Please help - two motherboards bricked in 2 weeks

[4en1vanm]

So let me start my tale of frustration and self doubt. As a longtime cryptocurrency user/investor I got the bright idea to try building a small mining rig since I have experience building gaming pc's. This was not a get rich quick idea but more a hobby that could pay for itself. So I did the research and decided to got with a MSI Pro Series SLI Plus Z270 ATX motherboard, Intel Celeron G3920 CPU, compatible 8gb's of ram, Corsair HX1000 Watt Platinum PSU, and Onvian Mining PCI-E risers. I also started with two RX 580's 8gb (Sapphire Nitro+ SE & Sapphire Pulse).

So here we go.

About two weeks ago I get my new motherboard and start building my mining rig. I get everything installed correctly (I think) and get ready to power it on. I click the power button and I get about a second of power - GPU's briefly buzz - and then nothing. I look around the board and realize that I forgot to plug in the CPU power connector from the PSU into the motherboard. (Complete noob mistake that I shouldn't have made but I din't think it could fry the board) So I plug the CPU connector in and hit the power.... Nothing... no indicator lights... never even posted. So I start trying to troubleshoot the problem - nothing worked. I reset the CMOS, unplugged everything but essential equipment, did the paper clip test on the PSU (It worked fine). I finally contacted the retailer and requested a replacement.

A few days later my replacement arrives - I surgically install everything and triple check every connector. I hit power and boom - everything is booting up!!! I almost start crying tears of joy as my rig starts humming along.

About two weeks go by and everything is working like a charm. I've modded the vbios on the GPU's to squeak out some extra MH/s and decide to order a third GPU.

Yesterday my third GPU (MSI RX 580 Gaming X 8gb) arrives and I plug it into my third PCI-E riser and start connecting all the power connectors - since I ran out of 6 + 8 connectors I used the 6 pin to SATA connectors that came with my risers and plugged two of my risers to the PSU (single cable for each one). I then used a 6 + 8 pin on the new GPU. After checking all the connections I hit the power button - and I once again get about a second of power - a click - then everything is dead.

My question to the community is what am I doing wrong or am I just really unlucky? I've built several gaming PC's that have never once had a problem. Are the cheaper motherboards really this fragile? Could the risers be causing a short on my motherboard? Does anyone else have any theories on what I'm doing wrong? I'm desperate for help at this point...

 

[Darkbreeze]

Off the top of my head, not factoring the initial board dying which could simply be bad luck, IDK, not plugging in the 8 pin CPU power connector shouldn't damage anything, it just usually won't power on, I'd say that for almost certain you do not have a capable enough power supply for 3 xRX580.

RX 580 has ~360w power draw, ACTUAL, from the wall ( https://segmentnext.com/2017/04/20/amd-rx-580-power-consumption-500w/) under max conditions, so for three cards and not even counting the demands of the rest of the system you are looking at potentially 1080w just for the cards.

That also does not factor in any changes to thermal design power created by your custom firmware installation tweaks, power spikes that rise beyond design power, the demands of the cpu, motherboard, memory, fans (If any), storage drives, cpu cooler or the fact that power supplies generally list their PEAK capacity specifications, not the amount that should be used continuously. This is much like a speaker's RMS values. Yes, it CAN provide that level of capacity, if it's a good unit, like yours, but that is not the level of demand it SHOULD be seeing on a full time basis.

Power supplies should be used which land the actual value needed on a regular basis somewhere in the 60-80% of total capacity capability percentile. You are FAR past that.

Also, depending on WHAT was included in your GPU bios tweak, it wouldn't be unreasonable to think that perhaps there's something in there that might be a factor in overdrawing the total MB PCIe power specification. That might be a stretch, but I've seen less proven firmware cause any number of issues in the past. Usually, unless the manufacturer releases a firmware for the purpose of CORRECTING a known issue, it's best to leave the firmware as is.

