Topic: [ANN] lolMiner-mnx, Mining MinexCoin(MNX) New version: v0.4 (July 22) - page 37. (Read 30772 times)

Heya,

That does make sense however it does not explain why v0.13 works flawlessly Any ideas?

You are running the rig headless?
Then this is a known issue of drivers in Windows, because drivers clock down the card when they detect no active display.
I belive that in  CUDA somehow one can deactivate this, but as far as I know there is no software solution I can do with OpenCL.
What you can do is plugging a fake HDMI plug - this is also used by some Ethereum Miners, see here:

https://www.reddit.com/r/EtherMining/comments/7qd4be/running_windows_headless_problem/

Alternatively if you have an old monitor in the attic this is ideal chance to do something useful with it

I can run two cards at full speed with v0.21 but as soon as I log out of teamviewer the hash rate halves straight away. If i log back in it goes back to full speed. Any ideas?

It works better then 0.20. On 290x I have 7400sol/s and on 270 I have 4300sol/s thanks to change in work size. On 0.20 it was 6600/3700sol/s for 290x/270. still this is low compared to optiminer where 280x can do 7300sols. Please work a little bit more on AMD cards. Thank a lot !!!!! This is for me only option because I have some issue with linux (ubuntu and amdgpu-pro)

Ok now V0.21 runs 12000 sols on both cards (1070s) after changing WORKB=8

With V0.20 I could run 2 x instances of the miner and both cards did 14000 sols but after a while card number 2 would drop speed between 3000 and 9000sols. CPU would run at 91% (Celeron G4560)
With V0.21 each card only does 7000 sols (1 or 2 instances it does not make any difference)

So two kards = half speed or two cards = same speed as with one? What happens with two distinc inszances of the miner vs one instance running both cards?

Hmm I did not test mixed GPU setup... tried two instances with one card each?

Edit: Looking over the code: it will be easy to simulate the old behavior ... (new code is very flexible)... I think I can do a simple switch to bring back the 0.2 way for Intel users and those AMD users where this caused less problems

Windows 10 and version .21 now enables me to run 2 x 1070 at the same time which is great but they only do half the speed.

Don't know what's wrong, but when I'm trying to start RX560 + R280 v0.21 is freezing or working with 0 S/s
v0.2 was stable for me

Great job!!!
Keep up the good work. Performance is fantastic but stability is key. I hope you worked out the kinks and it can stay stable. I'm running on 4 Pascal/Maxwell rigs with different versions of Linux and Windows 10 and will report back tomorrow on stability.

Nice work Lolliedieb! .21 works great compared to .20 (linux and 1070ti-s). Also liking the speed increase. Awesome work.

It work great, but only with display or TeamViwer

Windows 10, AMD Athlon II, 4GB RAM
3xGTX1070

With TV
http://i12.pixs.ru/storage/1/2/5/3x1070v021_8369054_29446125.png

Without TV
http://i12.pixs.ru/storage/1/6/8/3x1070v021_1306834_29446168.png



v021 more powerfull, but it overload GPU, and windows freezing
Windows 10, Intel i5, 8GB RAM
GTX1050 + v020
http://i12.pixs.ru/storage/1/7/4/GTX1050v02_4903780_29446174.png
stat
http://i12.pixs.ru/storage/1/8/0/GTX1050sta_4963692_29446180.png


GTX1050 + v021
http://i12.pixs.ru/storage/2/3/2/GTX1050v02_3250598_29446232.png
stat
http://i12.pixs.ru/storage/2/4/0/GTX1050sta_8045419_29446240.png

Hey guys ... any idea what I did last two days?

Well I have: maybe its was completing version 0.21 

Here are the download links:
Linux 64: Click me (v0.21)
Windows 64: Click me (v0.21)

I am quite happy with this ... why? Because it is like 0.2 should have been But see the release notes:

Introduced a complete new work thread model
The old solution caused massive problems with more then 2-4 Nvidia cards. Therefore I tried something new and what to say... I am happy with it. We tested it with one rig (I5) on 6-7 GPUs (GTX 1060) and what to say: It works This also means that the old --enable-nv-hotfix switch does no longer exist. The new solution will also increase the hash-rate on Nvidia hardware slightly (2-5%).

New parameters: --set-short-stats-interval / --set-long-stats-interval
The new parameters can be set to control the verbosity of lolMiner-mnx. It sets the time between the status lines printed to n seconds. If you set 0 or a negative number this disables the corresponding output lines. As for the --set-long-stats-interval this also controls the duration of the benchmark mode (that you can access when passing no user / server data to the miner)

Benchmark mode is now offline
Well I manged that no connection is required to test the capacities. Before it was mining to dev fee account, now this works even on machines without network connection.

Extending the parameter range for --set-work-batch
As before the default (2) is ideal for Nvidia GTX 10XX cards. The Nvidia professional cards (P100, V100 and Titan V) may profit from setting this to 1. For AMD cards: every even number will work well, but the higher the better. I measured that setting the parameter to 8 or even 12 (max) gave about 15% extra performance over the default. Setting a higher value also increases the memory use, so if the miner crashed on startup it may be better setting the value lower.
Even with this setting optiminer is still faster on AMD (to be changed in future ^^), but lolMiner-mnx consumes less energy (may be changed when GPU is better utilized). As far as I can estimate the two are now almost on par for sol/s per watt for RX 570 / 580 series. Try on your own which is more profitable in your concrete setup.

Some small changes in the stratum connection
Nothing to talk much about. I hope to improve stability with this. Note that the miner will now quit when it is not authorized from the server because the credentials are wrong.


Unfortunately there are two known issue:
Miner will crash on startup with Intel drivers
Well somehow the Intel drivers do not like the new work handling. Have to dig deeper why. For Intel 0.2 is the better version right now.

Miner will not quite when a GPU is detected with 0 sol/s
This feature had its problems in 0.2 and now with the new handling it does not work. But it will come back (with a switch to turn it on / off) in next versions.


Hope you also like the new version

-------
And here the current statistics:
Hashrate of all lolMiners out there: ~6.6 Msol/s
Market share: ~ 10,1 %

win10
one instance with one gpu works
two instances crash both of the miners within a minute (even if started from a different folder)
if more cards are started, hashrate drops

will try linux in the morning

keep up the good work and looking forward to new versions and fixes.. i don't mind the fee
thanks

maybe you should add logging? for tests in fields?

v0.2 worked fine for my 4-rigs.

ryzen 7 1700 + GTX1080*4(13.5-12kH/s)+1070*1(12kH/) CPU used 30-35%
ryzen 5 1500X GTX1080Ti*2(19-17KH/s)+1070Ti*2(13.5-13Kh/s)+1070*2(12.5-12KH/s) CPU used 95%
Pentium G 4560 GTX1080*2(13.5-13KH/s)+1070Ti*1(13.5KH/s) and CCminer another 5GPUs CPU used 95%
Celeron G 3930 GTX1070*1(13-12.5kH/s)and CCminer*1GPU + CastXMR(RX-VEGAs)*6GPUs CPU used 95%

thank you Dev this miner is great for pascal

Tested more v .2 . Hashrate is all over the place. 13k to 5k to 10k to 2k to 5 sols. Tested with three nvidia cards. 6x gpus restart the rig. Switched back to .13 for now.

On what? Obviously the solution I have does not work - in particular for Nvidia users that profit most from lolMiner-mnx ... So since bugs will not fix themselfs I'd rather start writing code It is not guarantied that what I try next is better than what we right now have but sure its worth a try

Ok, got 5*1060 gpu working

But hashrate is ridiculous