Hey there, by any chance, are you dual mining?
BTW 3GB cards can't mine ETH any more, and probably, given time, 4GB will go the same way, but at least right now, 4GB cards are ok.
Things to check, Nvidia driver, (cuda) install ok, or DDU, and clean install Nvidia drivers.

Post your batchfile content, it's helpful to get a better understanding about your rig and expectations.
Also, how much VM are you running?

OK, Ursul0, you surprised me, THANK YOU, apology heartly accepted.
Kudos also, most people are too pig headed to do it these days.

Hey so to be clear, I don't disagree with your views on cloning, simply that as a posted cloning solution to the problem tabled.
Assuming that will work in all and every case, is a generalisation, very much dependant on specific hardware, means and ability of the person doing it.
Those guys are running a single rig, and on a learning curve, and as you well know, that is a really tough combination.
Even if you have a tiny "other" rig, at least you have the vastness of differentiation available when you're trying to troubleshoot.

Anyway, all water under the bridge, they had already done an OS install.

As you asked, to clarify, why I don't clone for clients on mining rigs, (and I DO CLONE, also with Acronis, just as you do), but not for clients.
Why? Protection, mine and theirs.
Mining involves money, my clients money.

If I make a mistake, somewhere, and have cloned 50 rigs off that, and later I find out I messed up, that is a 50x mistake multiplier I'm not prepared to risk. You can mess up just as easily by something you don't do.
With single install, I'm isolating risk, constraining it to THAT rig.`
So to be clear here, this is my CHOICE in this context, but equally none of us want to clone/deploy our mistakes either.

While it might not be THE solution, it's the one I've adopted to date. I've not had to consider building more than one or two rigs, single builds, or tiny batches if you like, so clone rollout is not a big time saver for me, and certainly does nothing to build customer confidence.

Of course I have a bunch of reg files, and envrio tweaks, so the additional deploy time is minimal, and I usually multi task while doing those build stages, so I'm not idle in that time.

One thing I often see in tech support is what seems simple to you, me, is often too complex or outside the comfort zone of the people asking for help.
Usually I probe with a few questions to get a feel for their ability and experience level, and get some dialogue going, before settling on what advice to give, and walk them through it, (not so easy on a part time forum basis).
But assuming they have, blah, blah means and ability, etc etc, actually assuming anything about them at all, is my error, and usually going to cause them more issues.

Take your logical method of driver install, (extract and do it in devman), absolutely, I could not agree more, however, while that is known to you, it's not the AMD advertised method, they steer the user to setup.exe, which does usually work, time consuming as it is, but is that the "best" method?
Well, probably for the majority, yes it is according to AMD, which is why AMD release that way.
Is it the best method you have? No.

OK, so question to you, there are a lot of postings "blaming" issues on stable rigs, after updating to this recent (April) 1803 update.
For sure, some of those cases are possibly blaming 1803 unfairly, but that isn't the question.
Have you tried that update?
What issues, if any, did you encounter?

I have a feeling this is going to be a mini-plague for the coming months, and while "don't update" is one obvious answer, I'm curious to know, specifically why that breaks, (assuming it does), and how to solve that, (other than fallback etc).
If the answer is fallback, is that the only answer, or have you recovered a rig in that situation?
Cheers

No worries, wow, that is a nasty situation with win updates. But before you roll back, it would be very helpful to know, if the DDU, clean install, or alternatively a manual driver install in devman as Ursul0 pointed out, gets you anywhere.

Because one would hope that DDU/clean install-driver option is an available solution.

But indeed, you could be right, I have seen some systems in a real mess after a OS roll back, and while that isn't to say that applies in your case, it's always one of those nasty things that sow the seeds of doubt later.

Also, you might like to set all your NICs to metered connection, that at least for the most part, prevents MS from updating your rig, or if nothing else, gives you some warning when they try to.

