SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070 - page 44.

yslyung

legendary

Activity: 1500

Merit: 1002

Mine Mine Mine

Quote from: hawkfish007 on November 13, 2016, 02:41:24 PM

Quote from: mrb on November 13, 2016, 02:28:27 PM

Quote from: hawkfish007 on November 13, 2016, 02:24:52 PM

1 out 7 480s (4 GB) is hashing lower than others. Is there something that could be done to fix it? Not that I am complaining, it is hashing way higher than Claymore V4, just curious what is causing 1 GPU (dev6) to hash lower.

Assuming you run with --instances 2 (default), one of the 2 sa-solver processes may have crashed for this GPU. Does it go back up to 80+ sol/s after restarting silentarmy?

Most of the time the GPU at this particular location start low across several rigs, the second GPU from the very last one. I am using the following bat

c:\cygwin64\bin\python3.4m silentarmy --instances=1 -c stratum+tcp://zec.suprnova.cc:2142 -u hawkfish007.2H -p x --use 0,1,2,3,4,5,6
pause

Should I change it to --instances=2 or 0?

Thanks.

i asked the same question earlier but no answers yet, i rolled back to earlier version & the particular slower gpu goes back to normal but overall mining speed is a tad slower compared to the newer version but total speed is abt the same although the particular gpu is slower because of increased hashrate from newer version.

some other rigs works well no issue but on some others with the same setup & gpu make & model has this particular half speed.

if you open up gpuz & montitor that gpu, it's not really working to full speed, looks like only 1 instance instead of 2. (default 1 is actually 2)

hagie

hero member

Activity: 792

Merit: 501

Quote from: jakobito on November 13, 2016, 02:01:58 PM

Quote from: hagie on November 13, 2016, 10:39:11 AM

Quote from: krnlx on November 13, 2016, 09:21:37 AM

Made a cuda port of actual Silentarmy 5

speeds sucks. 92 s/s on overclocked 1070(opencl version give near 100)

Tested only on linux.

If anyone want

https://github.com/krnlx/nheqminer

too bad it dumped :

Code:

~/aaa/nheqminer/Linux_cmake/nheqminer_cuda_sa$ ./nheqminer_cuda_sa -l eu1-zcash.flypool.org:3333 -u t1M74VKHZDJnHwmwWiHUiMWkBkzgQW4VGeg -cv 0 -cs -cd 0 -t 0

        ==================== www.nicehash.com ====================
                Equihash CPU&GPU Miner for NiceHash vSA5 cuda port by maztheman,krnlx
        Thanks to Zcash developers for providing base of the code.
               Special thanks to tromp, xenoncat, mbevand
                     and eXtremal-ik7 for providing
              optimized CPU, CUDA and AMD equihash solvers.
        ==================== www.nicehash.com ====================

Setting log level to 2
[15:36:43][0x00007f8b99409780] Using SSE2: YES
[15:36:43][0x00007f8b99409780] Using AVX: YES
[15:36:43][0x00007f8b99409780] Using AVX2: NO
[15:36:43][0x00007f8b96069700] stratum | Starting miner
[15:36:43][0x00007f8b96069700] stratum | Connecting to stratum server eu1-zcash.flypool.org:3333
[15:36:43][0x00007f8b95868700] miner#0 | Starting thread #0 (CUDA-SILENTARMY) GRID K520 (#0) BLOCKS=56, THREADS=64
[15:36:43][0x00007f8b96069700] stratum | Connected!
[15:36:43][0x00007f8b96069700] stratum | Subscribed to stratum server
[15:36:43][0x00007f8b96069700] miner | Extranonce is d65d009971
[15:36:43][0x00007f8b96069700] stratum | Authorized worker t1M74VKHZDJnHwmwWiHUiMWkBkzgQW4VGeg
[15:36:43][0x00007f8b96069700] stratum | Target set to 004189374bc6a7ef9db22d0e5604189374bc6a7ef9db22d0e5604189374bc6a7
[15:36:44][0x00007f8b96069700] stratum | Received new job #c1bc667ea1f4a9aea140
Segmentation fault (core dumped)

But it's not the fault of your port - it's the same with the plain nheqminer 0.4b and the cuda_tromp solver. maybe again bad luck with the old computing (30) version.