Lastly, you don't generally want to choose a cheap ass motherboard for mining. Anytime you intend to run more than one GPU card, whether gaming or mining, you really do want to start factoring in quality over features because obviously there is going to be significantly more electrical stress than with any single card. I'd ask for a refund and buy a higher end board, and either get an additional PSU to run the non-card portion of the system off separately or a bigger power supply that can handle what you are asking it to, on a regular basis.

 

[4en1vanm]

Off the top of my head, not factoring the initial board dying which could simply be bad luck, IDK, not plugging in the 8 pin CPU power connector shouldn't damage anything, it just usually won't power on, I'd say that for almost certain you do not have a capable enough power supply for 3 xRX580.

RX 580 has ~360w power draw, ACTUAL, from the wall ( https://segmentnext.com/2017/04/20/amd-rx-580-power-consumption-500w/) under max conditions, so for three cards and not even counting the demands of the rest of the system you are looking at potentially 1080w just for the cards.

That also does not factor in any changes to thermal design power created by your custom firmware installation tweaks, power spikes that rise beyond design power, the demands of the cpu, motherboard, memory, fans (If any), storage drives, cpu cooler or the fact that power supplies generally list their PEAK capacity specifications, not the amount that should be used continuously. This is much like a speaker's RMS values. Yes, it CAN provide that level of capacity, if it's a good unit, like yours, but that is not the level of demand it SHOULD be seeing on a full time basis.

Power supplies should be used which land the actual value needed on a regular basis somewhere in the 60-80% of total capacity capability percentile. You are FAR past that.

Also, depending on WHAT was included in your GPU bios tweak, it wouldn't be unreasonable to think that perhaps there's something in there that might be a factor in overdrawing the total MB PCIe power specification. That might be a stretch, but I've seen less proven firmware cause any number of issues in the past. Usually, unless the manufacturer releases a firmware for the purpose of CORRECTING a known issue, it's best to leave the firmware as is.

Lastly, you don't generally want to choose a cheap ass motherboard for mining. Anytime you intend to run more than one GPU card, whether gaming or mining, you really do want to start factoring in quality over features because obviously there is going to be significantly more electrical stress than with any single card. I'd ask for a refund and buy a higher end board, and either get an additional PSU to run the non-card portion of the system off separately or a bigger power supply that can handle what you are asking it to, on a regular basis.

Thanks - I miscalculated the total energy draw needed from the RX580's on bootup. I planned for >225 watts per card based on some research I did - which in hindsight was too much for a 1000 Watt psu. Would drawing 1080+ watts put enough stress on a cheap board to fry it - I thought these PSU's had built in protections to prevent this? Can I rule out the pci-e risers causing the short? As for the modded GPU's - I did run them modded for about 24 hours and monitored their energy draw from GPU-z and the wall. On GPU-z I was getting about 150 watts each and the total rig was pulling about 480 watts from the wall while mining (For 2 gpus).

Based on your experience are cheap motherboard failures common with multi-gpu systems? I've always built gaming PC's using high end boards and have never opted for cheap boards...

Based on what you said I'll probably pick up another 1000/1200 watt PSU. I've already submitted the MSI board for a refund and purchased a Gigabyte Ultra gaming board.

Do you have any other advice??? Having 2k + of hardware failing over and over again has been pretty stressful.

 

[Darkbreeze]

The PSU didn't fail though, at least, that hasn't been determined yet. It MIGHT however have produced so much ripple at that peak expenditure, over a matter of days, that it boiled the motherboard caps into oblivion. While that TOO is not confirmed, it's certainly possible. Most units you throw a lot of the ripple, noise and voltage regulation test results right out the window if you are at or near 100% capacity for long periods of time.

Again, I cannot SAY that is the cause, only that it is A possibility.

Without getting specifically in DEPTH on the subject, and you can google for further relevant information, this topic is loosely touched up at the following link. There are tons of threads on this to be found though.

https://www.badcaps.net/forum/showthread.php?t=4306


Based on my experience, MOST cheap motherboards tend to NOT support multiple GPU card configurations, and those that do are usually limited to two card SLI and Crossfire configurations. Even those that technically support three or more cards, are not well recommended unless it's a high end board INTENDED for three or more cards, and mostly, those have disappeared with the introduction of the 1000 series graphics cards because the in game profiles for it are weak and mostly they are not needed since the capabilities of single cards have advanced so much that even for 4k there is not an overwhelming need for multiple cards anymore. Some yes, but not like there used to be.