Regedit
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\NetworkList\DefaultMediaCost]

Grant your user "folder" permissions, and change all NICs to dword 2.
e.g.
"3G"=dword:00000002
"4G"=dword:00000002
"Default"=dword:00000002
"Ethernet"=dword:00000002
"WiFi"=dword:00000002

It won't help you with your current problem, but might be useful on your rigs
Good luck.

Like I said, end of discussion!
Nothing I said was WRONG, and throwing all your toys out of your pram does nothing to further your point(less) posting.
You might like to also look up the word "apology", and be wise enough  (one day), to know when to do that.

Pay better attention to detail, it's all written down here after all, no need to paraphrase or morph what other people write.
If you don't understand it, that's fine, all you need to do is copy/paste and ask for help.

If you're  going to counter point what others post, at least give them the courtesy to take your time to read it first, understand it, and MAKE your counter points, specific and to the point. But do NOT blanket, with wide reaching statements like "all your ideas in this area is total crap"

Equally, you can look up the word redundant for yourself.

And again, some understanding or manners and protocol would do you no harm either.

Bullying a technical argument does you no credit, neither is it necessary assuming you're intentions are to be helpful.
That does not mean you assassinate other people in the process.

I suggest you let this go, you're out of order, and that is my final word on the matter.
I'm not going to be drawn into your pissing contest, this is a technical forum, sigh.
End of discussion!

Yes, better to keep these,  technical discussions, rather than personalising them.
It also helps you make your point!

OK, but like I said, as indeed did the original poster, they are a small outfit, they have a single rig, and Acronis cost what?
Another fifty bucks, to do what, make a single clone, from a disc they don't have?

Again, for the last time, cloning in the context raised is not a good solution, in fact it's not even possible.
End of discussion if you please.

A bit of attention to detail would not go amiss, nowhere did I say the dagpatcher was outdated, YOU said that!
What I wrote was it was redundant, that's it.

Well ok Ursul0, you're entitled to your opinion and abilities, but regardless, your response is not helpful, constructive, or even civil.
Something for YOU to think about, look up "good manners" in the dictionary for a start!

Furthermore, you take this totally out of context, bringing in virtual machines! Well you SHOULD know full well, as a hardware abstraction layer solution, that in itself "solves" most of the driver related issues likely to be encountered when trying to boot a different chipset off the wrong clone os etc.

Why you seem to think that is useful to these guys, is anyone's guess, other than to bolster your opinions, it provides no constructive, useful solution to the problem, you just add noise.

Granted you do have some good tips there, (apart from that redundant patcher), so one has to wonder why you didn't offer these suggestions sooner, although as I say, they are mostly not applicable as was already stated, these guys HAVE NO OS to clone from!


NOTE: there is no place for a pissing contest here on this forum! Lets try to keep it civil.

Hahaha, MAN you're hardcore, that is quite an achievement, WOW.
Don't tell MS anyone, they'll be looking for new ways to screw us over, lol.

Side note: Do you limit the Claymore log file size? or overwrite? Or maybe log to a USB storage device?

Interesting "solution". I'd suggest this is NOT a good idea. While svdinu might have the ability to get this to work, I'd suggest it will, in the long run, create a mass of issues, and even if you get it to work, it will take a LOT longer than a OS reinstall of Win10. (especially if you have an SSD, reinstall of Win10home, takes about 15min at most).

On cloning: Actually, unless the motherboard and peripheral devices are identical, it's quite likely the system will fail to boot, or have a bunch of driver issues, which, in the first place, IF you have the ability to solve, (and they are even solvable at all), you would simply know better than to try this solution to start with.

No criticism of svdinu here, simply that when troubleshooting, you need to ensure you are working to eliminate issues, and not introducing any new ones along the way. This solution, introduces a huge array of issues, and potential issues, and potential issues mean doubt later on when you have other problems with the rig.

