This repository has been archived by the owner on Jan 23, 2019. It is now read-only.
[Bug][NVIDIA][970] Slow hash rate after about 2hrs #22
Comments
|
@Blackskyliner Thanks for the detailed bug report. I think the GPU is completely off at some point, 40 Sol/s that you see is coming from your AVX2 CPU's as far as I can tell. Can you run only with |
|
About the remaining 40 Sol/s I also thought it may be my CPU only. But after some testing it only gets up to about 25 Sol/s. I will watch on my next mining run, going to bed now, if it happens again, on GPU only. Maybe I will also be able to get a good Thanks for the fast reaction 😃. Only CPU ( Only GPU ( |
|
So it happened again, but I guess the last time I was just lucky, a benchmark did not trigger the reset of that phenomen, I guess it was a lucky one. But it is indeed all-time-slow so my guess would stay that I somehow hit the slow-memory I guess. But why does it not reset though... As at the minute the program gets disconnected the CUDA stack should cleanup after itself or at least free stuff... EDIT: Added CUDA information. Running with debug enabled on slow job: Benchmark ( CUDA Information ( |
|
Closed all Applications (e.G. Mail, Firefox) aaaand it's working again. What I don't get I can even do graphics work on this machine while mining, okay the speed will be only at 50% most of the time, but it works. So why does it seem that some application combination can lock the GPU only for those mining relevant tasks... But I got aclue where to debug next time when this is happening. Maybe I can identify the application or combination which is causing that problem... Also compiled https://github.com/phvu/cuda-smi so next time I can get a view of the Memory usage (or maybe that program resetted the CUDA stack too much unknown variables atm. have to reduce those next time and try/check after each interaction with the system) Funfact about 42% of the GPU memory is always reserved. Its even reserved now, even after a decent cold reboot -- seems to be normal macOS behavior on my end... ./cuda-smi |
|
If you find a certain combination I'll try with my setup too (MP 4.1 with GTX960). Memory stats is more or less the same with yours |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Operating System: macOS 10.12.6 Sierra
Used GPU: GeForce GTX 970
Used Driver: NVIDIA Web Driver: 378.05.05.25f04
Version of Miner: 280fc55
Command used:
./nheqminer-gpu -cd 0 -t 3 -u MyZCashAddress.MinerName -p x -l eu1-zcash.flypool.org:3333Observed Behavior
Problem Guessing
-d 0before the bench reset, so we have no advanced debugging in the following logs-d 0session just to look if it's SW or HW related bug, as if it is the later one we should limit as stated above.Log of miner while incident and before bench-reset:
Log after restarting before bench-reset:
Log of bench-reset:
Log of starting miner after bench-beset:
The text was updated successfully, but these errors were encountered: