:270 f80 OC v7, Reset control for GPU 1, close miner right now if you want to use default control from Catalyst 17:59:48:270 f80 OC v7, Reset control for GPU 2, close miner right now if you want to use default control from Catalyst 17:59:48:270 f80 OC v7, Reset control for GPU 3, close miner right now if you want to use default control from Catalyst 17:59:48:442 f80 OC v7, Reset control for GPU 4, close miner right now if you want to use default control from Catalyst 17:59:48:442 f80 OC v7, Reset control for GPU 5, close miner right now if you want to use default control from Catalyst 17:59:48:457 f80 OC v7, Reset control for GPU 6, close miner right now if you want to use default control from Catalyst 17:59:48:457 f80 OC v7, Reset control for GPU 7, close miner right now if you want to use default control from Catalyst 17:59:49:583 f80 Restarting OK, exit... In other words, there seems to be no gpu reset process on linux. So, in linux, restarting seems to be abnormal. |