Start as you intend to go on! (known good etc)

I mention this here, as I say, not to be critical of svdinu, clearly he's trying to help, but when I read that someone formatted/reinstalled their OS, that tells me, they are possibly lacking the experience to know how to identify their issue, and have opted for reinstall, which is extreme, but which is a FAR better idea under these circumstances, (given their skill level)  than trying to clone a SSD, assuming you even have one to clone from.

From a personal POV, I HAVE cloned OS discs many times, would I try it in the context of this problem? NO.

In fact, I build identical rigs for people, and I STILL don't clone SSDs, (unless I'm installing bigger ones, or replacing them etc), but for a new build, I only install fresh, this ensures stability, and no cross pollination, (security for both the client, but also security of stability).

My 5 cents.
Happy Mining

OK, will it sounds like your PSU and topology is thought through, so maybe leave that for now.
Could you post the debug log, from the batch file I added above?

(Just to rule out the pool angles, as I have no experience with suprnova, and Ethermine, being pretty much the biggest there is, makes for a good choice when testing, even if you don't plan on mining with those guys).

Now that said, what SHOULD be happening at this stage is the init of OpenCL, (or CUDA if you're using Nvidia).
OpenCL is managed via the AMD driver, so assuming you don't have a problem with the pool, I think you need to investigate if OpenCL is available.

One way to do that, (with a only single GPU in the rig), is you use GPU-z, and see if the OpenCL check box is populated.

Now, I have also seen perfectly good working rigs, where that checkbox is NOT populated, and clearly that is wrong, so to be absolutely sure, you (again with only 1 GPU in the rig), could boot to safe mode, DDU, reboot, install the AMD driver, set to compute, reboot, and check GPU-z, I have never seen that fail to confirm OpenCL.

From there, try my batch file again, and post the log if you're still stuck.
Good luck

Also, VM? Set to 48000GB if need be.

Foot note, in case anyone needs convincing of the single GPU method. If you do a clean AMD driver install with 13GPU, this alone can take hours, more so, if it fails, and AMD themselves STILL have an open bug on init fail with more than 12GPU.

Often when testing theories, working through troubleshooting etc, this is time consuming, and for sure you probably work at testing, (proving) some angles that were perfectly fine to start with. Time is not wasted of course, as you can cross those off the list, elimination of them, takes you closer to the answer.
But all the same, if you have multiple GPU, that's an exponential multiplier on the possibilities of fault, and means more time to troubleshoot is a given.
Breaking it down to basics, (you only need to pull the USB cables from the MB, for the most part), is pretty quick, you don't need to physically remove the GPUs etc, means you can run DDU, and reinstall the AMD driver with the minimum time consumed, and get back to starting Claymore as quickly as possible.

AMD driver install time, is pretty acceptable with a single GPU, but 12 and 13 is really nasty, and I have seen that fail many times, so working on proof of concept with a single GPU is by far the fastest, and most likely to provide solutions imho.

BTW 18.3.4 is a stable driver for mining.
Note: there is NOTHING mentioned in any release notes from AMD, on any mining related improvements in the drivers release since 18.3.4.
That is not to say the newer drivers are bad, (some are though), but simply, 18.3.4 has been out for a while, rolls up all the mining related things you probably need for RX570/580 hardware, and has proved itself far more than any of the drivers released since then. So, if you're going to DDU etc, I'd suggest 18.3.4 is a good choice, known good, known stable, well proven choice, and when you're troubleshooting, you need to focus on elimination of doubt. Go with what you know, (is good).

Hibernation.

Win7, Win10.

It's quite common for mining rigs to have "small" SSD, just enough for the OS and Claymore, so probably doable on a 64GB SSD even.

However, with swap file, page file, and setting VM to 16GB or more, (16GB is the minimum), it's easy to see some big chunks of disc space be consumed.

Worse still, if you have 16GB of RAM, with a default Win install, you're going to see some 12~16GB stolen by hibernation.

