CUDA miner

Genoil

That

set(nvcc_flags -m64;--std;c++11;

looks like a typo. I'll have a look.

rether

Has anyone compiled 1.1 with CUDA 6.5?

If so, it'd be much appreciated if you could share! Trying it out on a 960 I just got a hold of on Win 10.

Genoil

@rether I was secretly hoping they'd done something about it in CUDA 8RC. But i guess it didn't happen

rether

Genoil said:

@rether I was secretly hoping they'd done something about it in CUDA 8RC. But i guess it didn't happen

I actually haven't tried it with the newest nvidia drivers. I'm still on 347 because it was such a nightmare to stop Win10 from automatically updating them. Did CUDA 8 theoretically fix any of the reasons for poor performance on Win 10?

Genoil

rether said:

Genoil said:

@rether I was secretly hoping they'd done something about it in CUDA 8RC. But i guess it didn't happen

I actually haven't tried it with the newest nvidia drivers. I'm still on 347 because it was such a nightmare to stop Win10 from automatically updating them. Did CUDA 8 theoretically fix any of the reasons for poor performance on Win 10?

Well there are some rumours about fixing those TLB issues with the final version of CUDA 8, but they may only apply to Pascal GPUs.

barabeku

@Genoil can you help me creating install guide for mining on CentOS, i got K40 and M40 GPU here. But really need to run on CentOS 7.
Help me please. or can i just run linux binary. but i need pre-compiled ones.

JPS

Ok, I couldn't help myself and just had to order 2x EVGA 1080 FTW to replace my 980Ti's. On stock, so should be here soon.
I'm gonna need some guidance setting up ubuntu I guess, anyone ?

rether

I did end

Genoil said:

rether said:

Genoil said:

@rether I was secretly hoping they'd done something about it in CUDA 8RC. But i guess it didn't happen

I actually haven't tried it with the newest nvidia drivers. I'm still on 347 because it was such a nightmare to stop Win10 from automatically updating them. Did CUDA 8 theoretically fix any of the reasons for poor performance on Win 10?

Well there are some rumours about fixing those TLB issues with the final version of CUDA 8, but they may only apply to Pascal GPUs.

I did end up trying the latest drivers with 1.1 on Win10 with my 960. Unfortunately, still really bad. I'll still have to use a CUDA 6.5 version it looks like.

malefactor

Hey Genoil,

Would prefer to use opencl if speed is equivalent to CUDA (simpler). But I'm only seeing ~9 MH/sec on a 6x750ti system, where an ancient build was seeing 22-23MH/sec. Any idea what might be wrong?

Then, switching to CUDA, I can't build, using CUDA6.5. Is this expected?

[ 35%] Building NVCC (Device) object libethash-cuda/CMakeFiles/ethash-cuda.dir//./ethash-cuda_generated_ethash_cuda_miner_kernel.cu.o
nvcc fatal : Unsupported gpu architecture 'compute_61'
CMake Error at ethash-cuda_generated_ethash_cuda_miner_kernel.cu.o.cmake:206 (message):
Error generating
/root/tmp/genoil-cudaminer-1.1.3/cpp-ethereum/build/libethash-cuda/CMakeFiles/ethash-cuda.dir//./ethash-cuda_generated_ethash_cuda_miner_kernel.cu.o


libethash-cuda/CMakeFiles/ethash-cuda.dir/build.make:55: recipe for target 'libethash-cuda/CMakeFiles/ethash-cuda.dir/./ethash-cuda_generated_ethash_cuda_miner_kernel.cu.o' failed
make[2]: *** [libethash-cuda/CMakeFiles/ethash-cuda.dir/./ethash-cuda_generated_ethash_cuda_miner_kernel.cu.o] Error 1
CMakeFiles/Makefile2:292: recipe for target 'libethash-cuda/CMakeFiles/ethash-cuda.dir/all' failed
make[1]: *** [libethash-cuda/CMakeFiles/ethash-cuda.dir/all] Error 2

Genoil

run cmake with -DCOMPUTE=50

Problem is TLB issues with gtx750 due to current DAG size. Can't do much about it.

malefactor

Thanks, that gets me past one problem, but then I hit something that never got fixed from earlier in the thread. A CUDA 6.5 lib is complaining that the gcc is too new.

