[ANN] dstm's ZCash / Equihash Nvidia Miner v0.6.2 (Linux / Windows)

[dstm]

@dstm

Is there any space for change requests? How about adding %subvendor% of the card in output, maybe turned on by parameter.

Thanks.

It's on my todo-list, it was already requested. It seems nvidias libraries don't provide this information on linux, I'll look into to it later again.

[dstm]

Could anyone please tell me what cuda is the best for this miner?

Any driver supporting CUDA 8 and above is fine.

[dstm]

I have a problem on one of my rigs since I started using 0.5.8. The miner crashes every like 2 days. So, what info can I provide to you so you can check it out? It was stable on 0.5.7 since it was released.

This is most likely due to overclocking, try to reduce it slightly. What messages does zm report after the crash?

[dstm]

<snip>
Version 0.5.7
<snip>

Hi,

Thanks for this, I find this version much more stable compared to 0.5.4 which I'd last tried on my Windows 10, Nvidia rig (2 x 1070, 1 x 1080) and also 6-7% better performance than EWBF.

I wanted to ask how to get email notifications whenever there are updates? The forum only offers notifications for the entire post/all new replies and I don't want that.

Let me know if there's a way to do that. Thanks! 

@DSTM - I'm still using ver 0.5.7 and gratefully paying the 2% devfee, but I have a question/suggestion - Why does the console output keep changing the order of the GPU stats it displays. It starts out as GPU0 onwards in order, but after an hour or two or three, it appears to randomize the order. See screenshot #1 linked below. This makes it quite confusing (and annoying) while viewing the console output to gauge real-time changes.



Also, the apparently random order itself keeps changing every so often, making it even more confusing to view. See screenshot #2 linked below.



Ensuring that the GPU rank/order is maintained would make it much easier to gauge real-time performance changes with a quick glance at the console, esp while tweaking for better performance.

(This is on Flypool. GPU0 & GPU1 are GTX 1080's, GPU2 and GPU3 and GTX 1070's. Driver ver 388.31. Using Windows 10 64-bit. Intel i5-4690k - this is my main PC, so..)

p.s. Also - re: the versioning updates question - can't you put this up on Github or somewhere similar so we can get push notifs for updated versions? This is a very minor issue though, the above issue is higher priority, at least for me. The randomized ordered display of GPUs is quite annoying

This is the order in which things actually happen on your system. It's on my todo list now.
Concerning notifications/distribution, right something like Github seems to be a much better option - I'll switch to Github or something similar.

[dstm]

This is my first post on this forum, despite being interested in crypto coins practically since beginning of BTC, so Hi everyone.

I'm thinking about building rig that would be most efficient. Everyone is comparing Sol/s or Sol/W, but as dstm stated (went practically through whole thread yesterday ), performance should be measured by I/s.
Do you know if anyone performed comparison based on I/s per W and how it relates to Sol/s and Sol/W in long term (as those are probablistic) for specific card models, brands? Could I ask you to submit your data here in following format:

Code:

Brand, Model, MB, I/s, W, Sol/W, Sol/s, %TDP (optional), +/-Core (optional), +/- Memory (optional), comments (optional)
Asus, GeForce GTX 750 Ti, 2048, 39.3, 27.19, 2.68, 72.76,,,,,

When adding yours, please include result posted by previous users and add yours at the end. Normally I would add link to some anonymous notepad like shrib under #nv_dstm_zcash_perf, so others can process and compare it as they wish, but unfortunately i can't post any links

24h runs averages would be best. If brand model name has been previously added by someone, use the same model / brand name so in longer term we can easy compare between products and models by filtering only by them. If possible provide results without OC/Undervolting.

Thanks and keep up good work!

Right, Sol/s are probabilistic - I/s is a better measure for performance especially if you're doing overclocking experiments - however zm includes an average for Sol/s - it will converge after some time - Sol/W is also an average so it will also converge after some time i.e. get pretty accurate. Btw. Sol/W is actually wrong - the units are wrong - the right units are Sol/J (Joule) - zm uses Sol/W since that's easier to understand for most people.