Considering one probably does not want a mining rig going to sleep or hibernating, tip of today would be to disable those, and then review your power settings, as those too can be hijacked by various apps etc,.

Open an admin command prompt

powercfg /h off


That's it, eliminates a potential cause of mining halt, and recovers a nice chunk of RAM you can then give to Claymore as VM if you need to.

Happy Mining

That sounds frustrating.
What do you have Virtual memory set to?
With 4GPU, 16GB should be fine, but if you want to absolutely exclude this possibility, set VM to 48000, (I have some 13 GPU rigs running that, and it's fine).

Also, side note, actually, I'll do a new post on this bit, hibernation!

If you didn't try it already, turn on debug and see if Claymore log adds some clues.

And if you really hit the wall, remove all the GPU but 1, boot to safe mode, run DDU, reboot, and install AMD driver with that single GPU, set to compute, reboot, and see if that works. Then add back the other GPU, build up to that etc, but get a proof of concept first.

Don't forget to switch off debug when things are running.

Good luck

Foot note: As always, if you get stuck, post your complete batch file content.
There are a lot of possible causes to this, so there is no one answer, but another common error is to command a clock the hardware can't deliver on. Claymore is waiting for some ack, which depending on what you sent to the hw, the driver, and probably the hw ability to cope with the impossible, might not send an ack, and Claymore can only wait, (and have the smarts to not wait forever, but trip a reset after 5 min).

But going back to basics, break it down, is a good method for debugging. Yes, it might be painful, but if done right and systematically, DOES deliver a good probability of finding a solution. Far more so than blindly stumbling around in the dark, hoping to get lucky, hmmm that reminds me of something else, hahaha.

Good morning jeremyismer

Thanks for adding additional detail.

Yeah, that's a tough one, I don't want to speculate on your specific hardware, as I have no experience with those MBs, but...

I did exactly the same thing as you initially, in that I had Dell Optiplex 7010 kicking about, and used that as my prototype rig. As it only had 3 PCIe slots, I used a PCIe 4 port expansion card, but that rig never had more than 5 GPU running off it, BUT it was REALLY unstable, I had a lot of issues every time I added a card, or changed anything, or even powered it down.

But I learned the importance of stability and uptime, and when building my first rig from scratch, opted for the ASUS Prime 270-a, and MAN was that like a breath of fresh air, it just worked... Until I added the 5th GPU, then it would not boot, or if it did, I would have only 4, or sometimes 3 or 2 GPU detected.

Now on that board, you need to do some tweaks in the BIOS to get beyond the 4 GPU, various youtube videos are out there about that, (guess what, they vary in detail, sigh), but essentially, the only critical change you need to make is, (on the Prime 270-a), called enable above 4G decoding.

The definition of “Above 4G decoding” is to allow the user to enable or disable memory mapped I/O for a 64-bit PCIe device to 4GB or greater address space, because the primary VGA card should always be mapped below 4GB address."

OK, so after that, I used the ASUS Prime 370-A, (same situation as above, (above 4G decoding), gets you up to 13GPU, and on the ASUS B250ME, specific mining mb, you have that feature enabled by default.

Now, as it happened, a lot of back ordered GPU started flooding in, after months and months waiting, and I was able to finish the rigs I built for other people, and even then had some spares kicking about, and again found myself with a spare Dell Precision 3650, so decided if I could use that to employ 5 "spare" GPU.

This was a tough one, 2 PCIe slots, so I used the 4port card in the 16x slot, and a single PCIe in the 1x slot, and no matter what combination, cards, ports, even tried some older AMD and Nvidia cards, but it would not boot with more than 4 GPU.

