Well I do not influence the clocks of your card at all - I just run a program there, but do not set any clocks or fan speeds nor voltages - nothing. lolMiner does not even read this values, because in OpenCL standard its not designated to do so.

So if your card is not following your power target setting, this is an driver / overclocking tool or cards bios bug, but non of the miner, since it just takes no influence in that.

It seems that this was a pool problem - almost all pools were down today for some time since the complete mining ecosystem of the coin was under attack.

As for the random crashes no solution yet, but always try to make it more stable. Already working on new ideas for that.

@ All that send me their mnx address because of the bug / new miner bounty campaign:
The payments just left my wallet!  
It remained approx 11 mnx unclaimed, that will reach the faucet soon.

You can see the transactions done on this address:
https://minexexplorer.com/?r=explorer/address&hash=XJ47v5a4Axv192U8Q9CjkidvwHh9CQZz3u
Here you see one outgoing transaction splitting the value in two parts, one smaller (to one of you) and one larger (as self transaction from the wallet software).
clicking on the larger one you can see the next outgoing transaction and so on. Overall there should be 30 transactions in this chain right now. The 31st will be the one to faucet

Thanks you all very much for your support!  

That is intetesting - so you set the platform number to a non existing value. Does this also happen with --platform AUTO?

Just updated the binary - had a small bug that --optim STD did not equal the default (when --optim is not set)

For next version I will try stabilizing the host part more - have an idea but must try how well this works

Time is running fast and coding never stops.

Although its only a small update, I would like to share lolMiner-mnx 0.31 with you

For those who do not want all the info whats new, here are just the download links:

Linux 64: Click me (v0.31)
Windows 64:   Click me (v0.31)

But why to download that without need? Well, this is new:

Fixed a bug in the Stratum Interface causing segfaults when receiving new jobs
This did not occur very often and most likely when the miner connected, but well - it happens from time to time and after some searching I found out why. But it seems I will have to do a new json interface for 0.4 - what fixes the segfault is rather a dirty fix and it only covers approx 80% of the possible reasons. Will try to come back with something better.

Introduced a new parameter "--optim"
This takes the two values OFF and STD (default). Setting the parameter to OFF will turn back some optimizations done for 0.3 that cause some lower end or older GPUs to run slower. So this kind of brings back the optimization level of 0.22 if you liked that more. Mid and high end cards will mine faster with this parameter not set (or the default STD), but also consume more energy. So the value OFF is also an option for those running into thermal problems.

If will extent the parameter in future for vendor specific optimizations. Lets see.

Introduced a new parameter "--use-compact-accept-msg"
This will replace the "Submitting Share / Share Accepted" message pair by * symbols attached to the 5 seconds statistics. This automatically activates --use-compact-short-stat, so we now have the following options for lolMiner-mnx to look like:

Default look	--enable-time-print
Code: [Device 0] Average speed (5s): 22714.3 sol/s [Device 1] Average speed (5s): 22586.9 sol/s Submitting share Share accepted [Device 0] Average speed (5s): 22544.1 sol/s [Device 1] Average speed (5s): 22776.2 sol/s Submitting share Share accepted Submitting share Share accepted Submitting share Share accepted [Device 0] Average speed (5s): 22045.2 sol/s [Device 1] Average speed (5s): 22045 sol/s	Code: 12:15:12 [Device 0] Average speed (5s): 22660.5 sol/s 12:15:12 [Device 1] Average speed (5s): 22526.5 sol/s Submitting share Share accepted 12:15:17 [Device 0] Average speed (5s): 22494 sol/s 12:15:17 [Device 1] Average speed (5s): 22618.8 sol/s Submitting share Share accepted 12:15:22 [Device 0] Average speed (5s): 22020.2 sol/s 12:15:22 [Device 1] Average speed (5s): 22010.2 sol/s Submitting share Share accepted 12:15:27 [Device 0] Average speed (5s): 23222 sol/s 12:15:27 [Device 1] Average speed (5s): 23155.6 sol/s
--use-compact-short-stat	--use-compact-short-stat --enable-time-print
Code: Average speed (5s): 22643.7 sol/s | 22736.9 sol/s Total: 45380.5 sol/s Submitting share Share accepted Average speed (5s): 22485.7 sol/s | 22536.9 sol/s Total: 45022.6 sol/s Average speed (5s): 22606.5 sol/s | 22537.9 sol/s Total: 45144.4 sol/s Submitting share Share accepted Submitting share Share accepted Average speed (5s): 22599.5 sol/s | 22529.1 sol/s Total: 45128.6 sol/s	Code: 12:21:43 Average speed (5s): 22705.1 sol/s | 22755 sol/s Total: 45460.1 sol/s 12:21:48 Average speed (5s): 22546.3 sol/s | 22667.3 sol/s Total: 45213.6 sol/s Submitting share Share accepted 12:21:53 Average speed (5s): 22578.1 sol/s | 22612.7 sol/s Total: 45190.8 sol/s Submitting share Share accepted 12:21:58 Average speed (5s): 22595.9 sol/s | 22635.1 sol/s Total: 45231 sol/s Submitting share Share accepted
--use-compact-accept-msg	--use-compact-accept-msg --enable-time-print
Code: Average speed (5s): 23167.9 sol/s | 23124 sol/s Total: 46291.9 sol/s ** Average speed (5s): 20888.2 sol/s | 22592.3 sol/s Total: 43480.5 sol/s Average speed (5s): 20381.9 sol/s | 22738.9 sol/s Total: 43120.8 sol/s Average speed (5s): 20863 sol/s | 22592.1 sol/s Total: 43455.1 sol/s Average speed (5s): 20359.9 sol/s | 22602.3 sol/s Total: 42962.2 sol/s *	Code: 12:23:52 Average speed (5s): 21990.8 sol/s | 22753.8 sol/s Total: 44744.7 sol/s 12:23:57 Average speed (5s): 21497 sol/s | 22487.4 sol/s Total: 43984.4 sol/s * 12:24:02 Average speed (5s): 20862.4 sol/s | 23229.8 sol/s Total: 44092.2 sol/s 12:24:07 Average speed (5s): 20294.7 sol/s | 22661.5 sol/s Total: 42956.2 sol/s ** 12:24:12 Average speed (5s): 20326.9 sol/s | 22620.1 sol/s Total: 42947 sol/s *

