Topic: Blew up 2 mining rigs with riser cables (Read 8325 times)

Hi guys, you may use any number of psu but you have to connect all the ground together (i mean the black wires).
I use an atx psu for motherboard, hard disk and risers. Then i have a second psu taken from a server (750watt - 70A on 12V) just for the vga.
Then connect 1 black wire from psu 1 to psu 2 and the voltages will be good.

If you don't short them you may test the voltage between the grounds (-) ... it could be 0.2-0.3 volts and that's not good.

 Easy to set up LINUX to automatically and reliably set your TDP as part of the bootup process.

 Windows - well, I mostly like Afterburner but it DOES occasionally lose track of settings for one or more cards....

I have no problems with 8x1080ti rig starting and running at 100% or even 110% TDP.
If you mean that if the rig is not cooled enough for running at high TDP, then GPUs will halt upon reaching 94C degrees or whatever temp limit was set.

One other note: on rig start up sometimes you may forget to run whatever S/W you are using to setup the TDP. This is especially important if you are operating 1080ti's and accidentally forget this step and start at 100% TDP and your rig is laid out for say 65%.

Nothing will happen because there is no potential differences, and however if there will occur a lack of power on some components, system will shut itself down immediately... Also nothing will happen if you lose powering on one of PSU - it will cause the same.

I had many times back to my mining rigs to see that the motherboard is not powered due to PSU failure but the orher one is running still powering risers or GPU's external input or conversely - the cards were not powered but the mobo (system) was running with black screen, but after reboot it was everything OK anyway.

Guys, come on these are not 80's electronics, I know it is chinese crap done for $10 but we still have some protective circuit inside of it.

Regards

i killed 2 Asrock H110 BTC Pro because of crap soldering on risers

after checking the whole rig to the point of mental exhaustion and throwing in the towel, i started to check the soldering points on the riser components and sure enough i found 2 pieces (from the same batch) that had fused solder points, after replacing them and buying a third mobo problem solved

moral of the story is dont trust non quality certified equipment and at least do a visual inspect before use

 The length of the USB cable linking the riser to the "shortie board" that connects to the motherboard DOES NOT AFFECT THE POWER DRAW.

 That is NOT the issue at all.

 Backwards riser - possible but that more likely would cause the system to not power up at all.

 I'm more inclined to think there is an issue with the "add second PSU" adapter - one reason I like the B250 Mining Pro 19-slot MB as it has that part built into the MB and common-links their grounds the easy AND HARD TO MESS UP way.

^^^ This is a perfect example of how to start a fire either by crossing voltages or one psu crapping out while the other is still on.

*NEVER* power the riser and gpu from different psu's.

*ALWAYS* power the riser and gpu from the same psu.

Try downgrading your system a bit. Follow KISS. (Keep it simple, stupid.)

Maybe start with a dual-on board- system, Using 1 PSU, then add risers/GPUs 1 by 1.

This might help you pinpoint where the fuck up is happening. But might not prevent you from doing it again. lol

I agree, while I don't mix my riser power with 2 PSUs (I use 1 PSU for the board and all risers, the second just to power cards) I do know people that do mix the PSUs and their systems don't fry (why take a chance). 

It sure sounds like a riser put in backwards. 

I was thinking if you put your riser connection into PCI-e slot "Backwards", then it sure burnt it.

see this thread, not sure if it is, and might help in case

https://bitcointalksearch.org/topic/burnt-pcie-slot-and-riser-1983900

I don't get it how people can manage to burn their rigs when going into mining stuff, you should learn more not only about that fantastic crypto world but also get some knowledge about electronics, power distribution and how does this all works together in theory and whats happening inside of all your machinery... I used to power my risers stright from the PSU without any capacitors, voltage regulators on boards just because its not even needed if you know what are you doing it's not needed at all, chinese just put this stuff on your boards to look nice and make bigger profit of worth nothing electronic components. To long cords, bad connectors and that all make higher resistance of circuit and failure at all, also ground loops, bad grounding which may cause inductions and amperage rise or falling rapidly.


It doesn't even matter if you have efficient PSU, you could even use 12V car battery to power your risers or GPU's, it doesnt matter if the potentials are fine so it belongs on how you connect PSU's together and thats it.

 Thanks, Reagrds.

3 GPUs per rig is a waste of space and money if you ask me, but to each their own.

I have never had a single riser burning issue and I've connected over a 100 GPUs to various rigs (6 or 8 cards).

The "secret" is to power risers with molex or pci-e cables and avoid the shitty sata connectors that are just molded on the cables. The real risk is to have 2 cables inside the connector melt together. These connectors and sata cable are not rated for high amperage. Also do not power more than 2 risers per molex cable. Get at least ver 006 risers.

With a good quality PSU (do not go cheap on PSU) you're fine.

And just some basic safety rules: avoid to place rigs near water sources or materials that are easily flammable. I've seen some people on here stacking like 10 rigs in a room on the carpet floor. That's just asking for trouble.

2014

As were said above, if you use 2 PSUs then:
Rule #1: Do not mix PSU when powering risers and GPUs: each GPU with its riser should be powered with only one PSU.
Rule #2: Wire both PSUs, rig frame (if it's metal) with strong cable to avoid "ground bounce".
Rule #3: If you use risers, do not plug any GPU directly in MoBo, it is common reason of burnouts.


Just perfect explanation I couldn't make due to problems with technical English vocabulary.

Must be the risers, you have done nothing wrong in the text you posted. No curcuit shortage, nothing. Expensive mining rig u got there  Problems with minin rigs are almost always riser related. What we need is a proffesionall riser company.

I got into mining and being an IT guy I noticed a lot of the guys in this forum either purposely give bad information, or are really that dumb and stumbled into this to make money and have no real understanding of blockchain tech, hardware, GPU's or power.

Now we have mobo's designed for mining, now we just need some nice PSU's.

Got the risers from a Hong Kong vendor.

When I powered up, I saw sparks jump from the 1x slot and heard sizzling (both times). After that, though, there was nothing to see. No burned contacts, no burned traces that I could see. I could smell it though.

I've subsequently reused the "killer" risers, after doing a full electrical continuity test with a multimeter, and verifying no shorts or missing connections, and they work perfectly now that they are powered from the primary supply.

I can only guess that during power up, there was some sort of "ground bounce" problem, which caused the ground voltage on the PSUs to pull apart. I suspect that as the 2 PSU grounds were only tied together by a single piece of 18 G wire on the dual-PSU adaptor, that this was not enough to ensure full equipotential bonding, and that it was the ground current that fried stuff.

Perhaps, if all your PSUs are very strongly grounded together, then you could avoid this problem.

I can see why if the risers are on the same PSU as the mobo, this problem can be avoided. The cards will provide a very low impedance ground plane between aux power input and the 16x slot. This will provide a very strong cross-ground connection between the 2 PSUs. If the card and riser are on the same PSU, then any ground currents have to go through the PCI-e signal and shield wires.