3 motherboards fried trying to get a 7th GPU work

[pkefal]

Hi all,

I have some Sapphire RX580 Pulse 8GB cards with a custom bios. PSU is HXi 1200 from Corsair. 6 of them are connected through 006C risers and the risers get their power directly on their 6-pins using Corsair's cables and some good quality splitters I got. It's basically one0 6 pin from the PSU cable, to one 6 pin to 2 - 6 pin splitter that only power the risers. The GPUs all have their own power cables.

I tried adding the 7th GPU, which had it's own cable for the GPU and from the same cable, it also powered the riser (the PSU cable had two 6 pins). Trying to start it, it fried the mobo (MSI Z170A Gaming M5). I thought there was some kind of bad luck, so, I used some of the mobos I had for the next rigs, I changed the PSU cable and the riser, but trying the same thing on two more Gaming M5s, resulted in the same thing. They get fried. Along with them, it killed the DRAM, the disk and also the CPU. The CPU wasn't completely fried but it wasn't functioning properly so I had to write that off.

Does anyone have any clue what might be happening? I'm afraid to try again as I fire I'm going to fry more stuff. Any ideas are welcome.

[bathrobehero]

I don't get the 6-pin splitted to feed risers.

The 6-pin or 8-pin cables that you plug into the GPUs only carry 12V but the riser also needs some of the 5V feed included in SATA/molex.
Edit: nevermind I see there are risers that only take 6-pins and stepping the 12V down.

Whatever the case might be, I think either the PSU is faulty or something is very wrong since I never heard people also frying stuff like CPU/RAM/disks with the motherboard.

I'd be curious to see some pictures.

[ben8jam]

I read this just as I'm about to turn on my MSI Z170a M5 for the first time.......

[Maxumark]

Did the system work with the 6 GPUs prior to adding the 7th?

Is there any chance that ANY of the cables connected to the PSU are not Corsair PSU cables even if they look the same?

Use a multi meter and check the output at the cable end and see everyone has the voltage and polarity that it should.


 

[philipma1957]

You attached the risers wrong

The data cables or USB cables were in the 1x pcie slots backwards

Been there done that.

You should read my thread here I have great 3 card riser free board builds

[bathrobehero]

You attached the risers wrong

The data cables or USB cables were in the 1x pcie slots backwards

Been there done that.

You should read my thread here I have great 3 card riser free board builds

How is that possible to plug them in backwards?

[VoskCoin]

How did you determine the mobo was fried?

[Mattthev]

How did you determine the mobo was fried?

Pretty bad smell and you can see burnt place on mobo.

I would say bad cables or wrong connection...
But only time I saw burn everything was like 8 years ago when my PSU failed only RAM, HDD and fans survived. Since then only parts on mobo burnt they use better protection now. My PC what I'm using when I'm not on laptop has burnt 4pin PWM controller and it was my fault I touched GPU power with metal frame, but GPU has no problem at all
So my toughts are the mobo... also do you use multi or single rail? I was starting thread about this few days ago I have Corsair HX 1200.

[bathrobehero]

How did you determine the mobo was fried?

Pretty bad smell and you can see burnt place on mobo.

I would say bad cables or wrong connection...
But only time I saw burn everything was like 8 years ago when my PSU failed only RAM, HDD and fans survived. Since then only parts on mobo burnt they use better protection now. My PC what I'm using when I'm not on laptop has burnt 4pin PWM controller and it was my fault I touched GPU power with metal frame, but GPU has no problem at all
So my toughts are the mobo... also do you use multi or single rail? I was starting thread about this few days ago I have Corsair HX 1200.

In case of a single rail setup if the PSU can't deliver as much as you're trying to pull it will just shut down and not burn anything so it has to be something else.

[Mattthev]

In case of a single rail setup if the PSU can't deliver as much as you're trying to pull it will just shut down and not burn anything so it has to be something else.