[dstm]

Does the recent Meltdown and Spectre OS patches affect the DSTM Miner's performance?
Has anybody done any before/after benchmarks?

From what, I read the recent Meltdown patch (KPTI) affects any programs that call the Kernel frequently such as netwoking, SSD disk access and CUDA memory copy operations from host to GPU.

How often does the DSTM Miner call the Kernel, anyway ?

I'm looking into this on both Operating-Systems. Not all OS-patches are finished currently... so yes, I'm aware of the of the increased systemcall time.

[dstm]

Hi, dstm! Big thanks for your miner.
I gets a lot of RECV, farm reconnects and keep working, but I think, that shares a lost.
2018-01-08 7:44:12|   ========== Sol/s: 2326.0 Sol/W: 3.46  Avg: 2293.5 I/s: 1249.0 Sh: 7.86   0.99 178
2018-01-08 7:44:15|#  recv failed: 10054
2018-01-08 7:44:15|#  reconnecting
2018-01-08 7:44:26|#  connected to: eu1-zcash.flypool.org:3333

2018-01-08 0:58:46|nvmlDeviceGetTemperature failed (this is my farm goes to sleep! Fixed)
2018-01-08 7:31:52|>  GPU0  32C  Sol/s: 0.2    Sol/W: 3.45  Avg: 473.8  I/s: 0.1    Sh: 1.82   0.99 78 
2018-01-08 7:31:52|   GPU3  33C  Sol/s: 0.1    Sol/W: 3.29  Avg: 452.8  I/s: 0.0    Sh: 1.47   1.00 78 
2018-01-08 7:31:52|   GPU2  33C  Sol/s: 0.1    Sol/W: 3.39  Avg: 465.2  I/s: 0.1    Sh: 1.40   1.00 109
2018-01-08 7:31:52|>  GPU4  32C  Sol/s: 0.3    Sol/W: 3.87  Avg: 463.6  I/s: 0.2    Sh: 1.97   1.00 553
2018-01-08 7:31:52|   ========== Sol/s: 465.6  Sol/W: 3.48  Avg: 2324.2 I/s: 251.2  Sh: 8.32   1.00 179
2018-01-08 7:31:53|   GPU1  33C  Sol/s: 0.1    Sol/W: 3.39  Avg: 464.9  I/s: 0.1    Sh: 1.64   1.00 78 
2018-01-08 7:32:06|   GPU1  40C  Sol/s: 310.9  Sol/W: 3.38  Avg: 463.6  I/s: 169.5  Sh: 1.63   1.00 78 
2018-01-08 7:32:06|   GPU0  38C  Sol/s: 345.2  Sol/W: 3.45  Avg: 472.8  I/s: 183.6  Sh: 1.80   0.99 78 
2018-01-08 7:32:07|   GPU3  40C  Sol/s: 327.9  Sol/W: 3.29  Avg: 451.8  I/s: 176.5  Sh: 1.46   1.00 78 
2018-01-08 7:32:07|#  recv failed: 10054
2018-01-08 7:32:07|#  reconnecting
2018-01-08 7:32:18|#  connected to: eu1-zcash.flypool.org:3333
2018-01-08 7:32:21|#  server set difficulty to: 0004189374bc6a7ef9db22d0...
I changes overclocking for minimum (tdp 50%), but RECV still my guest.
Sorry for my bad english.

This is not due to overclocking, it's a network issue, are you on wifi?

[b2s]

dtms check last bminer changelog

"* Minor performance improvement for GTX 1060"

Maybe he is using your miner inside somehow?

[somethingiswrong]

Anyone got problems since the latest version of dstm's zm miner was updated on SMOS?
SMOS restarts the miner at least once a hour.
Maybe this is a SMOS problem?
I'm using Nvidia GPUs on a ASrock H110 BTC+ PRO...