[ 38%] In file included from /usr/local/cuda/include/cuda_runtime.h:59:0,
                 from <command-line>:0:
/usr/local/cuda/include/host_config.h:82:2: error: #error -- unsupported GNU version! gcc 4.9 and up are not supported!
 #error -- unsupported GNU version! gcc 4.9 and up are not supported!

I'm aware of the TLB DAG issues on 750ti, but you said many times you expected OpenCL perf to be the same as CUDA. I'm seeing a lot less (9MH/sec vs 22 MH/sec). Is there something else I could do to make that work? I hate fiddling with CUDA versions.

Now, I'm trying CUDA 7 to see if it is happy & compiles.

malefactor

Retrying build with CUDA 7 fixes it, and your latest miner is functional, running about 22 MH/sec across 6 750ti cards. Would really appreciate it you could pull a rabbit out of your hat wrt TLB issue..

curzelcroz

hello guyz. can anyone help me pls. im having difficulty setting up my ethereum. I download genoils cpp-ethereum. can any one tell me the step-by-step guide becuz im not sure what im doing.

nethyb

Hi genoil,

On the new 110 branch I’m no longer able to submit shares using a local proxy server.

I’ve been using your miner (0.9.41-genoil-1.0.8) successfully to solo mine using 3 rigs in combination with EtherProxy (https://github.com/sammy007/ether-proxy) for quite a while.

Everything is working fine on your master branch but when I checkout 110 and try the new DAG-in-Memory version 0.9.41-genoil-1.1.3 the miner never submits a solution to the proxy.

As an aside - I’ve also tried with with dwarf’s proxy - same result - no solutions submitted.
Is there possibly an issue with solo mining or using proxy’s with the new version?
cheers, and just donated, thx.

Genoil

@nethyb thanks for the bug report, I'll investigate

Genoil

@nethyb and many thanks for the donation, much appreciated!

Mo35

nethyb said:

Hi genoil,

On the new 110 branch I’m no longer able to submit shares using a local proxy server.

I’ve been using your miner (0.9.41-genoil-1.0.8) successfully to solo mine using 3 rigs in combination with EtherProxy (https://github.com/sammy007/ether-proxy) for quite a while.

Everything is working fine on your master branch but when I checkout 110 and try the new DAG-in-Memory version 0.9.41-genoil-1.1.3 the miner never submits a solution to the proxy.

As an aside - I’ve also tried with with dwarf’s proxy - same result - no solutions submitted.
Is there possibly an issue with solo mining or using proxy’s with the new version?
cheers, and just donated, thx.

not tried 1.1.3 yet , but 1.1.2 copiled with -DCOMPUTE=52 works fine with dwarf proxy , running few rigs for couple days without hickups

nethyb

Genoil said:

@nethyb thanks for the bug report, I'll investigate

I should also have said i'm using the miner with OpenCL for my AMD cards

Etherion

Where can I download 1.1.3 for Windows 7 64Bit?

Mo35

https://github.com/Genoil/cpp-ethereum/tree/110

check under releases or compile yourself

Etherion

@Mo35 Thanks. Got it and it is running now.

@Genoil You say alot that Compute 2.0 is not as good as the Open_CL 1.2 for older cards. But I get 7.1MH/s from my old Q5000 with your CUDA miner compared to 5.2 MH/s using -G. I get the same 7.1MH/s with QTMiner but more than 10% failed shares. So yours is faster i.m.o.

I see you have removed -C CPU. I have a dual Intel® Xeon® Processor E5-2699 v4 Workstation that can do about 5.5MH/s on the CPU's. So far it covers the power and a bit more. But as 1.1.3 does not support CPU any more i need am running 1.0.8 and 1.1.3.

chrizzo

Is 1.1.3 final or in a RC state? Reading about much problems, so is stay on 1.0.8 this time.