I played about with that for some weeks, building up to the conclusions, it lacked the ability to cope, whether by hardware design, or BIOS limitations, (something Dell are renowned for, their goal, (understandably) is mass market, stable, reliable, nothing special or pushed too hard, means lower support costs for them, and steady revenue streams for replacements after 3 years warranty expires).
At the same time, having 5 GPU kicking about and no way to use them, I forked out for another Prime 270-A, which I chose over the 370-A, simply because I wanted to pull the i7 6700 CPU and 16MB 2400DDR RAM from the Dell precision, so in my case, that additional rig, was only really costing me the new MB, and time, as always ;-)

Interestingly, I noticed 2 other things,

1. Mining with a HDD sucks, staggering how slow that was to do everything, boot, init, you name it.
2. Mining with an i7, even an older mid range 6700, was a lot faster to init Claymore, than all my other rigs, i3 7100 or i3 8100. Probably that is obvious, but all the same it just goes to illustrate how much code is executing in Claymore during the init, so that was the real surprise I mean here.


So... I'm sorry man, I don't have any real solutions for you, I would have liked to get some definitive answers on those Dells, but time is always divided between things, and looking at the money draining away from not using those 5 GPU, that was the kick in the arse to build another rig.

BTW: One mistake I made early on, was to opt for a celeron CPU, and while that worked ok, it means you have to use one of your mining GPU to drive a monitor, and that in itself brings in a whole raft of issues, mostly when you're debugging driver problems, init issues etc. Now, that was my first from-scratch rig, and I didn't fully appreciate what the loss of intel graphics would mean, so I swapped that out for a i3 7100, and on all rigs since, having intel graphics is a check-box item for me since. It means, no matter what crap MS or AMD or drivers or PSUs through at you, at least you can get a display on it.
The i3 uses about 1W more than the celeron, (on paper), in reality, for mining, once everything is running, I don't see any difference in power consumption from the celeron.

So, in conclusion, you could be flogging a dead horse there by adding the 5th GPU, and weighing up the cost of 4 GPU doing nothing while you work on that, possibly unsolvable issue, is a situation probably already very much playing on your mind.

So... time to go shopping maybe? MB, CPU, RAM? Or by a 2nd hand bare-bones maybe.

Good luck man, lmk how you get on.

Hey there,
So you have debug on, what do your logs tell you?

Try adding

-di 0

Start with GPU0 only, would be a good thing to try, as you could be overloading your PSUs.
If it's a new rig, never been running/stable, break it down, get through the init on one GPU, prove your config and basic hw/sw enviro, and take it from there.

jooi, what PSU are you using, and how many GPU on each, how many risers per cable?

Kudos on the 13gpu rig man, not so easy to get those stable.

Good luck.


footnote:
Try this.

setx GPU_FORCE_64BIT_PTR 1
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_SINGLE_ALLOC_PERCENT 100
EthDcrMiner64.exe -epool ssl://eu1.ethermine.org:5555 -ewal wallet.rig -epsw x -di 0 -gser 1 -esm 0 -etha 0 -ethi 16 -eres 2 -erate 1 -estale 1 -asm 1 -platform 1 -y 1 -dcri 9 -wd 1 -ftime 5 -r 1440 -cclock 1200 -cvddc 900 -mclock 2100 -mvddc 850 -tstop 83 -tstart 50 -tt 60 -fanmin 40 -fanmax 100 -ttdcr 80 -ttli 80 -mode 1 -dbg 1 -altnum 3 -mport -3333 -mpsw whatever -logfile logs\

Hey jeremyismer
TBH, I would not waste time on MB, system ram, or even PSU. (in terms if variables, there is nothing to tune, and little to gain, (assuming you have a stable rig right now)
The key point with PSU is, they are a big portion of your asset cost, so ensuring they run at optimal efficiency vs number of GPU is a quick and easy calculation, that's it, there is nothing else to tune there, no more variables to consider, (other than maintenance and mtbf)

The BIGGEST variable, (loss of income) that I see, is simply this windows update, driver updates, or people's actions, resulting in  taking rigs off line.
If you let MS do that, and even if you don't have followup issues, costing more down time, or a nightmare trying to work out what changed, broke etc, but it all adds up to down time.

