|
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*
|
|
|
|
Hi all out there having problems with passwords or multi-GPU in Windows ... Unfortunately I included some broken .bat file where I confuse PASS and PASSWORD... so here is how it should look like: @echo off
setlocal enableDelayedExpansion
Rem #################################
Rem ## Begin of user-editable part ##
Rem #################################
Rem Insert your pool name here
set "POOL=eu.minexpool.nl"
Rem Insert your pool port here
set "PORT=9999
Rem Insert your user and worker names here
set "USER=XQZoBeGPZJoLhTdkcF8bScsNH5Wyesd9g4.worker0"
Rem Insert your worker password here
set "PASS=x"
Rem Insert the index of the platform you want to run on. Run lolMiner-mnx.exe --list-devices
Rem to see all available platforms and the available devices for each platform
set "PLATFORM=0"
Rem Insert your devices to run the miner here. For multiple GPU pass a comma separated list, e.g.: 0,1 for using two GPUs
set "DEVICE=0"
Rem Uncommend the next line (by removing Rem) to enable the Nvidia CPU load fix
Rem set "HOTFIX=--enable-nv-hotfix"
Rem #################################
Rem ## End of user-editable part ##
Rem #################################
set "PARAMS=--server %POOL% --port %PORT% --user %USER% --pass %PASS% --device %DEVICE% --platform %PLATFORM% %HOTFIX%"
:MINE
Start /wait /high lolMiner-mnx.exe !PARAMS!
goto :MINE One more fix for 0.21 in the bank  |
|
|
|
GPU load is only 10-15%
But I asked for CPU not GPU ^^ Better try that way: #!/bin/bash
POOL=mnx.suprnova.cc
PORT=7077
USER=andreassantonio.Andreassant3
PASSWORD=1
PLATFORM=0
DEVICE=0,1,2,3,4,5
cd "$(dirname "$0")"
while true
do
./lolMiner-mnx --server mnx.suprnova.cc --port 7077 --user andreassantonio.Andreassant3 --pass 1 --platform 0 --device 0,1,2,3,4,5 $@
if [ $? -eq 134 ]
then
break
fi
done
By the way: you did not use the hotfix ... you are running on AMD? Then I would recommend to use optiminer, that one is faster for AMD right now. For Nvidia you should add the --enable_nv_hotfix, because else it is likely that 6 GPUs will cause a lot of load and a segmentation fault - unless you have some Ryzen Threadripper or Intel equivalent in that system. for me too
2x1070+2x1060 = speed like 1x1070
and segmentation fault
Can you show how you start the miner and what is the output of lolMiner-mnx --list-devices? |
|
|
|
How is your CPU load? Usually each GPU gets its own worker thread that is capsuled of the others. As for the .bat problem: did you write in a password? When --pass is send to lolMiner the next element is treated as password. If you do "set "PASS=" this kind of kills the --device parameter. What you can do is figuring out ideal commandline to run the miner without the .bat and then copy over the whole string to bat... that will also work  |
|
|
|
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") As usual here are die Download links: Linux 64: Klick Me (v0.2 beta) Windows 64: Klick Me (v0.2 beta) Change log:Added simple watchdog scripts for the minerWell these aren't really high level software, but for the random crashes this will do the job. On linux restart of crashed miner is almost guarantied with that, for Windows I am not sure about what happens when Windows thinks that miner is hanged up. Lets see. If there are more problems I will pay some of the batch experts here to do a proper watchdog for lolMiner  Please do not leave the miner unattended in the beginning. Due to the long list of modifications I did not test this very deeply... Added multiple GPU supportUnder the hood there were a lot of changes required but here it is: Just add --devices=a,b,c to run lolMiner on your devices a,b and c on the selected platform. As before use --list-devices to see which devices are there on your machine. For each device there will be sol/s displayed separately plus every 60 seconds the average over all cards over the minute. When one GPU does not send any sol/s for about two minutes the miner is closed so it can be cached up by the watchdog for a restart. Added --set-work-batch switchThis switch will control the load that is send to the GPU at once. The value may be between 1 and 4. For Nvidia GTX 1XXX series I observed that 2 (default) is ideal, while AMD Mainstream graphic may work faster (5 %) having this set to 4. Most extreme example I have seen was some Nvidia Tesla card jumping almost 25% when this is set to 1. So you see: there is no ideal value for all cards out there and it is not much tested yet. But having a switch more is better then no option to choose Fixed immediate reconnect bugSome pools had trouble with lolMiner that after an unsuccessful connection attempt it directly tried to reconnect. Now it will wait some time before to give the pool a chance to recover from the problem. Fixed boost async IO crashSometimes it happened that one thread closed the pool connection while an other tried to write on it causing the miner to crash. Now there should be a block taking this exception and print a warning. Afterwards