Marvell9

This Nvidia mining really anoys me thiough part of why im getting rid of my 24 970s is sometimes a rig that was working fine after a reeboot (windows 8.1) will suddenly keep submitting invalid shares.

It can go on for days like this wasting power if i dont notice and the only resolution is to fully reinstall the OS

tried everything from deleting the dags , rolling back to older versions of the miner etc. Seems to happen randomly today it happened again. A six card rig didn't submit any shares all day

if you look at the log you see

"Not accepted "

I would rather the miner stops working than constantly doing this and wasting power and giving the illusion its still mining.

I used to have six six card 970 rigs but this annoyance is causing me to get rid of them all and just focus on AMD.

Genoil

Etherion said:

@Mo35 Thanks. Got it and it is running now.

@Genoil You say alot that Compute 2.0 is not as good as the Open_CL 1.2 for older cards. But I get 7.1MH/s from my old Q5000 with your CUDA miner compared to 5.2 MH/s using -G. I get the same 7.1MH/s with QTMiner but more than 10% failed shares. So yours is faster i.m.o.

I see you have removed -C CPU. I have a dual Intel® Xeon® Processor E5-2699 v4 Workstation that can do about 5.5MH/s on the CPU's. So far it covers the power and a bit more. But as 1.1.3 does not support CPU any more i need am running 1.0.8 and 1.1.3.

Yes you're right, my new Compute 2 code is faster. I don't have CPU mini G in 1.1amy longer because the DAG is no longer in system RAM.

As for what is RC and what not, once I move to 1.2 development, 1.1 becomes the main stable release, etc.

ethfan

@Genoil, can you use static linking so that it is possible to just do it once and then deploy the binary to multiple rigs instead of repeating the compile/link cycle on every rig... For Linux

dlehenky

@Genoil Any verdict on the sync stratum implementation?

Logicaluser

@Genoil Just ran across what looks to be the same Stratum issue I reported a few weeks back with 1.0.8 for the first time since switching some rigs to 1.1.3 the other day

Past reports here: https://forum.ethereum.org/discussion/2227/cuda-miner/p53

Got the offline notification from pool, connected to the machine to find the endless loop of IO operation aborted/could not resolve host... only to have it resolve itself and go right back to mining the second I teamviewer in.

Didn't catch the initial cause, but here's what happens before/after I teamviewer into the machine.
(duplicate lines trimmed down a bit)

Oh, and this is with --work-timeout set to 300 seconds, and on a wired network connection with a few dozen mining rigs, yet the issue seems to impact just one or two rigs at a time (and thus far only AMD AM3 platform rigs)

Win7 x64, 5 GPUs, AMD 15.7.1.

@ i 09:10:28| Connecting to stratum server us2.ethpool.org:3333
Could not resolve hostus2.ethpool.org:3333, The I/O operation has been aborted because of either a thread exit or an app
lication request i 09:10:28| Reconnecting in 3 seconds...
i 09:10:29|openclminer4 Solution found; Submitting to us2.ethpool.org ...
i 09:10:29|openclminer4 Nonce: 0x2ae642a139bc818a
X 09:10:29|stratum Handle response failed: An established connection was aborted by the software in your host machin
e
i 09:10:31| Connecting to stratum server us2.ethpool.org:3333
Could not resolve hostus2.ethpool.org:3333, The I/O operation has been aborted because of either a thread exit or an app
lication request i 09:10:31|stratum Reconnecting in 3 seconds...
i 09:10:34|stratum Connecting to stratum server us1.ethpool.org:3333
Could not resolve hostus1.ethpool.org:3333, The I/O operation has been aborted because of either a thread exit or an app
lication request i 09:10:34| Reconnecting in 3 seconds...
i 09:11:04| Connecting to stratum server us1.ethpool.org:3333
Could not resolve hostus1.ethpool.org:3333, The I/O operation has been aborted because of either a thread exit or an app
lication request i 09:11:04|stratum Reconnecting in 3 seconds...
i 09:11:07|stratum Connecting to stratum server us1.ethpool.org:3333
i 09:11:07|stratum Connected to stratum server us1.ethpool.org : 3333
i 09:11:07|stratum B-) Submitted and accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:07|stratum :-( Not accepted.
X 09:11:08|stratum :-( Not accepted.
X 09:11:08|stratum :-( Not accepted.
X 09:11:08|stratum :-( Not accepted.