So, pick what you like most
(Edit: you may see some speed differences between cards ... well the card 0 in this example running into heat problems rather soon and throttles - just that you know)




Did you try the new --exit-on-zero parameter of 0.3? That should close the miner after at most two minutes after the first 0 sols was detected and the scripts provided with the miner (.bat file in Windows, .sh in Linux) should be able to restart it. Does this work for you? If you want a quicker restart you can set the long stats interval shorter.

The latter is fine - for most Windows systems we have platform 1 when an Intel CPU is there.

Can you check on CPU load with all 7 cards? If the overall system load is >70% this may reduce speed because the Miner does not get enough attention by the operation system and thus the work is not scheduled fast enough.



It can be that your system runs out of memory. The Windows AMD drivers (not sure about the Linux ones) require that at the creation time of the GPU buffers there is as well RAM or virtual system memory available as requested GPU device memory. So with HIGH this means that virtual memory file has to be about 7.5 GB per card!

This video gives a good hint, how to increase virtual memory size on your system (For lolMiner it should be more then the 16g in the video, but the way to set it is the same) :
https://www.youtube.com/watch?v=ZL0m7tVt_DU

You could also try running less cards or --set-work-batch 2G instead of --set-work-batch HIGH in the beginning. Would like to know if this changes your cards from not working to working

Edit: Other reason indeed can be some driver issue - there is reason, that on Windows the default is 2G instead of AUTO or HIGH ... just because we had some trouble with AMD driver ...

what happens when starting first 4 GPUs in one instance and then after a minute or so the next in a second? Virtual mem is mostly required on startup as far as I know.



Hmm - never tested such a mixed rig. Does this also happen when you run them sorted. e.g. starting with device 0,3,5,7 (only the Ti) - does it work then?

Nothing... seems to be a bug - will have an other look to code what could go wrong there. Is the problem reproduveable?

No all fine. 0.22 was very slow with AMD Polaris cards - it should be more with 0.3 and high work batch numbers.

Lets get this one more stable first xD
I don't know how well the algo will work for other Equihash parameters. Especially the 192/7 consumes a lot of memory and lolMiner currently profits a lot from lots of memory any ways. Of cause I could give a try.
But one thing is certain: unless I find a golden way to get it more stable I would never do a ZCash (200 / 9) version for Windows, because that would cause extreme fud due to miner not stable xD



Well I can allow it to shut off monitor but was not sure if this is a good idea - because Windows going to power save is not good for mining performance  Never the less - doing a switch to turn it of is not so hard. I will include one in next version.


Depends on the Video driver. On Nvidia this is done automatically I gues, AMD handles things differently. If your miner runs without crashes, then you are fine.

I have heard from some people that the lower end Nvidia cards GT 1030, GTX 1050 (Ti) and older cards ( 8X00M, GTX 750 Ti and lower)  are slower with the new version and that for GTX 1060 its a close calculation because it is quicker, but consumes more energy as well.