Yeah, but Corsair HX series can be switched to both and it comes set to Multiple.

[pkefal]

A lot of replies, thank you guys.

Yeah, but Corsair HX series can be switched to both and it comes set to Multiple.

So my toughts are the mobo...

I haven't changed anything to the setup on the PSU, it's as it came out of the box so I'm not sure if it's multi or single. The HXi required the Link connection enabled/installed to allow you to change this and I haven't done anything for it, so it's not easy to try. Also, you mention you think it's the mobo. I've seen and read here of setups on the Gaming M5 with 7 GPUs, how can it be the mobo? And let's say the first mobo was the case.. 2 more? So 3 in total faulty? Lastly, why would single vs multi rail make such a dramatic difference and cause things to fry?

Did the system work with the 6 GPUs prior to adding the 7th?

Is there any chance that ANY of the cables connected to the PSU are not Corsair PSU cables even if they look the same?

Use a multi meter and check the output at the cable end and see everyone has the voltage and polarity that it should.

Yes, it did and it currently does with 6 GPUs. No, all the cables are Corsair cables, they were part of the Corsair package plus the extra ones I bought were from Amazon and they are from Corsair as well (I bought two extra kits to have spare cables to use), I mean they came in a corsair box and bag, I don't think they are counterfeit. I will use a multi meter to measure the output, that was my idea after all this shitfest as well, but I was trying to poke people's brains in case they had a similar experience or any other ideas.

At this point, I have an AXi 1200 and 7 other GPUs waiting to be assembled into another rig, but I'm afraid to try, before I understand what happened. Even if I was trying to draw more power than what the PSU can handle, it should just have shut it down, not fry everything. On the other had, looking at how much power 6 cards are drawing now at the wall, even under-volted, it's about 1080 to 1120.. so I have doubts if it can handle the 7th card now, I was expecting the system to be around 920-950 with 6 cards when I did my calculations. I might have to get a secondary PSU and link them together.

[pkefal]

I don't get the 6-pin splitted to feed risers.

The 6-pin or 8-pin cables that you plug into the GPUs only carry 12V but the riser also needs some of the 5V feed included in SATA/molex.
Edit: nevermind I see there are risers that only take 6-pins and stepping the 12V down.

Whatever the case might be, I think either the PSU is faulty or something is very wrong since I never heard people also frying stuff like CPU/RAM/disks with the motherboard.

I'd be curious to see some pictures.

Yeah, I was surprised by this as well. I've been building PCs for a long time, for an internet cafe chain back in Europe (about 5-6 per day + fixing and diagnosing on some more), I never had such a case. I was annoyed by the whole thing.

I do suspect the PSU as well, or at least the specific output port of the PSU where that cable plugs in but at the same time I have trust in Corsair not to screw this up or at least catch it in their QA.

Is there anything specific you want to check at the pictures or just the whole setup? Currently is with the 6 GPU.

[HashZilla]

Did you use the same risers in same positions?  Possibly bad riser like this....

https://bitcointalk.org/index.php?topic=1649106.0

[philipma1957]

You attached the risers wrong

The data cables or USB cables were in the 1x pcie slots backwards

Been there done that.

You should read my thread here I have great 3 card riser free board builds

How is that possible to plug them in backwards?

I will photo.

see the standard 1x slots  see the small slot with a little wall


https://i.imgur.com/HlJ3Zuw.jpg

only allows for this


but people using the 16x full slot
can do this which is fine


but they can do this. This is a short  which can kill the card the riser the slot the mobo


 

[pkefal]

Did you use the same risers in same positions?  Possibly bad riser like this....

https://bitcointalk.org/index.php?topic=1649106.0

Same risers, same positions for all of the 6th. But I did change the riser on the 7th once the first one was fried. Also, how can a riser, cause such havoc?