Latest windows updates, sigh, here we go again, some people are seeing rigs down for  hours, only to then limp along for days afterwards, many poor people are still trying to debug their rigs update issues.

Even 3 hours of down time, that is time you NEVER get back. (mining time = money)

Tuning for hash rate increases, now THAT is worth it, even worth the down time, because those small gains will accumulate over time, offset the losses, and increases profit accumulation. (assumes you're mining operations are longer term, (some years).
(more so, if you strategically ensured you have the same make model GPUs, as you can rollout your gains to each one with minimal effort and downtime)

So, imho, the ONLY thing you should focus on is maximising hash rate, to find the optimum power consumption efficiency.

Anything else, anything costing you down time, is time your investment is making no money, in fact it's costing you money to do nothing.

This is more an economics situation than a tech one, assuming you want to earn from it.

Anyway, hope it goes well mate, offer to look at your BIOS remains open.
Cheers

No worries, but to be clear, Claymore has no dependency on sharing. (Equally, sharing is a weakness in your system security).
This sounds more like a windows folder permissions issue.
Assuming YOU don't have the need to share that folder, (and knowing Claymore does not), you might like to check the next tab over to the right, (from sharing).
Right click the folder > Properties, (security tab), and grant your (normal) windows user account, (read & execute), or even full control.
At least that closes a potential security weakness, (sharing), and also should solve this appropriately.
Happy mining

Claymore 11.7 seems to be a tad more sensitive to overclocks.

To be clear, claymore does no such thing!
All claymore is doing is "sending" the clock value YOU specify, that is all, nothing else.

If it fails, your clock setting is wrong, or you're working too close to the limit of your hardware.

What has changed/improved over time  in Claymore is ability to leverage more from the hardware, so if you're cutting it too fine, or dual mining, you might tip it over the edge.

But regardless, claymore has no such "sensitivity", it simply does what you ask of it, and you need to ensure that is rational for your rig/hadware.

Hey there, sorry, I'm not suggesting you use the console every time, just one time, so you can read the output.

BUT: If you are saying, for example, your desktop shortcut to your batchfile executes and closes, yet opening a console and executing the batchfile works, claymore starts Mining etc, then that is probably a windows permissions or security settings, thing, or possibly your antivirus related problem.

But essentially, if the console execution is working ok, that proves 99% is ok, os, drivers, batchfile content, claymore, etc etc.
So if you have a shortcut init fail, you can shift focus away from the above.
Cheers

Updates, sigh.
Well, OK...
You have 30+ miners? WOW, so this should be well within your means to solve.

First: Work out if you have a hardware problem. (Take one/some of those problem GPU to a good rig and test)
Given you have 30+, I assume you long ago recognised the absolute necessity for a dedicated development/testing rig, so use that one, which of course you would NEVER update, if you expect it to act as your benchmark. ;-)

Check my post on power supply cables, I mention this, as it's easy to waste time thinking only about the stuff you most recently changed, and back track from there, but if the rig was off for a while, during updates, and subsequent issues, it's not impossible you have other problems, hardware related set of some time bombs etc.

I'd suggest not wasting time, and MAN that wastes time, on anyone using the Aug 2017 blockchain drivers, certainly not for purposes of troubleshooting, they are beta after all, and specifically tagged as not stable/released.

But to be clear, make sure you eliminate the hardware before expending time on the software.
(You could also pull all GPU from that rig, drop in a known good GPU, riser, cable etc), and see what that reveals.

All the more easy when you have as much hardware as you have.

If you're 100% certain the hw is ok, set log to debug, see what is going on before/after the hang event
Hope you get some sleep too.
Good luck man.

I'd suggest taking a look back through your logs, and assuming you changed nothing your end, if this worked before, it might be explained by the pool being unavailable, or unreachable.
Which of course is exactly what that is saying, so the question is why?
Maybe try pinging that address, or pasting the relevant sections into your browser, check if it's up etc.
Good luck man.