Still, if a board SAYS it supports multiple cards, whether in an SLI/Crossfire configuration or not, it SHOULD be capable of it. Unfortunately, that's not ALWAYS the case. There are a LOT of things that cheap boards SAY they support or can do, and the reality is that much like anything, you get what you pay for.


With a dual PSU configuration you may, likely, need to do some research on USING them together. The power_ok signal and other specifications built into the design of modern power supplies and motherboards is such that simply connecting one of the power supplies to the graphics cards, without it being connected via the 24 pin ATX connector to the motherboard, will likely NOT turn on without additional modification or hardware in order to allow it to turn both power supplies on. You can find information on this here:

Google results. Two power supply mining configurations.


Aside from ALL of that, there could clearly be other/different factors involved as well. Might want to pull the CPU and make damn sure there are no bent motherboard or CPU pins/contacts. Make certain that the memory is 100% fully seated, not partially seated but LOOKING like it's fully seated. This happens, and can cause shorts/disconnects. Generally that would not fry something, but would create errors, but who knows. Double check every single connection on the system to be sure everything is fully seated.

Probably also a good idea to make sure that the 8 pin EPS power connector to the motherboard is ORIENTED correctly, with the keyway lined up to match. While it IS rare, I HAVE seen people mistakenly plug them in backwards, somehow (Still not sure how much force it takes to do THAT) and eventually or instantly fry a motherboard and/or CPU. Pretty doubtful, but so is two sequential motherboard failures.

 

[4en1vanm]

The PSU didn't fail though, at least, that hasn't been determined yet. It MIGHT however have produced so much ripple at that peak expenditure, over a matter of days, that it boiled the motherboard caps into oblivion. While that TOO is not confirmed, it's certainly possible. Most units you throw a lot of the ripple, noise and voltage regulation test results right out the window if you are at or near 100% capacity for long periods of time.

Again, I cannot SAY that is the cause, only that it is A possibility.

Without getting specifically in DEPTH on the subject, and you can google for further relevant information, this topic is loosely touched up at the following link. There are tons of threads on this to be found though.

https://www.badcaps.net/forum/showthread.php?t=4306


Based on my experience, MOST cheap motherboards tend to NOT support multiple GPU card configurations, and those that do are usually limited to two card SLI and Crossfire configurations. Even those that technically support three or more cards, are not well recommended unless it's a high end board INTENDED for three or more cards, and mostly, those have disappeared with the introduction of the 1000 series graphics cards because the in game profiles for it are weak and mostly they are not needed since the capabilities of single cards have advanced so much that even for 4k there is not an overwhelming need for multiple cards anymore. Some yes, but not like there used to be.

Still, if a board SAYS it supports multiple cards, whether in an SLI/Crossfire configuration or not, it SHOULD be capable of it. Unfortunately, that's not ALWAYS the case. There are a LOT of things that cheap boards SAY they support or can do, and the reality is that much like anything, you get what you pay for.


With a dual PSU configuration you may, likely, need to do some research on USING them together. The power_ok signal and other specifications built into the design of modern power supplies and motherboards is such that simply connecting one of the power supplies to the graphics cards, without it being connected via the 24 pin ATX connector to the motherboard, will likely NOT turn on without additional modification or hardware in order to allow it to turn both power supplies on. You can find information on this here:

Google results. Two power supply mining configurations.


Aside from ALL of that, there could clearly be other/different factors involved as well. Might want to pull the CPU and make damn sure there are no bent motherboard or CPU pins/contacts. Make certain that the memory is 100% fully seated, not partially seated but LOOKING like it's fully seated. This happens, and can cause shorts/disconnects. Generally that would not fry something, but would create errors, but who knows. Double check every single connection on the system to be sure everything is fully seated.