SNIP, removed ~40 lines of X 09:11:08|stratum :-( Not accepted.


X 09:11:08|stratum :-( Not accepted.
X 09:11:08|stratum :-( Not accepted.
i 09:11:08|stratum Subscribed to stratum server
m 09:11:08|main Mining on PoWhash #b1f826ad : 79.22MH/s [A2697+37:R235+0:F0]
i 09:11:08|stratum Authorized worker 0xb90a82ec61627885eab72f4253939285ba40c91d.MinerC
i 09:11:08|stratum Received new job #45937a1d
i 09:11:09|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:11:09|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:11:09|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:11:09|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:11:09|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
m 09:11:10|main Mining on PoWhash #45937a1d : 89.15MH/s [A2697+37:R283+0:F0]
i 09:11:12|stratum Received new job #e7cd02d7
i 09:11:27|openclminer3 Solution found; Submitting to us1.ethpool.org ...
i 09:12:54|openclminer3 Nonce: 0xc8aa53dd184544ce
i 09:12:54|stratum Received new job #b4a15502
i 09:11:12|openclminer4 Solution found; Submitting to us1.ethpool.org ...
i 09:12:54|openclminer4 Nonce: 0x3f97ecf0711994e6
i 09:12:54|stratum Received new job #1b9a9112
i 09:12:54|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|stratum Received new job #f919a3f5
i 09:12:54|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|stratum Received new job #6e2199c8
i 09:12:54|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:54|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|stratum Received new job #4728abf6
i 09:12:55|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:55|stratum Received new job #f6815d2d
i 09:12:55|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
m 09:12:56|main Mining on PoWhash #f6815d2d : 51.91MH/s [A2697+37:R283+0:F0]
i 09:12:56|stratum Received new job #ba710477
i 09:12:56|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|stratum Received new job #77d140be
i 09:12:56|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:56|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|stratum Received new job #0bc5c452
i 09:12:57|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82
X 09:12:57|stratum :-( Not accepted.
X 09:12:57|stratum Read response failed: End of file
i 09:12:57|stratum Reconnecting in 3 seconds...
i 09:12:57|openclminer0 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer1 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer2 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:57|openclminer3 set work; seed: #c3ddddc8, target: #0000000112e0be82
i 09:12:58|openclminer4 set work; seed: #c3ddddc8, target: #0000000112e0be82

Genoil

@Logicaluser try -SV 2 flag. It's a new stratum client. Curious whether or not this would solve the issue.

kps

I'm just starting with eth mining, so I have installed a genoil cuda miner, using the standard -G -M benchmark, i seem to be getting about 42 mhash/s, however when i use actual pool mining with -G -F / --opencl-platform 1 arguments, the hash rate seems to be between 7-10mhash/s, unless i'm reading it wrong, example

m 22:07:26|main Mining on PoWhash #04f644c0 : 7.31MH/s [A0+0:R0+0:F0]
m 22:07:27|main Mining on PoWhash #04f644c0 : 9.91MH/s [A0+0:R0+0:F0]
m 22:07:28|main Mining on PoWhash #04f644c0 : 10.83MH/s [A0+0:R0+0:F0]
m 22:07:28|main Mining on PoWhash #04f644c0 : 9.13MH/s [A0+0:R0+0:F0]

I use win 10 with the newest 368.39 drivers, and i have 2 GTX 1080 gpus. Am i doing something wrong or is this about the limit with win 10 ? I can install linux for ethereum mining purposes if that improves the hash rate greatly, also i didn't compile genoil miner, i have just used the pre-made binary

work

@kps benchmark uses epoch 0 dag, which will be much faster then the current dag. It's been discussed a lot on these forums already, but nvidia just sucks for ethash mining because of TLB smashing issues. There is hope that a later driver release may improve things.