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

[Trololino]

Tested the windows 0.5.2 miner and its working faster for my 1070, but significantly slower for my 1080 rig (regular 1080's not Ti).

Hopefully it will be improved in the next patch

[N0pr3s3n7]

Hey guys , first review and problems:

First of all miner is better and more effective by 2-3 % than EWBF. I've been running it on several machines , got improved hashrate  (GTX 1060  3gb from 280 ~ to 300 ~ , GTX 1070 from 470 to 490 , GTX 1080 from 540 to 550 , gtx 1080 ti from 720 - 740  , all vary bettwen + - 1-5 sol/s)
Machines are working fine with ETHcontrol , or anykind that has some watchdog that either restarts miner or knows what to do. But so far they are running stable.


Problem :

I got one rig with 7 cards and it seems that it cannot work good with dstm miner. It has 5x 1070 , 2x 1060 3g


First of all , to run it properly I had to lower power limit ALOT on all cards

Standart settings at afterburner:  80 pl / +150 core / +540 memory  -----------> at this point miner and Ethcontrol comes in some loop that whenever miner is started it's always killed and restarted , because it hangs out too much time or shows 0 sol/s for cards (e.g. miner takes a lot of time to start and doesnt start at all)

Settings to start: 65 pl / +150 core / + 540 memory -----------> miner starts but this is what happens , after "1round of hashing good" it starts hashing super pooorly. I dont know what should be the case , all my rigs are identical , just this is the only one with more than 6 cards. All have same windows / drivers etc etc etc . CPU seems to be stuck at 100% at rig 1 though when I run ZM miner , probably some problems with more than 6 cards rig on win 10 ?


edit: I've tried a lot of different settings but in vain. Problems seems to be that  POWER for cards makes crazy jumps (35-80 % PL for example) . Only time it worked fine for arround 5 mins , was because power limit somehow settled down at exactly 80 % (what was set) and it was running okey. Making any changes to OC settings results in various jumps in power and thus cards start to hashing super poorly. This is the only thing i've discovered.

Thx for the report, this is very helpful.

I'm currently heavier on CPU load.
It looks like your CPU is to slow to provide enough work for the GPUs. CPU load increases with the total solution rate of your system - so this is most likely why your 7card rig is having troubles since it requires more cpu time. What CPU are you on?

It's intel 1151 , G3930  , it's not impressive but it's not on the downside of any processors.

edit. I got such processor on all my machines , and also planning to add up to total of 8 cards on each of them soon , so this might be huge problem to run ur miner @ them

[Andrey09]

Thx for the report, this is very helpful.

I'm currently heavier on CPU load.
It looks like your CPU is to slow to provide enough work for the GPUs. CPU load increases with the total solution rate of your system - so this is most likely why your 7card rig is having troubles since it requires more cpu time. What CPU are you on?

It's intel 1151 , G3930  , it's not impressive but it's not on the downside of any processors.

edit. I got such processor on all my machines , and also planning to add up to total of 8 cards on each of them soon , so this might be huge problem to run ur miner @ them

I have same cpu -and don't have problem. 8 GPU in rigs.

All fine, stable h/s
Aroud 95-100% cpu LOAD (G3900) on 8 gpu + teamviewer.

[djazher]

Hey, thanks for the good work.

Actually testing it on 2 6x EVGA 1070 rigs. OC settings : PL -65% | Core +120 | Mem +500 | AVG 4.3 sol/W.

So far, the hashrate is improved by 4% compared to EWBF.

+ Does the temp controller overtake AB fan control, as I understand when OP says "It is recommended to set your fan speed to 100% when using this setting.", or am I totally wrong ? So far, it does not, and fans just run at max speed. I definitely don't want to use 100% fans, they usually run at 45% max, and I'm pretty fine with that.

+ Is there a way to use different colors to separate displayed information ?

+ Maybe worth implementing an auto-restart function, as during a 1h network maintenance, I found my rigs had exited the program. Short connection shortage restart as soon as it re-connects, but longer ones seem to just exit.