Therefore I will try to build 0.31 soon with a switch to turn back the optimizations if desired.

Thanks all for feedback so far!

Very strange... could only be that your driver really stops the OpenCL execution, because Kernel and host side are now dead lock free - so even when code does not finish properly it should schedule new Work for the GPU.  Just ordeted own hardware to do tests at home by the way.



As it appears we have first known problem with low end cards (gtx 750, gt 1030, gtx 1050...) for these chips the new algo seems in fact slower. I think I will do a switch to turn off code changes for this cards.

As said about the 0 sols: no idea why this happens. Anyone else?

Hi folks,

well it took some time for me, but finally it is there: lolMiner-mnx 0.3
And its the first month birthday of the miner as well - time to party 

As usual here are the download links:

Linux 64: Click me (v0.3)
Windows 64:  Click me (v0.3)

It will take a bit to describe what is new, so lets start.

Introduced a complete new Stratum Interface

As you may know the stratum interface of 0.22 was very buggy and caused serious problems on connection losses as well to the user as to the dev fee pool. Therefore 0.3 now has a complete new Stratum interface that should handle the situation much more gracefully and stable.

That said the new interface also comes with a new feature: fail-over support. You now can pass comma separated lists  to --server, --port, --user and --pass. The miner will try to connect to the first server in the list with the first credentials. Once there are too many reconnection attempts for this first connection the miner will go to the next pool in list with the 2nd user name and so on.
With --max-connection-attempts <number> you can define how many connection attempts to be used before the miner switches over to the next pool & credentials in list. The default is 5.

This new stuff also applies for the dev fee connection. To bring this down the current three biggest pools have to be down simultaneously. Moreover the dev fee connection will try to use the same pool as the user for the fee, so the miner does no longer damage the balance of the pools by sending all shares to one pool. If this is not possible, it will do a round robin through all pools in the integrated list.

Even if the fee is completely blocked the miner will now continue to work, but slower then usual, because it will pause in between for reconnection attempts. The performance loss is approximately 5-10% depending on card and settings, but it will continue to work.

Improvement of the OpenCL code

As for the OpenCL code I did some performance improvements, that help hiding the memory latency (by almost 40%). This results the following performance table. I only listed the devices available for testing.

Nvidia

	lolMiner-mnx 0.22	lolMiner-mnx 0.3
GTX 970	unknown	10.2k sol/s
GTX 1060	9.9 - 10.1k sol/s	10.8-11.1k sol/s
GTX 1070ti	15.2k sol/s	17.5k sol/s
GTX 1080 FE	13.4k sol/s	17.1k sol/s
GTX1080 Ti	19.5k sol/s	22.5k sol/s

AMD

	lolMiner-mnx 0.22	lolMiner-mnx 0.3
RX 570 8GB	6.9k sol/s	8.9k sol/s
RX 580 4GB	7.4k sol/s	9.1k sol/s
RX 580 8GB	7.6k sol/s	9.8k sol/s
RX 580 4GB	9.9k sol/s	10.3k sol/s
with Mining BIOS
Vega 64	16k sol/s	~19k sol/s (to be confirmed)

All of the above numbers were generated with the Linux version of the miner with current amdgpu-pro or cuda drivers.

Note that the performance of AMD cards may be limited in Windows, see the commends to –set-work-batch below.

Beside this I added some new comfort features & changes:

--platform and –device
The platform parameter now understand the keyword AUTO, which is also default. This one will select the platform with the most graphic cards automatically. If there are two with the same number, e.g. and Intel integrated GPU and one AMD / Nvidia graphic card it will prefer the non-Intel platform.

Device now detects two special values ALL and ALLGPU (default), that will select all devices or all graphic cards in the selected platform. Note then when using --device ALL on a processor with integrated graphics the miner will mine as well on the integrated graphics and on the CPU part. For most configurations this is not recommended.

--set-work-batch
There are now three special values 2G, AUTO and HIGH. When the work batch is set to 2G the miner is allowed to use 2 gigabytes of graphic memory at most. At level AUTO it will use at most 60% of the graphic cards memory and at level HIGH 90%. Note that AUTO and HIGH operate on per-GPU level, so mixing graphic cards with different memory sizes is no problem.

For Windows and AMD graphic cards there currently is a problem with lolMiner-mnx and the driver preventing that the miner works with more then 2 gigabytes of memory. Therefore the default for the Windows version is 2G, while the default for the Linux version is AUTO. This causes lolMiner-mnx 0.3 to be slightly faster on Linux compared to Windows when using AMD graphic cards.