the miner tries to reconnect. Should be more graceful than before. Fixed invalid share bugDue to some unclean cleanup of nonce informations the first ordinary mining attempt after the shares of dev fee were computed was messed up. This did not happen every iteration, but will have cost about 1% profitability on pool side. Pools did not punish sending these faulty shares till now, but will do in future. Now this is fixed and should provide you slightly better income. Alert: Due to this bug which may be serious for some pools I would like to announce that the 0.1x versions are kind of deprecated now. If possible switch to newer version asap. [Nvidia] Fixed high CPU load after long runtime bugWhen lolMiner.mnx ran for a long time (yea indeed that happend) the CPU time it consumes increased. Now lolMiner resets the settings for the --enable-nv-hotfix switch internally. That should help reducing CPU load on a long term. For rigs with many cards it still may be high and it still may be better to run lolMiner-mnx on less cards then are available. [Windows only:] Switched some of the compiled in librariesThe libraries included in 0.13 were identical (but invisible) to those of 0.11, while the ones in 0.12 were newer. Now in 0.2 the ones fit to those of 0.12 which were more stable for some users. Hope this helps stabilizing the miner in Windows. ------- And here the current statistics: Hashrate of all lolMiners out there: ~5.25 Msol/s Market share: ~ 9,6 % Till now I collected 8.1 mnx as dev fee, half of this will be spread out as bug bounty. Also: Announcement:At the first monthly birthday of lolMiner-mnx, which is March 12th there will be a new version reducing the dev-fee to 1.5%. |
|
|
|
Danke für die Zahlen Das mit dem schließen ist so eine Sache die ich noch nicht ganz verstehe - es ist unter Linux exakt der selbe Code und da klappt es gut. Dieser crossbuild ist derzeit noch nicht ganz das gelbe vom Ei, aber besser als keinen Windows build an zu bieten. Der hotfix hat damit auch nichts zu tun, denn der soll eigentlich nur die CPU Last klein halten. Gut zu wissen, dass das offenbar nicht immer klappt oder nur Anfangs - mal sehen ob das noch besser geht. Eine alternative, die ich mir vorstellen kann ist später einen Core auf 100% last laufen zu lassen der permanent schaut ob eine GPU fertig ist und die dann wieder anwirft. Dann ist die Last zwar hoch aber nur auf einem Kern bei mehreren GPUs. Damit sollte dann auch ein i5 bis zu 8 GPUs stemmen können und ein Celeron 3xxx (wie auf dem Octominer board) sollte wenigstens 6 packen) ... nicht ideal aber für große Systeme besser als der ist-Zustand ^^ ------------- Edit: Current Hash for all lolMiners out there: ~5,25 Msol/s Market share of lolMiner: ~8,5% |
|
|
|
Isnt same issue on lolMiner-mnx from Lolliedieb? Dev fee is targeted on dedicated port (which isnt published on main page) on suprnova?
Currently lolMiner fee is working very simple: One out of 50 GPU runs the miner uses a 2nd connection (yes to suprnova) for fee submission. From the current difficulty and the much lower hashrate thats 1-2 shares per minute I would guess. There are no further connections and no other algorithm running but the one in the main thread that mines for you. You can monitor this in wireshark if you like. In a future version I will switch from suprnova to an own server, but else not much will change. lolMiner will never run an other algorithm in background - for fee it just does what it does for you. Currently lolMiner has 2%, but I promised in the lolMiner thread that I will spread 50% of my first month income to those people helping me finding bugs during the first two weeks of existence. Furthermore after this first month next version will reduce fee to 1.5% and in first post of thread I already gave some conditions on which I will lower it to 1%. And if I ever stop giving support for this I will clean up and release sources. Hope this serves you |
|
|
|
Maybe you can tell me which driver to install on Ubuntu 16.04 to use AMD card?
There are two that work. One option is to install amdgpu (opensource) with mesa-opencl-icd and libclc-amd-gcn. This is the free variant that will work with any Linux distribution. For Ubuntu with long term support you also can try amdgpu-pro driver (without the other two libs mentioned above). These are the official AMD binary drivers and are likely faster than the free ones but a bit more tricky to install. But currently for an AMD rig running Linux you should prefer optiminer because its currently about 30% faster then lolMiner on AMD. Will try to close that gap in future of cause |
|
|
|
|
Hi all (that have problems with occasionally dropping hashrates with --enable-nv-hotfix):
Let me start with an explanation what exactly is the problem and how the hotfix-switch works.
The problem is that in the miner the CPU has to wait for the graphic card to complete its work. Afterwards the CPU will continue doing target comparison and then submit while the graphic card gets a new work to process.
