Decred stuck at <5% GPU load about excavator

[nicehashdev]

That sounds as a driver crash in the middle + failures to launch kernel, for which we haven't been able to find a way to detect properly. We will take a look how to detect kernel failures.

[nirvanaosc]

It only happens with Decred though, if we don't restart excavator and we change again to another algorithm by removing/adding workers, excavator work flawlessly, but when it goes back to Decred, it happens again.

[nicehashdev]

This is the same issue that can happen with ccminer - launch failure that isn't reported as error so excavator keeps going on.

If your cards are OCed, reduce OC.

[nirvanaosc]

Ok, I found the issue(s).

There are two things here:

The first one is that if you

• start a worker with custom arguments (i.e. lbry with B=6144,TPB=768,NPT=16)
• stop the worker
• add a new worker on decred without custom arguments
  it will use the custom ones used in lbry. I assume it's because it's reusing the CUDA thread with the custom arguments, because if I add a new decred worker without stopping the lbry one (hence creating a new thread) it start it with default arguments. With this arguments decred doesn't work properly (I assume it's because TPB is out of range).

The second one is that it doesn't error when creating with arguments out of range.

[nicehashdev]

Yep, that could be it! Nice catch, thank you!

You can verify if this is the case by calling API algorithm.list which will also list workers and their parameters. If there is something you did not add, then this is trully the issue.

Will be fixed in next version.

[nirvanaosc]

Looks like algorithm.list always shows the parameters empty:

Add lbry workers with arguments B=6144,TPB=768,NPT=16

[21:33:24][0x000006c8][info] wrkr3-3 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16
[21:33:25][0x00001ec8][info] wrkr5-5 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16
[21:33:27][0x000006d8][info] wrkr7-7 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16
[21:33:27][0x0000177c][info] wrkr6-6 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16
[21:33:30][0x00002108][info] wrkr0-0 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16
[21:33:30][0x000011d8][info] wrkr1-1 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16
[21:33:30][0x00000ae8][info] wrkr2-2 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16
[21:33:31][0x000020b4][info] wrkr4-4 | Algorithm: CUDA-lbry parameters: B=6144,TPB=768,NPT=16

{'algorithms':
[{'algorithm_id': 0,
'name': 'lbry',
'connected': True,
'got_job': True,
'details': {'total_shares': 10,
'total_accepted': 10,
'total_rejected': 0,
'last_efficiency': 1.0},
'address': 'lbry.eu.nicehash.com:3356',
'login': 'x:x',
'workers': [{'worker_id': 0, 'device_id': 0, 'params': [], 'speed': 391009781.61415625},
{'worker_id': 1, 'device_id': 1, 'params': [], 'speed': 391009781.61415625},
{'worker_id': 2, 'device_id': 2, 'params': [], 'speed': 391009781.61415625},
{'worker_id': 3, 'device_id': 3, 'params': [], 'speed': 426823087.4625873},
{'worker_id': 4, 'device_id': 4, 'params': [], 'speed': 393728667.535854},
{'worker_id': 5, 'device_id': 5, 'params': [], 'speed': 418158225.5044178},
{'worker_id': 6, 'device_id': 6, 'params': [], 'speed': 415705440.8525755},
{'worker_id': 7, 'device_id': 7, 'params': [], 'speed': 415705440.8525755}]
}],
'id': 1,
'error': None
}

Remove workers
Add decred workers with no arguments

[21:33:43][0x000011d8][info] wrkr1-1 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16
[21:33:43][0x000006d8][info] wrkr7-7 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16
[21:33:43][0x0000177c][info] wrkr6-6 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16
[21:33:43][0x00001ec8][info] wrkr5-5 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16
[21:33:43][0x000020b4][info] wrkr4-4 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16
[21:33:43][0x00000ae8][info] wrkr2-2 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16
[21:33:43][0x00002108][info] wrkr0-0 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16
[21:33:43][0x000006c8][info] wrkr3-3 | Algorithm: CUDA-decred parameters: B=6144,TPB=768,NPT=16

{'algorithms':
[{'algorithm_id': 1,
'name': 'decred',
'connected': True,
'got_job': True,
'details': {'total_shares': 0,
'total_accepted': 0,
'total_rejected': 0,
'last_efficiency': 1.0},
'address': 'decred.eu.nicehash.com:3354',
'login': 'x:x',
'workers': [{'worker_id': 0, 'device_id': 0, 'params': [], 'speed': 363666194920.3757},
{'worker_id': 1, 'device_id': 1, 'params': [], 'speed': 360316232231.9314},
{'worker_id': 2, 'device_id': 2, 'params': [], 'speed': 396066294173.94855},
{'worker_id': 3, 'device_id': 3, 'params': [], 'speed': 351756358245.81464},
{'worker_id': 4, 'device_id': 4, 'params': [], 'speed': 368280100586.36176},
{'worker_id': 5, 'device_id': 5, 'params': [], 'speed': 400639096003.26666},
{'worker_id': 6, 'device_id': 6, 'params': [], 'speed': 373859946414.0466},
{'worker_id': 7, 'device_id': 7, 'params': [], 'speed': 342559374791.3434}]
}],
'id': 1,
'error': None
}

[nicehashdev]

Will take a look into this too.

[t4nja]

Both issues have been fixed. Thank you for the feedback!