[philipma1957]

Did you use the same risers in same positions?  Possibly bad riser like this....

https://bitcointalk.org/index.php?topic=1649106.0

Same risers, same positions for all of the 6th. But I did change the riser on the 7th once the first one was fried. Also, how can a riser, cause such havoc?

it is a dead short   trust me.

   I killed the slot the gpu and the psu when I did it.  My guess is  I had a 750 watt psu and you have a beast to do 7 gpus so you feed it more juice and killed more gear.

I do smaller rigs and rarely bother with  a lot of risers.  Shorts are a big reason and my double cataract repaired eye are okay but not great for depth perception and close up stuff.

[adaseb]

He said he changed the riser the 2nd time he tried it. What are the chances that 2 risers sequentially were both internally shorted.

Really hard to diagnose without actually being there and seeing what the issue can be.

I know back in the litecoin days people destroyed motherboards due to uncut ribbon risers however with USB risers that's not an issue these days.

[pkefal]

He said he changed the riser the 2nd time he tried it. What are the chances that 2 risers sequentially were both internally shorted.

Really hard to diagnose without actually being there and seeing what the issue can be.

I know back in the litecoin days people destroyed motherboards due to uncut ribbon risers however with USB risers that's not an issue these days.

Thank you for the comment. Yeah, I understand, it's quite difficult, but at the same time I was almost out of ideas and things to try and 3 mobos, although I can RMA them, it's not a "cheap" thing to try.. I wonder if there is an easier way to test those without risking any gear.. Plug them in and start the PSU with the paper clip method to see what kind of power goes to the riser(s)?

Another thing is this: Currently, the system draws 1080 to 1100 Watts.. with the 7th card, this would have been over the 1200 Watts the PSU supports. However, in that case, it would have just shut down, not fry stuff. Is that assumption wrong? I don't think so, but I'm asking just in case, plus the PSU has short and OCP protection, although the short protection does kick-in in specific conditions so I don't know if a shorted riser will do the trick and trip the protection.

[pkefal]

Did you use the same risers in same positions?  Possibly bad riser like this....

https://bitcointalk.org/index.php?topic=1649106.0

Same risers, same positions for all of the 6th. But I did change the riser on the 7th once the first one was fried. Also, how can a riser, cause such havoc?

it is a dead short   trust me.

   I killed the slot the gpu and the psu when I did it.  My guess is  I had a 750 watt psu and you have a beast to do 7 gpus so you feed it more juice and killed more gear.

I do smaller rigs and rarely bother with  a lot of risers.  Shorts are a big reason and my double cataract repaired eye are okay but not great for depth perception and close up stuff.

I did change the riser the second time as I suspected it might be a cause. I also checked the risers for obvious soldering issues on the capacitors or the PCI 1x connector that goes on the mobo to see if there is any shorting (e.g. solders touching etc). As far as I can tell, I don't see anything that stands out or not normal.

[adaseb]

He said he changed the riser the 2nd time he tried it. What are the chances that 2 risers sequentially were both internally shorted.

Really hard to diagnose without actually being there and seeing what the issue can be.

I know back in the litecoin days people destroyed motherboards due to uncut ribbon risers however with USB risers that's not an issue these days.

Thank you for the comment. Yeah, I understand, it's quite difficult, but at the same time I was almost out of ideas and things to try and 3 mobos, although I can RMA them, it's not a "cheap" thing to try.. I wonder if there is an easier way to test those without risking any gear.. Plug them in and start the PSU with the paper clip method to see what kind of power goes to the riser(s)?

Another thing is this: Currently, the system draws 1080 to 1100 Watts.. with the 7th card, this would have been over the 1200 Watts the PSU supports. However, in that case, it would have just shut down, not fry stuff. Is that assumption wrong? I don't think so, but I'm asking just in case, plus the PSU has short and OCP protection, although the short protection does kick-in in specific conditions so I don't know if a shorted riser will do the trick and trip the protection.

Don't think this is it, since you said as soon as you started the computer it started going on fire. Not when mining. So the PSUs had almost no load.