Regards

Try adding some flags in CMakeLists.txt for older compute capabilities 3.5 or 5.0. For 750Ti i solved it adding 5.0 flags:

set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};; -gencode arch=compute_50,code=sm_50;-gencode arch=compute_52,code=sm_52; -gencode arch=compute_61,code=sm_61 )

afaik the compute version is detected and used - definitions to use 3.0 are in the CMakeList ..

It still dumps - must be something different ..

regards

laik2

sr. member

Activity: 652

Merit: 266

Just put 2 more RX480 and started getting this:

Quote

[ 327.726578] VM fault (0x02, vmid 2) at page 1641426, read from 'TC4' (0x54433400) (72)
[ 327.726582] amdgpu 0000:03:00.0: GPU fault detected: 147 0x07c84802
[ 327.726597] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x001A82F1
[ 327.726613] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x04048002
[ 327.726629] VM fault (0x02, vmid 2) at page 1737457, read from 'TC4' (0x54433400) (72)
[ 327.726634] amdgpu 0000:03:00.0: GPU fault detected: 147 0x01484802
[ 327.726648] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x0017865A
[ 327.726664] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x04048002
[ 327.726680] VM fault (0x02, vmid 2) at page 1541722, read from 'TC4' (0x54433400) (72)
[ 327.726684] amdgpu 0000:03:00.0: GPU fault detected: 147 0x08304802
[ 327.727325] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x001BCEF3
[ 327.727865] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x04048002
[ 327.728406] VM fault (0x02, vmid 2) at page 1822451, read from 'TC4' (0x54433400) (72)
[ 327.728411] amdgpu 0000:03:00.0: GPU fault detected: 147 0x07d04802
[ 327.728943] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x001B280A
[ 327.729472] amdgpu 0000:03:00.0: VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x04044002
[ 327.729998] VM fault (0x02, vmid 2) at page 1779722, read from 'TC5' (0x54433500) (68)

I've installed some old powered PCIe risers which are non usb/nor original asrocks but cards are recognized.
Additional power to the motherboard H81 Pro BTC is not installed...may that be the problem?
It's not so often but a few times every 10mins... and mostly doesn't seem to stop mining unless it completely freezes.

hawkfish007

hero member

Activity: 895

Merit: 504

Quote from: mrb on November 13, 2016, 02:28:27 PM

Quote from: hawkfish007 on November 13, 2016, 02:24:52 PM

1 out 7 480s (4 GB) is hashing lower than others. Is there something that could be done to fix it? Not that I am complaining, it is hashing way higher than Claymore V4, just curious what is causing 1 GPU (dev6) to hash lower.

Assuming you run with --instances 2 (default), one of the 2 sa-solver processes may have crashed for this GPU. Does it go back up to 80+ sol/s after restarting silentarmy?

Most of the time the GPU at this particular location start low across several rigs, the second GPU from the very last one. I am using the following bat

c:\cygwin64\bin\python3.4m silentarmy --instances=1 -c stratum+tcp://zec.suprnova.cc:2142 -u hawkfish007.2H -p x --use 0,1,2,3,4,5,6
pause

Should I change it to --instances=2 or 0?

Thanks.

Ursul0

sr. member

Activity: 857

Merit: 262

Quote from: tc61 on November 13, 2016, 02:35:27 PM

Question,

the flag --instances , what does this actually mean? I know you can put 0 , 1 or 2 but what does it exactly do?

thxx

it runs instances of the opencl?

tc61

hero member

Activity: 494

Merit: 500

Question,

the flag --instances , what does this actually mean? I know you can put 0 , 1 or 2 but what does it exactly do?

thxx

mrb

legendary

Activity: 1512

Merit: 1028

Quote from: hawkfish007 on November 13, 2016, 02:24:52 PM

1 out 7 480s (4 GB) is hashing lower than others. Is there something that could be done to fix it? Not that I am complaining, it is hashing way higher than Claymore V4, just curious what is causing 1 GPU (dev6) to hash lower.

Assuming you run with --instances 2 (default), one of the 2 sa-solver processes may have crashed for this GPU. Does it go back up to 80+ sol/s after restarting silentarmy?

hawkfish007

hero member

Activity: 895

Merit: 504

1 out 7 480s (4 GB) is hashing lower than others. Is there something that could be done to fix it? Not that I am complaining, it is hashing way higher than Claymore V4, just curious what is causing 1 GPU (dev6) to hash lower.

Ale-x

newbie

Activity: 36

Merit: 0

Does anybody know how to run SAv5 after reboot in screen? OS: Ubuntu 16.04.
I've tried to put in rc.local smtng like this:

Code:

screen -S miner -d -m /home/user/silentarmy/silentarmy --use 0,1,2,3,4 -c stratum+tcp://equihash.eu.nicehash.com:3357#xnsub -u MY_WALLET.Rig1 -p x

but it works only if I run rc.local from command line but not after reboot.
Huh

UPD: its need to add path to screen, but now it starts from root. How to run it from specified user?

UPD2: SOLVED! Final string:

Code:

sudo -iu username /usr/bin/screen -S miner -d -m /home/user/silentarmy/silentarmy --use 0,1,2,3,4 -c stratum+tcp://equihash.eu.nicehash.com:3357#xnsub -u MY_WALLET.Rig1 -p x

Also I use amdgpu-pro-fans.sh script to set fan speed after boot.

Anybody use overclocking in Linux?

osnwt

sr. member

Activity: 353

Merit: 251

Quote from: Biodom on November 13, 2016, 01:48:50 PM

which low end CPU did you try? G1840, Sempron or Pentium G32**?
There is no point for me to try Optiminer if G32** (58 or 60) is not enough. Most of my rigs are on these.

My rigs use G3240, it is enough of it for all miners.
Low end was not mine, some single core Celeron 2800MHz, don't know more - was not interested, just helped to run the rig.

mrb

legendary

Activity: 1512

Merit: 1028

Quote from: cncr04s on November 13, 2016, 01:01:21 PM

The developer of this miner should correct the mining.authorize stratum command to include the expected number of parameters (params) as being an array of the worker name and password. It is currently sending only an array of one item, the workername and is breaking existing stratum code which has always included, and is defined as per protocol description as needing worker name and password.

example, {"id": 12, "method": "mining.authorize", "params": ["worker"]}

needs to be , {"id": 12, "method": "mining.authorize", "params": ["worker",""]}

This currently improper command breaks compatibility between this miner and some pools/services.

Will fix. But in the mean time you can just specify:

Code:

-p ""

on the command line and it will do the proper thing.

jakobito

newbie

Activity: 7

Merit: 0

Quote from: hagie on November 13, 2016, 10:39:11 AM

Quote from: krnlx on November 13, 2016, 09:21:37 AM

Made a cuda port of actual Silentarmy 5

speeds sucks. 92 s/s on overclocked 1070(opencl version give near 100)

Tested only on linux.

If anyone want

https://github.com/krnlx/nheqminer

too bad it dumped :

Code:

~/aaa/nheqminer/Linux_cmake/nheqminer_cuda_sa$ ./nheqminer_cuda_sa -l eu1-zcash.flypool.org:3333 -u t1M74VKHZDJnHwmwWiHUiMWkBkzgQW4VGeg -cv 0 -cs -cd 0 -t 0

        ==================== www.nicehash.com ====================
                Equihash CPU&GPU Miner for NiceHash vSA5 cuda port by maztheman,krnlx
        Thanks to Zcash developers for providing base of the code.
               Special thanks to tromp, xenoncat, mbevand
                     and eXtremal-ik7 for providing
              optimized CPU, CUDA and AMD equihash solvers.
        ==================== www.nicehash.com ====================

Setting log level to 2
[15:36:43][0x00007f8b99409780] Using SSE2: YES
[15:36:43][0x00007f8b99409780] Using AVX: YES
[15:36:43][0x00007f8b99409780] Using AVX2: NO
[15:36:43][0x00007f8b96069700] stratum | Starting miner
[15:36:43][0x00007f8b96069700] stratum | Connecting to stratum server eu1-zcash.flypool.org:3333
[15:36:43][0x00007f8b95868700] miner#0 | Starting thread #0 (CUDA-SILENTARMY) GRID K520 (#0) BLOCKS=56, THREADS=64
[15:36:43][0x00007f8b96069700] stratum | Connected!
[15:36:43][0x00007f8b96069700] stratum | Subscribed to stratum server
[15:36:43][0x00007f8b96069700] miner | Extranonce is d65d009971
[15:36:43][0x00007f8b96069700] stratum | Authorized worker t1M74VKHZDJnHwmwWiHUiMWkBkzgQW4VGeg
[15:36:43][0x00007f8b96069700] stratum | Target set to 004189374bc6a7ef9db22d0e5604189374bc6a7ef9db22d0e5604189374bc6a7
[15:36:44][0x00007f8b96069700] stratum | Received new job #c1bc667ea1f4a9aea140
Segmentation fault (core dumped)

But it's not the fault of your port - it's the same with the plain nheqminer 0.4b and the cuda_tromp solver. maybe again bad luck with the old computing (30) version.

Regards

Try adding some flags in CMakeLists.txt for older compute capabilities 3.5 or 5.0. For 750Ti i solved it adding 5.0 flags:

set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};; -gencode arch=compute_50,code=sm_50;-gencode arch=compute_52,code=sm_52; -gencode arch=compute_61,code=sm_61 )

ioglnx

sr. member

Activity: 574

Merit: 250

Fighting mob law and inquisition in this forum

Quote from: cncr04s on November 13, 2016, 01:01:21 PM

The developer of this miner should correct the mining.authorize stratum command to include the expected number of parameters (params) as being an array of the worker name and password. It is currently sending only an array of one item, the workername and is breaking existing stratum code which has always included, and is defined as per protocol description as needing worker name and password.

example, {"id": 12, "method": "mining.authorize", "params": ["worker"]}

needs to be , {"id": 12, "method": "mining.authorize", "params": ["worker",""]}

This currently improper command breaks compatibility between this miner and some pools/services.

Feel free to fix it.

Biodom

legendary

Activity: 3892

Merit: 4331

Quote from: osnwt on November 13, 2016, 01:33:26 PM

Quote

I was assuming the devfee on that miner was crazy high since he wouldn't tell you what it was. Some report Optiminer as unstable, and it is Linux only and harder to set up. Are you comparing solves over time at pool?

It is quite stable, there are issues with kernel error messages in logs, but it usually does not stop mining.
No yet watchdog if some card stops mining, but it is planned.
There is a Windows version (posted 2 hours ago, early alpha).

Quote

I am just observing no one has complained about the 15% devfee, yet everyone complained about 2.5% when there were a ton of slow, unstable, and/or pool locked miners at the time.

Miner is slow on low end CPUs. I tried it on one such machine and had around 80 sol/s for RX470 x 2, but 155 with SA5. At the same time with a better CPU the same 2 cards give with Optiminer ~240 sol/s.

which low end CPU did you try? G1840, Sempron or Pentium G32**?
There is no point for me to try Optiminer if G32** (58 or 60) is not enough. Most of my rigs are on these.
Thanks

osnwt

sr. member

Activity: 353

Merit: 251

Quote

I was assuming the devfee on that miner was crazy high since he wouldn't tell you what it was. Some report Optiminer as unstable, and it is Linux only and harder to set up. Are you comparing solves over time at pool?

It is quite stable, there are issues with kernel error messages in logs, but it usually does not stop mining.
No yet watchdog if some card stops mining, but it is planned.
There is a Windows version (posted 2 hours ago, early alpha).

Quote

I am just observing no one has complained about the 15% devfee, yet everyone complained about 2.5% when there were a ton of slow, unstable, and/or pool locked miners at the time.

