Doesn't do anything on GTX 670

[markasoftware]

I have a 2gb ASUS GTX 670. When running just ./silentarmy to the donation address on Arch Linux, it gives this output, then nothing:

Connecting to us1-zcash.flypool.org:3333
Solver 0.0: unexpected banner "/home/markasoftware/Downloads/silentarmy/sa-solver: /opt/cuda/lib64/libOpenCL.so.1: no version information available (required by /home/markasoftware/Downloads/silentarmy/sa-solver)"
Solver 0.1: unexpected banner "/home/markasoftware/Downloads/silentarmy/sa-solver: /opt/cuda/lib64/libOpenCL.so.1: no version information available (required by /home/markasoftware/Downloads/silentarmy/sa-solver)"
Stratum server sent us the first job
Mining on 1 device

No sol/s or any other progress indication. Are kepler GPUs supported? Only 42% of my GPU memory is in use according to nvidia control panel.

EDIT: The GPU does show up in ./silentarmy --list

[mbevand]

I don't know what is the cause of the pb. Sorry SA is at the moment not supported.

[markasoftware]

i think I found the source of the issue anyways, because it occurred with other miners (genoil) too: The DAG is now over 2gb so does not fit in the vram of my 2gb GTX 670.

[slashafk]

Hello, same problem with my GTX780, I start Silentarmy, and then nothing after :

Stratum server sent us the first job
Mining on 1 device

I did try to rerun it with -vvv to see more information, I got those line :
To solvers: 0000000000000000000000000000000000000000000000000000000000004000 123853#651 040000005735736eedeeaa7676280ed0fb0e26485d97f7de4465471211d9c3460000000064d7b546eeca3f9ea53e13fceb03a313d9f9e641b85250f67570f2af7f67c444000000000000000000000000000000000000000000000000000000000000000030152f59e7d5471c 00000000

And nothing after after, like the process did hang or something :s

If I then use :

sa-solver --mining

And paste the mining stuff :
0000000000000000000000000000000000000000000000000000000000004000 123853#651 040000005735736eedeeaa7676280ed0fb0e26485d97f7de4465471211d9c3460000000064d7b546eeca3f9ea53e13fceb03a313d9f9e641b85250f67570f2af7f67c444000000000000000000000000000000000000000000000000000000000000000030152f59e7d5471c 00000000

It does the mining and find solutions ...

SILENTARMY mining mode ready 0000000000000000000000000000000000000000000000000000000000004000 123852#808 04000000ed99d8f67b531acf11a0adcfb60a9f50864979a8b03bd220d5c4361b00000000a408c98e77be4fffc8f9373f5894f811f0ea3d397f711a02ed65adaa07115b9d0000000000000000000000000000000000000000000000000000000000000000c6142f59ed92471c 00000000 status: 1 0 status: 16 0 status: 23 0 status: 41 0 status: 51 0 status: 60 0

So I think that the problem is not the solver itself, but the python daemon somehow lose the process that he is suppose to watch.

Does it works with python3.6/linux ? asyncio problem ? Encoding problem ? (Not using utf8 by default)

Seems to work pretty well until :
for devid in self.solver_procs:
self.solver_procs[devid].stdin.write(job.encode('ascii'))

[slashafk]

Well I did investigate a bit more, it appears that the opencl version embedded by Cuda doesn't return a "required information"
sa-solver: /opt/cuda/lib64/libOpenCL.so.1: no version information available (required by sa-solver)

That fail the following test :

        #if banner != "SILENTARMY mining mode ready":
            #print('Solver %s: unexpected banner "%s"' % (devid, banner))
            #proc.kill()
            #self.cleanup_solvers(devid)
            #return

So the sa-solver gets killed.

I did comment it out, and it works perfectly :).

I guess :
stderr=asyncio.subprocess.STDOUT
return the error at the place of the banner, resulting in a crash :s.

You should may be manage stderr in another subprocess or ignore this one.

[C0rn3j]

Same issue on Arch with GTX 660. Using GTX 970 there is no issue.

[Smodel]

Hi all i am having the same issue with 1080 Ti on Ubuntu. Any solution ?

[C0rn3j]

Unrelated to this project, but I ended up using ccminer-git from AUR. While I still can't use it with my 660, it's way faster than silentarmy on my GPU.