[PCMining]

Guys, I have uploaded this DSTM's Equihash Nvidia Miner in our pool.
You can try pre-configured windows versiyon...


http://pcmining.xyz:8080/getting_started

PS: Don't forget to edit wallet adresses with yours...

[raven1322]

Rig 1:
7x ASUS STRIX 1080 Ti:
EWBF: 690 Sol/s in Average
dstm: 720-740 Sol/s in Average

Rig 2:
7x ASUS STRIX 1080 Ti:
EWBF: 687 Sol/s in Average
dstm: 705-720 Sol/s

DSTM wins

That's a nice improvement for sure ... I'm off to give this a try (Win10) ... over my 14 rigs this could make quite a difference 

Also the lower Sol/s on Rig2 is because it uses 1x1600W EVGA G2   compared to Rig1 that uses 2x1200W EVGA Platinum P2; I fool myself that Rig1 has a better Power input than Rig2, I could be wrong.
Also the settings is  Core +200 Memory +200 with Powerlimit at 80%. Each GFX uses around 195-200W. The temperature for each card moves from 39->47 degree Celsius. I will re-read the thread and try to fix the Api like I had on EWBF.

Thanks dtsm for a wonderful miner!

You must have won the gpu lottery for you to OC those 1080ti’s like that and still maintain those temps. I can only bring up my strix oc’s to +105/+135, afterwards frequent crashes occur. Also, temps stabilize around 54-58degC.

[asdpof]

Very good!! Got about 100 extra Sol/s usina this miner with 1080Ti.

Thanks !

[ivlog]

Windows 0.5.2 alpha

This is an alpha, so it might contain bugs, pls report them if you hit any.
Launch options are described on the OP, or simply run zm.exe without any arguments to get a description.
If you get an error about missing VCRUNTIME140.dll you need to install the 'Visual C++ Redistributable for visual studio 2015'

Big thx to santan for providing access to a windows system.

add plz info about installation Visual C++ Redistributable for visual studio 2015 on first page.

[DevonMiner]

Well I've been testing this today, and like many others I like it, certainly faster than EWBF - nice work DSTM - Thanks!

Only one rig is playing up, saying 'Application zm.exe has been blocked from accessing graphics hardware' ... I'll tweak Afterburner and see what happens.

I don't go for the fastest settings with full-on power, but settings that give a good bang per watt.

ZOTAC 1080ti 11gb 4-card Rig:
63% Power
+240 Core Clock
-229 Memory Clock
70% Fans
= Average 2600 sols @ 702w at the wall

ZOTAC 1070 8gb Mini 5-card Rig:
Power: 64%
Core Clock: +225
Memory Clock: -220
55% Fans
= Average 2065 sols @ 568w at the wall

Happy to have ditched EWBF for a faster miner with an active developer 

[PickAxe131]

The miner crash from 3-5 hours usage. Is there any --log function in the miner?

[CarlOrff]

Good job for this Windows version.
But, sorry to write that : I don't care about it ! I'm a Linux user.
Any project DSTM, for the next Linux version ?

[Nippy-ua]

max. support card ?

[xenonwon]

1060 6gb sols?

[chancefx]

--api needed

[dstm]

Hey, thanks for the good work.

Actually testing it on 2 6x EVGA 1070 rigs. OC settings : PL -65% | Core +120 | Mem +500 | AVG 4.3 sol/W.

So far, the hashrate is improved by 4% compared to EWBF.

+ Does the temp controller overtake AB fan control, as I understand when OP says "It is recommended to set your fan speed to 100% when using this setting.", or am I totally wrong ? So far, it does not, and fans just run at max speed. I definitely don't want to use 100% fans, they usually run at 45% max, and I'm pretty fine with that.