I have different cards on my rigs (7950, 7970, 390/390X, 470). I had 1555 sol/s with SA5, but 2525 sol/s with Optiminer. DevFee is really 15% but it is above hashrate shown, giving total performance of 2900 sol/s - twice from SA5 and Claymore's 4.0.
No one complained since it is not even written in the README about DevFee (IIRC), not everyone sees that. Also unlike Claymore that mines for some time and it gives rate drops, this miner mines devfee in parallel with main jobs. So there is no any drops in rate visible.
Pool reported hashrate is equal to reported by the miner. DevFee is not included and is "outside" of that rate.

Miner is slow on low end CPUs. I tried it on one such machine and had around 80 sol/s for RX470 x 2, but 155 with SA5. At the same time with a better CPU the same 2 cards give with Optiminer ~240 sol/s.

cncr04s

member

Activity: 70

Merit: 19

Owner of MiningRigRentals

The developer of this miner should correct the mining.authorize stratum command to include the expected number of parameters (params) as being an array of the worker name and password. It is currently sending only an array of one item, the workername and is breaking existing stratum code which has always included, and is defined as per protocol description as needing worker name and password.

example, {"id": 12, "method": "mining.authorize", "params": ["worker"]}

needs to be , {"id": 12, "method": "mining.authorize", "params": ["worker",""]}

This currently improper command breaks compatibility between this miner and some pools/services.

ioglnx

sr. member

Activity: 574

Merit: 250

Fighting mob law and inquisition in this forum

You have to recompile it then.

changesl

member

Activity: 160

Merit: 10

Quote from: ioglnx on November 13, 2016, 12:15:34 PM

You can also go for:

Code:

#define COLL_DATA_SIZE_PER_TH		(NR_SLOTS * 9)

Increased the speed of GTX1800 by 10% and GTx1070 about 5% on Windows.
Test run was now 18h without issues or crashes or something.

i just edit param.h as below, correct? seems no change with default with zawawa r6

_{// Number of collision items to track, per thread
#ifdef __NV_CL_C_VERSION
#define COLL_DATA_SIZE_PER_TH (NR_SLOTS * 9)
#else
#define COLL_DATA_SIZE_PER_TH (NR_SLOTS * 9)
#endif}

toptek

legendary

Activity: 1274

Merit: 1000

Quote from: hawkfish007 on November 13, 2016, 12:16:04 PM

Quote from: jimmy_007vn on November 13, 2016, 12:10:42 PM

Quote from: hawkfish007 on November 13, 2016, 12:06:02 PM

Quote from: jimmy_007vn on November 13, 2016, 11:56:14 AM

I went to https://cygwin.com/index.html to download cyqwin, but i don't know where to select "Python3: Py3K language interpreter" .please help

Click on Phython then select Py3K language interpreter by selecting couple of check boxes (forgot what they are called). Install and enjoy!

Can you give me some detail about selecting Py3K. I downloaded Cyqwin 64bits windows version and phython 3.6 i believed.

Select Py3K from the list

tip if you type Python3 in the search box it puts the Py3K language interpreter at the top of the Python3 list instead looking thu the long list .

ioglnx

sr. member

Activity: 574

Merit: 250

Fighting mob law and inquisition in this forum

Quote from: TIKCrazy on November 13, 2016, 12:30:43 PM

Quote from: krnlx on November 13, 2016, 09:21:37 AM

Made a cuda port of actual Silentarmy 5

speeds sucks. 92 s/s on overclocked 1070(opencl version give near 100)

Tested only on linux.

If anyone want

https://github.com/krnlx/nheqminer

can somebody build this for windows? Thanks!

Well as mentioned previously he is using a Linux only lib that makes is close to impossible to make it for windows without deep changes.
Maybe if he tells us which parts of mm_malloc.h he used and if there is a equivalent for Windows.

Topic: SILENTARMY v5: Zcash miner, 115 sol/s on R9 Nano, 70 sol/s on GTX 1070 - page 44. (Read 209309 times)