EDIT: nevemind it seems to be pool timeout

[MinerGTX]

Does the recent Meltdown and Spectre OS patches affect the DSTM Miner's performance?
Has anybody done any before/after benchmarks?

From what, I read the recent Meltdown patch (KPTI) affects any programs that call the Kernel frequently such as netwoking, SSD disk access and CUDA memory copy operations from host to GPU.

How often does the DSTM Miner call the Kernel, anyway ?

I'm looking into this on both Operating-Systems. Not all OS-patches are finished currently... so yes, I'm aware of the of the increased systemcall time.

Does the dstm miner perform the Blake2b hash on the CPU and transfer the results to the GPU by calling the CUDA driver and the Kernel each time ?
If "yes" then it can be affected by the increased systemcall time.

[SpecterHome]

Hi, dstm! Big thanks for your miner.
I gets a lot of RECV, farm reconnects and keep working, but I think, that shares a lost.
2018-01-08 7:44:12|   ========== Sol/s: 2326.0 Sol/W: 3.46  Avg: 2293.5 I/s: 1249.0 Sh: 7.86   0.99 178
2018-01-08 7:44:15|#  recv failed: 10054
2018-01-08 7:44:15|#  reconnecting
2018-01-08 7:44:26|#  connected to: eu1-zcash.flypool.org:3333

2018-01-08 0:58:46|nvmlDeviceGetTemperature failed (this is my farm goes to sleep! Fixed)
2018-01-08 7:31:52|>  GPU0  32C  Sol/s: 0.2    Sol/W: 3.45  Avg: 473.8  I/s: 0.1    Sh: 1.82   0.99 78 
2018-01-08 7:31:52|   GPU3  33C  Sol/s: 0.1    Sol/W: 3.29  Avg: 452.8  I/s: 0.0    Sh: 1.47   1.00 78 
2018-01-08 7:31:52|   GPU2  33C  Sol/s: 0.1    Sol/W: 3.39  Avg: 465.2  I/s: 0.1    Sh: 1.40   1.00 109
2018-01-08 7:31:52|>  GPU4  32C  Sol/s: 0.3    Sol/W: 3.87  Avg: 463.6  I/s: 0.2    Sh: 1.97   1.00 553
2018-01-08 7:31:52|   ========== Sol/s: 465.6  Sol/W: 3.48  Avg: 2324.2 I/s: 251.2  Sh: 8.32   1.00 179
2018-01-08 7:31:53|   GPU1  33C  Sol/s: 0.1    Sol/W: 3.39  Avg: 464.9  I/s: 0.1    Sh: 1.64   1.00 78 
2018-01-08 7:32:06|   GPU1  40C  Sol/s: 310.9  Sol/W: 3.38  Avg: 463.6  I/s: 169.5  Sh: 1.63   1.00 78 
2018-01-08 7:32:06|   GPU0  38C  Sol/s: 345.2  Sol/W: 3.45  Avg: 472.8  I/s: 183.6  Sh: 1.80   0.99 78 
2018-01-08 7:32:07|   GPU3  40C  Sol/s: 327.9  Sol/W: 3.29  Avg: 451.8  I/s: 176.5  Sh: 1.46   1.00 78 
2018-01-08 7:32:07|#  recv failed: 10054
2018-01-08 7:32:07|#  reconnecting
2018-01-08 7:32:18|#  connected to: eu1-zcash.flypool.org:3333
2018-01-08 7:32:21|#  server set difficulty to: 0004189374bc6a7ef9db22d0...
I changes overclocking for minimum (tdp 50%), but RECV still my guest.
Sorry for my bad english.

This is not due to overclocking, it's a network issue, are you on wifi?

No, that farm is on lan, another on wifi and has a much less recv (2-3 days uptime). However I wrote to flypool and they answered
"Thank you for your question, generally we see this error from clients with low internet connections or now wired connected rigs."
But WiFi farm has much low recv.
To exclude all factors, I reistall MSDN Win7_64 Pro Licensed with integrated updates, install 388.73 driver, vc++2015 and now has a very smooth flight for the 2 days with the only one recv (but in the same time with my friend, so that is because of flypool).
Big thanks to rotating me from overclocking, to the internet, its much easier to fix.
 