The restriction does not apply for Nvidia graphic cards. For this GPU it should be save to use AUTO on any operation system.

New parameter: --use-compact-short-stat

This parameter will make lolMiner-mnx to print the short statistics in a similar style as the long statistics but without the surrounding dashed lines. Therefore one single line will be used for all devices instead of one line per device.

New parameter:  --exit-on-zero

This parameter will cause lolMiner-mnx to check the speed of each device when printing out the long statistics. If an device is detected with 0 sol/s (thus it was idle on the long statistics intervall) the miner will quit, so it can be restarted from a watchdog script.

With all those changes I still may announce that working configuration files from 0.22 can be used with 0.3 and should work as before. But to use the new features you may like to modify it 

Beside the new comfort work I also tried to hunt down the memory access crashed and the reasons for reported 0 sol/s. This said it is natural that I may have overseen a problem, but it should be better now. If not please report.

That said there are two more smaller changes:
1. The dev fee is now reduced to 1.5% 
2. There is a new readme in .pdf file format shipped with the miner. This is now much closer to a real manual and should help to configure the miner


Bugs is a good keyword.
The fee collector collected slightly over 72 MNX for my early testers. That said I decided to round it up to 75 to make is a smooth number - or just because I have a good day 
But still I do not have all addresses yet, so here a remember to the following people still missing - Remember you have time until this Friday, 12:00 UTC  to claim your part (!)

From bitcointalk forum:

From telegram:

Beside this I will do one more announcement. Unfortunately I am not good enough in doing batch programming for a cool Watchdog, so I will put up an contest for that during next week. More details will follow.

Ok, maybe readme is not so precise, so I will try to explain it slowly:

What is a platform?

First of all it is important to know that OpenCL can be run on tons of different processors, but requires a driver for the processor. Platform is hereby the OpenCL terminology for an OpenCL capable driver. Different to CUDA OpenCL allows multiple different OpenCL drivers on one computer, since there may be multiple OpenCL capable devices in the system. Especially when your rig has an Intel CPU it often happens that the Intel graphics driver are installed as well, because they ship with Windows.

LolMiner now requires to know which driver (platform) to use, since it only can use one at once. This is controlled by --platform <number>. The platforms are numbered by integers starting with 0 (first driver), next 1 (2nd driver) and so on.

How to find out which platforms are there?
lolMiner offers the option --list-devices. When lolMiner is started with this parameter it will print out a list with all platforms that were found on the system. Additionally for every platform all devices are printed that are found within a platform. Note that we use the term device, because beside the GPUs (which are most common devices) the miner can also run on a CPU. For example on an AMD APU system --list-devices will often print out one platform (AMD) and then two devices, one for the integrated GPU and one for the CPU. The same is true for the Intel platform when your processor has integrated graphics.

After printing out the miner will close directly. To see the output in Windows you must run cmd (the Windows command line), navigate to the directiory where lolMiner-mnx.exe is in and then run lolMiner-mnx.exe --list-devices directly from the command line. Writing the --list-devices to the configuration file is not supported and running in the batch will end up in the miner closing instantaneously and you do not see a print out.

 How to set --platform and --device?

After running the --list-devices parameter you know what drivers lolMiner found with which GPUs. Also the order should remain always the same. Just look in the printout which number is attached to the platform you want to use (It prints out "Platform <number>: your platform name") . To select this platform use --platform <number> in the configuration file.

Below platform number and name --list-devices prints a list of all devices within the platform. Usually this list starts with the first GPU (index 0), then prints more GPUs and afterwards the CPUs and other devices in your platform. To select the GPUs to run the miner on add --devce 0,1,2...,n for the first n-1 devices you want to use. E.g. when 4 GPUs were detected before you can use --device 0,1,2,3 .


Hope this helps configuring


This is something that is a bit specific to AMD platform. Their OpenCL compiler is extremely aggressive and uses certain amount of memory and time to compile the code before it is send to your GPU. This is also the reason why the miner takes about 10 seconds before the "Start mining..." is printed.  I observed that it usually is enough to have enough swap space (Linux) or virtual memory file (Windows) to start the miner on AMD. More RAM should not be required since this has only effect on the first 10 seconds of the miner running before the actual mining starts.



Odd ... I just tested it in Linux - will check if thats a Windows issue and try to fix it for next version



The overall stratum interface has a lot of problems in 0.22. Therefore I made a complete new one for 0.3. This will handle things more gracefully.

Yes that may be. The main loop currently goes like this:

Look if any card has completed some work successfully.
If yes --> Offload old work, send it to tester against target, count sols found and Issue new work
if no --> Wait for a short time and look again.