The waiting for the GPU to complete is done very differently by vendors OpenCL drivers. For AMD and Intel the runtime sets a kind of callback that is automatically triggered when the GPU has completed. So far so good. On Nvidia it is rather simple: There is a tiny loop that frequently tests if the GPU is ready with its work and then exits once this is done. (By the way: in CUDA one can choose having the one or other solution) Unfortunately that simple this loop seems: it produces 100% CPU use for almost nothing useful.
What now does the hotfix. Well I try to measure the runtime of the GPU code and then let the primary mining thread sleep for a while such that it returns shortly before the work is done. Then after the sleep the Nvidia loop is entered to wait for actual completion. Currently of the full time the thread sleeps almost 90% of this time. This 90% can be used by other processes to do other work or let the CPU idle around and allow it to cool down. Simple solution isn't it?
Well unfortunately not, because the sleep command on many systems gives away the "active task" flag of the task and only guaranties that the thread sleeps for a certain time. When my thread actually awakens massively depends on the system load. If the system is idle it is very well working and also if enough resources are avail. But when a lot of miner instances are running, some important system demon jumps in with high priority or whatever taking systems attention (e.g. anti virus scanner), then it may be that the sleeping time is much longer than it should be. And this causes the drop in hash rate.
I observed that one can mitigate (not fully solving) this problem by adjusting the priorities. In Windows task manager you can set the priority of lolMiner from "normal" to "higher" (please not "highest" or "realtime" or something like that ... will crash your system for sure) and on Linux based systems you can start it with "sudo nice -5 ./lolMiner-mnx" to increase its priority. The consequence will be that the system scheduler will pay more attention towards the miner getting the computation time it requires.
This will not completely solve the problem but will make the drops more rare and less high.
|
|
|
|
guter miner lolliedieb! schön wäre wenn man mehrere gpu´s in einer konsole starten könnte
meine rx vega64 schafft 16k sol´s mit blockchain driver
Ich tüftel da gerade dran - hoffe das Feature in der nächsten Version (1-2 Wochen) fit zu haben Muss dafür unter der Haube etwas ändern aber sollte machbar sein. |
|
|
|
is there in miner some watchdog script?
There is one made by a community member earlier in this thread (first 3-4 pages). But I plan an official one for Windows version in 0.2 (hopefully beeing ready next 7 days) |
|
|
|
Ubuntu 16.04, G3900, 4GB
I can start only 5 threads. System has freeze when I start 6th thread. With 7 threads I can't tested. Optiminer work correctly on this system. Now start 4 GPU with lolMiner-MNX and 3 with optiminer
Its surprising that the G3900 does 5 cards at all. That CPU is very weak (only two threads) and due to Nvidia high CPU load with OpenCL and the work to be done sha-ing the results it is great to know it does 5 of these. Optiminer somehow found a better solution to handle the Nvidia OpenCL CPU load. I will try if I also can do better because the current situation where hash-rate drops from time to time and more cards make the system freeze is well ... undesirable ^^ Maybe thinking one month ahead I think I would like to implement a switch shifting the sha256 calculation to GPU as well ... then a G3900 should be able to power 6-8 cards when the tiny processor does not have to do the target calculation itself... |
|
|
|
|
I just made an interesting observation regarding headless displays and maybe also applies for ZenoDiac:
In Windows 10 it happens thst the order of Platforms is volatile when Intel GPU drivers are installed, because with monitor attached or much stress on GPU the Nvida (may also happen with AMD) is platform 0 but without anything platform 0 is Intel and platform 1 is Nvidia. Therefore it can happen that you instead run the Intel iGPU or even CPU when in wrong order.
I will try implementing something that allows auto detection, but currently when platform 0 device 0 is only giving 3 digit sols where there should be more than 5k, then try again with platform 1 device 0. Sometimes this may fix it.
|
|
|
|
|
Hi out there.
There was a bug in the initinal binaries of lolMiner-mnx
Everyone that took his copy before 16:30 UTC should reload the current binaries!
|
|
|
|
|
Both binaries updated - should work now.