No it doesn't. ZM's temperature controller is a proper PID-controller - it measures the temperature and adjust the performance continuously. Fans on the GPU use also some kind of controller. So you'll get 2 controlling loops with feedback. So what might happen is: zm lowers the performance -> fan gets slower -> temp rises again -> zm lowers the performance even more -> fan gets slower again .... and so on - something like this or something in between might happen. You don't have to set the fan speed to 100% however it should be set to something constant to avoid 2 controlling loops.

+ Is there a way to use different colors to separate displayed information ?

Currently not. What information is hard to read currently / should be highlighted?

+ Maybe worth implementing an auto-restart function, as during a 1h network maintenance, I found my rigs had exited the program. Short connection shortage restart as soon as it re-connects, but longer ones seem to just exit.

Currently zm retries for about 10min and gives up. This is hardcoded. It exits after 10min such that you'll get notified if you're using notifications on process exit.

[dstm]

The miner crash from 3-5 hours usage. Is there any --log function in the miner?

Did you overclock your GPUs? No there is no log option, however zm checks all calls and prints an error message if something goes wrong.

[dstm]

max. support card ?

It's stated on the OP.

Compatible with devices having Compute Capability 5.0 and greater i.e. it should run fine on Maxwell/Pascal GPUs.

[dstm]

--api needed

ZM supports telemetry via an web interface or by json-rpc, check the '--telemetry' option.

[zzz.to.the.moon]

Hey guys , first review and problems:

First of all miner is better and more effective by 2-3 % than EWBF. I've been running it on several machines , got improved hashrate  (GTX 1060  3gb from 280 ~ to 300 ~ , GTX 1070 from 470 to 490 , GTX 1080 from 540 to 550 , gtx 1080 ti from 720 - 740  , all vary bettwen + - 1-5 sol/s)
Machines are working fine with ETHcontrol , or anykind that has some watchdog that either restarts miner or knows what to do. But so far they are running stable.


Problem :

I got one rig with 7 cards and it seems that it cannot work good with dstm miner. It has 5x 1070 , 2x 1060 3g


First of all , to run it properly I had to lower power limit ALOT on all cards

Standart settings at afterburner:  80 pl / +150 core / +540 memory  -----------> at this point miner and Ethcontrol comes in some loop that whenever miner is started it's always killed and restarted , because it hangs out too much time or shows 0 sol/s for cards (e.g. miner takes a lot of time to start and doesnt start at all)

Settings to start: 65 pl / +150 core / + 540 memory -----------> miner starts but this is what happens , after "1round of hashing good" it starts hashing super pooorly. I dont know what should be the case , all my rigs are identical , just this is the only one with more than 6 cards. All have same windows / drivers etc etc etc . CPU seems to be stuck at 100% at rig 1 though when I run ZM miner , probably some problems with more than 6 cards rig on win 10 ?


edit: I've tried a lot of different settings but in vain. Problems seems to be that  POWER for cards makes crazy jumps (35-80 % PL for example) . Only time it worked fine for arround 5 mins , was because power limit somehow settled down at exactly 80 % (what was set) and it was running okey. Making any changes to OC settings results in various jumps in power and thus cards start to hashing super poorly. This is the only thing i've discovered.

Thx for the report, this is very helpful.

I'm currently heavier on CPU load.
It looks like your CPU is to slow to provide enough work for the GPUs. CPU load increases with the total solution rate of your system - so this is most likely why your 7card rig is having troubles since it requires more cpu time. What CPU are you on?

It's intel 1151 , G3930  , it's not impressive but it's not on the downside of any processors.

edit. I got such processor on all my machines , and also planning to add up to total of 8 cards on each of them soon , so this might be huge problem to run ur miner @ them

Try using older versions of drivers, in Linux that had a huge impact on CPU usage.

--api needed

API is present, see my older conversation with dstm on how to read it.

Edit: wrong quote

[dstm]

Good job for this Windows version.
But, sorry to write that : I don't care about it ! I'm a Linux user.
Any project DSTM, for the next Linux version ?

Yes ofc, there were couple request I'll implement in my next version, currently I'm working on CPU usage reduction which both OS will benefit from.