Well while this works most of the time this loop has one flaw: If due to OC or other circumstances the OpenCL code fails in its execution (it may be enough one bit flipping in memory to make this happen) or the driver is forced to cancel the current work the condition on successful execution does not get fulfilled and the miner may deadlock. I will try to come up with something more stable or that is able to recover better from this situation, but I see a pattern that this happens more often when OC is in the game.

Look at the first lines of output: your miner tries to use the Intel Integrated Graphics! Its likely that you have to use --platform 1 or --platform 2 to get your cards detected and the miner running.



Hmm ... no I think thats some unsave freeing of resources. I hope this is fixed with 0.3.

Hi all - unfortunately I have an other announcement to do, but this time a bit sad  


The provider hosting the pool where 0.22 sends its dev fee too just gave a notice that today between 16:00 and 17:00 UTC there will be a short downtime of about 15 minutes.
What does that mean for lolMiner? I have to admit I do not know, since I changed some parts in 0.22 that should make the miner continue working, but I only can hope that it really does, because some of you had crashes when the miner had no stable connection.

So what you can do is: montitor the rig and set it offline when there is a crash and resume after an hour, or load older version 0.21 that uses an other pool for this single hour.
I am very sorry for this, it will be better with 0.3 - I am currently hunting some smaller bugs in the optimized code, but the stratum interface is now much more fail resistant. I think I get things done till Monday


Edit:
If you search the download links for 0.21 ... they are here:
https://bitcointalk.org/index.php?topic=2933939.msg30763779#msg30763779

Not in next version, but likely later - we are just at the start of what this miner could be





Could it be this one here:
https://support.microsoft.com/en-us/help/2718841/0xc0000005-error-message-and-windows-explorer-stops-responding-on-a-co

I hope this is fixed with new version since it has much better thread handling (in particular on the stratum side of code)

@ol92: hmm ... have not tested much two seperate instances ... will have to do so.


@ALL:
Important Announcement (In particular for the early adopters of lolMiner):

I grabbed all the messages of this forum thread / private messages and telegram messages from the first two weeks and finally came up with an excel sheet of names and contributions to the lolMiner success.
As promised: 50% of the first month income will be for you - to get the exact amount we will have to wait till Monday (should rename it to Moneyday ^^), but here are the names of those who are eligible to claim their fraction of the fee:

From Bitcointalk forum:


From Telegram:

If your name is on the list, what to do?

Create a new mnx address and send it to me in a private message, either here or in telegram (if you decide to do in telegram but are here in bitcointalk list do not forget your nickname). Please do so before Friday, March 16th. 12:00 UTC

I think my name is missing...
You contributed with sol/s report, virus check or bug report in the first to weeks (before Feb 27)? Well then also send me a private message with link to the post and I will add you if I have overseen the contribution.

And how much will  be payed out?
I can not say jet precisely. Currently we are at about 45 MNX to be payed, but with next half week still counting this will be approx 55 to 60 in total. Also as said some posts before: unclaimed shares will be split: 50% go to the faucet and other 50% go to people who gave me an address to send to.  From the splitting: 15% go to people sharing hash-rates, 4% for virus checks, 37 for bug reports (in two categories depending on details), 15% bug confirmations, 15% for providing fixes and solutions for users,  10% private testing and 4% for misc contributions.

From the numbers I have till now is that least payed value will be at approx 0.35 mnx and max between 4 and 5 mnx.  I hope you agree that I keep the exact amount of each one in the list closed, but every one will get a private message reply to the msg with the address where I give the transaction.

So, thats it for the moment, looking forward completing v0.3


------------
Current stats:
Hash-Rate of all lolMiners out there: 49.8 Msol
Market share: 52,5 % 

You are amazing guys, thanks for that support!

Well not in 0.21 and 0.22. Before it was very extreme, but the more recent version should not have this problems. By the way I myself also get about 13.5k on a founders edition GPU but have seen 1070 cards going above that. It depends extremely on the cards clocks, especially memory timings (rather then memory clock), so for every card there is kind of sweet spot in mem clock.

I hope that its a bit more consistent with next version


You really should not use that version any more. I have to admit that its just instable. First versions of this miner were rather like tropical fruits in mid Europe: get ripe at customer xD

Almost ... you forgot one "-"
its  lolMiner.exe --list-devices

If you put that in a bat the problem will start, print but close instantly ... therefore writing it to bat is no good idea. With cmd is better


No, there is no maximum - we had people running 6+ cards. Only problem may be if that one is used for graphics output or the CPU is too  weak, because each solution found has to be checked by CPU if its below target and then send to host. It seems somewhere your rig runs out of resources.