It seems 13 is a bad luck number ^^
|
|
|
|
Hello Dev! Whats wrong?  Win10 CPU: DualCore Series E8700 GPU: 1063*1 Thanks for reporting. Links for v 0.13 removed .... there is a small bug with shares submit ... will be fixed in 10 minutes Edit: Fixed for Linux download. But Windows will take some time. |
|
|
|
And good news #2 for today: lolMiner v0.13 is there. And here is the link to thread: Click meThis one should get rid of the large number of disconnects due to wrong number of parameter send and connection timeouts. Also did a step to improve (hopefully) Windows stability. Of cause was not able to test this much because of the short release cycle but too many new things to hide this version from you. Happy mining |
|
|
|
umm...Although the miners seemed to be stable, they suddenly crashed earlier... It seems to be unstable yet.  Windows10 x64, GTX 1080 Good job, Lolliedieb. No reason to get sarcastic  Well I did not plan to have an other version today, but there are to much changes to hide it from you: Time for version 0.13As usual here are die Download links: Linux 64: Click here (v 0.13)Windows 64: Click here (v 0.13)Change log:Resolved the "wrong number of parameter" disconnect bugNow lolMiner-mnx is more flexible regarding this and interprets other sizes correctly. Thanks to Raquemis for helping me finding this. Fixed certain bugs with the stratum connector.Some small fixes that will keep the connection more stable and prevent the miner from disconnecting / crashing. A small adjustment to the "--enable-nv-hotfix"It happened that the system clock was to precise to measure the timings for the switch. Now modified the routine that takes care of this. CPU load might be increased slightly, especially for cards with very few compute cores, but the impact of the switch to the hash-rate should be decreased. [Windows only] lolMiner-mnx is now a static build for WindowsOn some machines stability issues came from older libraries in the system conflicting with the ones shipped with lolMiner. This should not happen any more. I can not promise that this fixes all Windows issues but it gives me more control over the situation. Note that for Windows build also the structure in the ZIP has changes, since the whole ZIP now only contains two files. ---------- Current hash-rate of all lolMiners out there: ~3.09 Msol/s Market share: ~6.3% |
|
|
|
As the software's dev I am obviously interested in a possible solution. I hope this does not mean frequent reconnects to pool since this is a lot of stress for the mining pool. Anyways - If you help me getting things stable that would not be for free anyways - remember I announced to distribute half my first income from dev fee to all that help me stabilize this miner and to improve the experience with it Also I would like to know more details about your system, since you seem to have massive stability problems. Valuable parameters for me: Exact Windows version, CPU you use and maybe a detailed error message if there is one. Some complained about having a libstdc++6 error on Windows. If you also had this can you check if there is an other version of this library anywhere on your system except for in the lolMiner directory. I have the feeling that this is sometimes the reason for the crashed. As said - this is not for free - every detailed description that helps reducing these bugs make you people eligible for claiming your fraction of what I want to distribute Ps: Due to a very productive live chat with Raquemis (dev / host of minexpool.nl) last evening I can announce that we found one glitch in lolMiner. Fixing that will eliminate the wrong parameter size issue you currently have sometimes on minexpool.nl and optiminer.eu. Also we found a 4 years old bug (!) in z-nomp ^^ Ok, nothing serious, but funny that it took so long for someone to find it ^^ Also for all AMD miners: under Linux optiminer currently is about 50% ahead to lolMiner, while lolMiner almost doubles the hash-rate of NVidia cards. This is currently just an optimization issue. As soon as my miner is stable I will introduce more optimizations for both GPU vendors lifting lolMiner hopefully on par (or even beyond) |
|
|
|
For all out there with stability issues: I just released a new version (0.12) of lolMiner-mnx. I can not guaranty that I canceled out all issues we had, but its unlikely that the new version is worse than the old Here is the change-log for the new version: v 0.12 beta:Changes to the "--enable-nv-hotfix" switchThe hotfix is now a bit less aggressive compared to v 0.11. It may be that the CPU load is slightly up compared to the previous version, but the new algorithm is doing better when recovering from a time with overall high CPU load. If the old version worked better at your side you can still download the old version v0.11 here (Win 64) and here (Lin 64) . In general for the miner monitor the load during the first minutes of running. If you have an overall system load above 60% I would recoment running less instances of lolMiner and use the remaining graphic cards for mining with an other coin. Fix of the "compile() error"The miner now also runs on older AMD implementations with the official AMD drivers. The most ancient card I tried was a Radeon HD 7770 (GCN 1, released February 2012 (!) ) and it runs fine with about 600 sol/s when using Windows 10. Fix of a "lucky user" bug.There was one strange bug in the miner causing a segfault when too many solutions were found with one equihash run. Usually on average about 2 solutions are found, 4-5 can happen, but I had one report of 10 solutions in one run. v 0.11 was not prepared for this, but now it is ^^ Improvements to reconnect function.Before there were multiple threads testing if the stratum connection is still valid and that may have issued a reconnect. This sometimes caused that thread #1 wanted to restart the reconnect attempt just started by thread #0. Now there is exactly one thread left that is able to issue a reconnect when an error is detected. Also after the first reconnection attempt the miner will now wait for 30 seconds to give the pool time to respond before the next reconnection attempt is taken. Shipment with new libraries in Windows versionWe had some reports about libstdc++ library crashes in Windows version. I updated all libraries in Windows version to newer versions. This does not guaranty that the stability issues are gone, but maybe the new versions help. At least I had no issues with the new versions in first testing. Happy hashing Edit: @Racquemis1: You got some private message. |
|
|
|
|
Show Posts