[WaveRiderx]

Could some of you guys check your devfee stats

Tonight I ran 1 of my rigs about 4 hours it produced this:

58 hits of * and 568 hits of +

That is much more than 2%.

Hopefully there is a genuine explanation in there.

I'm about the last person to defend the devfees charged by some of these miners, because many of them are still beta or even alpha-level in functionality/reliability, but I suspect the explanation here is that the devfee shares are mined at a much lower difficulty to better ensure a share is earned during the short mining period allotted.

But this is just a theory on my part.

I think your theory is right.

Beside, zm is mining devfee on zcash flypool (see here: https://zcash.flypool.org/miners/t1NEpmfunewy9z5TogCvAhCuS3J8VWXoJNv insane, am i rite? )
and WaveRiderx is probably mining another coin on another pool with obviously different difficulty,
so counting * an + is a waste of time.

I've checked the devfee, it is exactly 2%.

Ok, thanks guys, good to hear. 

[skrill79]

Sorry for the newbie question, which might be already answered - dstm is for zcash, or all mineable coins with equihash algo, i mean can dstm be used for zclassic, or zencash for example? Thanks in advance.

[Protofabtt]

I think your theory is right.

Beside, zm is mining devfee on zcash flypool (see here: https://zcash.flypool.org/miners/t1NEpmfunewy9z5TogCvAhCuS3J8VWXoJNv insane, am i rite? )
and WaveRiderx is probably mining another coin on another pool with obviously different difficulty,
so counting * an + is a waste of time.

I've checked the devfee, it is exactly 2%.

I am curious as to how you determined which pool/wallet zm was using?

[MagicSmoker]

Sorry for the newbie question, which might be already answered - dstm is for zcash, or all mineable coins with equihash algo, i mean can dstm be used for zclassic, or zencash for example? Thanks in advance.

Yes, just use an appropiate wallet address and mining pool.

[rabbi-t]

v.0.5.8 pool zen.suprnova.com
periodically i get such message in log
#  connection closed by server r:0
#  reconnecting
#  connected to: zen.suprnova.cc:3618
#  server set difficulty to: 000f0f0f0f0f0f0f0f0f0f0f...
#  connection closed by server r:0
#  reconnecting
#  connected to: zen.suprnova.cc:3618
share counter drops to 0
gpus getting cold for some minutes, then the process starts again until next reconnect.
is this a bad thing? never got share zero-drop with ewbf.

[jp26jp]

Will DSTM be adding additional algos?

[djn053]

V0.5.8 windows 10 with 6X1080
when I run a fews hours，always auto quit cmd，and then do nothing，It waste a lot of time until I know it.
how can I fix it？PLS help！

[lankyman]

I have a strange issue with Linux version.
I`m new with Linux so the issue could be simple.
I installed nvOc v0019-1.3.

I started mining with ZM 0.5.2 - It`s working like a charm!!!
I set up a telemetry to 0.0.0.0:42002. In windows I make the same and it uses the address of computer. And I could enter it via, for example - 192.168.0.1:42002.
With windows version it works, but with Linux version telemetry does not open via browser.

Could you please say me were I`m wrong.

Here is my launch
./zm --server eu1-zcash.flypool.org --port 3333 --user t1R5MDRcwGKqLr9n7v4o1s8TrvjQsVeWkyc.ZM_Ubuntu_1080TI --pass x --telemetry=0.0.0.0:42002
Thank you in advance.

Command line looks fine.                                                                                                       
Did you got your IPs right?                                                                                                   
Does 'http://127.0.0.1:42002' work if try it in a browser on the same system zm is running on?

How can you make it run on specific graphics cards only.
If i have 5 GUPS and i only want to use GPU 1-3
example, -d 0,1,2

[Supercoiner111]

Pool failover still on the books for the next release?