Probably also a good idea to make sure that the 8 pin EPS power connector to the motherboard is ORIENTED correctly, with the keyway lined up to match. While it IS rare, I HAVE seen people mistakenly plug them in backwards, somehow (Still not sure how much force it takes to do THAT) and eventually or instantly fry a motherboard and/or CPU. Pretty doubtful, but so is two sequential motherboard failures.

Thanks - one final question. The motherboard I bought to replace my MSI board has 3 pci-e x16 slots (16,8,4) and three PCI-e 1x slots conforming to PCI E 3.0 standards. How many GPU's could I safely run on this board assuming I have the correct power supplies? It does support 2 way and 3 way AMD cross fire & Quad GPU CrossfireX. The online documentation from people on this has been very spotty - most seem to not know what they're talking about. The typical comment I see is it can support 6 with pci-e risers. Do you have an opinion on this?

 

[Darkbreeze]

SLI and Crossfire support are unimportant, to the extent that you are not TRYING to run those types of configurations. Totally different methodology involved. What it DOES affect though, is having a board that doesn't crumple under the power demands of multiple cards. Generally, even boards that don't specifically support multi card SLI/Crossfire configurations can support multiple individual cards, but since few people use those kinds of configurations, they rarely get tested OR often they are very low demand, slot power only models that don't push things. Still, since each card is only drawing 75w per slot, it's rarely a problem unless you've maxed out the PCIe circuit to some degree, electrically.

I'm actually NOT the best person to ask about mining specific configurations, as I haven not done any mining myself, so I probably lack some insights when it comes to this area anyhow. Still, even with only a single card, cheap boards are still cheap boards, and have a higher failure rate than more expensive boards, especially if they are pushed to their limits like I would assume a mining configuration would do.


Honestly I can't tell you where to go to find people who are knowledgeable in both Bitcoin practices AND computer hardware, because from what I've seen so far you are looking at mostly a lot of low hardware knowledge but high BC knowledge miners for the most part. I imagine a bit more looking around in the bitcoin forums might land you eventually on a page where somebody with a high degree of experience and knowledge in BOTH aspects might be found offering some advice, but it's likely going to take a bit of digging.

 

[4en1vanm]

SLI and Crossfire support are unimportant, to the extent that you are not TRYING to run those types of configurations. Totally different methodology involved. What it DOES affect though, is having a board that doesn't crumple under the power demands of multiple cards. Generally, even boards that don't specifically support multi card SLI/Crossfire configurations can support multiple individual cards, but since few people use those kinds of configurations, they rarely get tested OR often they are very low demand, slot power only models that don't push things. Still, since each card is only drawing 75w per slot, it's rarely a problem unless you've maxed out the PCIe circuit to some degree, electrically.

I'm actually NOT the best person to ask about mining specific configurations, as I haven not done any mining myself, so I probably lack some insights when it comes to this area anyhow. Still, even with only a single card, cheap boards are still cheap boards, and have a higher failure rate than more expensive boards, especially if they are pushed to their limits like I would assume a mining configuration would do.


Honestly I can't tell you where to go to find people who are knowledgeable in both Bitcoin practices AND computer hardware, because from what I've seen so far you are looking at mostly a lot of low hardware knowledge but high BC knowledge miners for the most part. I imagine a bit more looking around in the bitcoin forums might land you eventually on a page where somebody with a high degree of experience and knowledge in BOTH aspects might be found offering some advice, but it's likely going to take a bit of digging.

Thanks for the advice - just an update - received my new Gigabyte motherboard and plugged in 2 of my 3 gpus and it's been running good for 2 days now. I could instantly tell the new board was higher quality/better design based on my visual inspection I have a 1200 Watt corsair PSU getting delivered tomorrow. I'm going to swap out the 1000 w for the 1200 w and see if I can get the third GPU running. I'll update with my progress.

 

[Darkbreeze]

Sounds good. Fingers crossed.