[ANN] lolMiner-mnx, Mining MinexCoin(MNX) New version: v0.4 (July 22)

[Wa4A]

Windows 10, AMD Athlon II, RAM 4GB

3xGTX1070 do not work correctly

On start low speed

http://i6.pixs.ru/storage/9/3/3/02v3x1070s_1956580_29432933.png

http://i6.pixs.ru/storage/9/1/5/02v3x1070e_8729672_29432915.png

http://i6.pixs.ru/storage/9/2/4/02v3x1070e_1118952_29432924.png

http://i6.pixs.ru/storage/9/2/9/02v3x1070e_9714662_29432929.png

Same threads typing in same output in one of the time without queue.


1xGTX1070 show 10kSol/s, less than prev release

Likewise! Installed 5 cards 1060 * 3. When you run the first two cards - the hash holds about 10k, when I run all 5 cards, the speed drops to 100 sol

Add: Really! When you run 2 cards or two copies on one card, each card issues 10k +, but when you start the third instance, the speed starts to divide and the result is already 7-8k for each!

[1713974518]

1713974518

[Lolliedieb]

I see that there are many problems with that ...

I currently only have a two cards system and for this it works fine. I will rearrange the complete miner thread stuff for 0.21 and I hope this will fix it *crossing fingers*

[Wa4A]

I see that there are many problems with that ...

I currently only have a two cards system and for this it works fine. I will rearrange the complete miner thread stuff for 0.21 and I hope this will fix it *crossing fingers*

Wait)

[salshicha]

So, a bit earlier then I expected but since it is lolMiners birthday (one week old now) lets have a party

Time for version 0.2
(Also known as: "The first feature release")

Ok, got 5*1060 gpu working

But hashrate is ridiculous

[Lolliedieb]

I see that there are many problems with that ...

I currently only have a two cards system and for this it works fine. I will rearrange the complete miner thread stuff for 0.21 and I hope this will fix it *crossing fingers*

Wait)

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

[Dirtpaw]

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.

[oomurashin]

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

[Nihrupka]

maybe you should add logging? for tests in fields?

[dejan_p]

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

[Lolliedieb]

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 <n> / --set-long-stats-interval <n>
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 %

[kdb]

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

[Dirtpaw]

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

[mnxfl]

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.

[r12x2]

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

[mk111]

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.

[Lolliedieb]

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

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

[Lolliedieb]

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.

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?

[mk111]

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.

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?

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)

[mk111]

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

[elektrofusion7]

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)