Topic: Genoil's ethminer - CUDA driver error (Read 1951 times)

I've seen that error and also CUDA error in func search (illegal memory access). Dialing back the overclock a little bit worked for me.

what about reducing the overclock eh ?

Same problem today. I am suspecting one card is causing the problem but not sure yet.  Hopefully someone works this out.


--cuda-block-size 128 --cuda-grid-size 8192 less than
--cuda-block-size 64 --cuda-grid-size 16384
64 and 16384 best so far defailt --cuda-parallel-hash <1 2 ..8> *[4]

  i  08:00:45|main  Solution found; Submitting to 127.0.0.1:8080 ...
  i  08:00:48|main    Nonce: 26f776b0fe1d452d
  i  08:00:48|main  B-) Submitted and accepted.
CUDA error in func 'ethash_cuda_miner::search' at line 346 : unknown error.
  X  08:01:06|cudaminer0  Error CUDA mining: unknown errorCUDA error in func 'ethash_cuda_miner::search' at line 346 : unknown error.
CUDA error in func 'ethash_cuda_miner::search' at line 346 : unknown error.
CUDA error in func 'ethash_cuda_miner::search' at line 346 : unknown error.
CUDA error in func 'ethash_cuda_miner::search' at line 346 : unknown error.

yes, agree with that. 7 hours of lost mining time is terrible. Hope the developer add a restar feature soon.

I was lucky to get Samsung memory in all 6 cards although my 6th card is a pain and gives errors if OC more than 450 mem clock speed. All other cards are at 850 mem clock, 90 power.

If my rig crashes while im sleeping thats potentially 7 hours mining lost so would be a negative trade vs claymore and its shutdown/restart option.

One crash in three days, thats acceptable to my standards... but mine crashes in 20-30 minutes.... not good

my 1060's memory are micron, what are yours?
I haven't tried running stock, just to check. But if running stock gets me the same hashrate as claymore.. ill just use claymore with dual mining.

I've been asking around for a crash/hang detect and auto restart solution but there doesnt appear to eb anything.

I was using Claymore and getting 136MH off 6 x Gigabyte 1060s and with EthMiner Im getting 151 MH!

So definately worth sticking with Ethminer so long as it doesnt crash.  In 3 days i've had 1 crash that was alerted to me by ethermine.org emailing me to say my miner was offline and it was errors in the CMD window with machine not hanging so I just remoted in and restarted but I still lost an hours mining

yep, confirmed. same happens with my MSI 1060 6gb (micron memory) at 600mhz memory 0mhz core.
It runs for maybe 20 minutes and then crashes. Doesn't even restart, which is a shame.

I will be using Claymore till when I find a fix as the optimized 1060 ethminer is definitely faster.

bump

I'm getting this error too, did you ever find a fix?

edit: and I'm on win10 using the latest 1060 optimized code.

Hi,

I am running genoil's cudaminer on Ubuntu 16.04 and after a random amount of hours, CUDA driver crashes and miner hangs:

CUDA error in func 'search' at line 346 : unknown error.
  ✘  05:07:10|cudaminer3  Error CUDA mining: unknown error
CUDA error in func 'search' at line 346 : unknown error.
  ✘  05:07:10|cudaminer4  Error CUDA mining: unknown error
CUDA error in func 'search' at line 346 : unknown error.
  ✘  05:07:10|cudaminer6  Error CUDA mining: unknown error
CUDA error in func 'search' at line 346 : unknown error.
  ✘  05:07:10|cudaminer2  Error CUDA mining: unknown error
CUDA error in func 'search' at line 346 : unknown error.
  ✘  05:07:10|cudaminer5  Error CUDA mining: unknown error
CUDA error in func 'search' at line 346 : unknown error.
  ✘  05:07:10|cudaminer1  Error CUDA mining: unknown error
CUDA error in func 'search' at line 346 : unknown error.
  ✘  05:07:10|cudaminer0  Error CUDA mining: unknown error
  m  05:07:12|ethminer  Mining on PoWhash #21d3071c : 4.72MH/s [A1496+9:R0+0:F0]
  m  05:07:14|ethminer  Mining on PoWhash #21d3071c : 0.00MH/s [A1496+9:R0+0:F0]
  m  05:07:16|ethminer  Mining on PoWhash #21d3071c : 0.00MH/s [A1496+9:R0+0:F0]
  m  05:07:18|ethminer  Mining on PoWhash #21d3071c : 0.00MH/s [A1496+9:R0+0:F0]

Did anyone mod this miner to auto restart after a driver crash, or any other workarounds?

Thanks