der8auer finds major problem with RTX 5090 power connector

[gingerbee]

Damn even Philly D is talking about this, this won't be good for Nvidia.
@17:58

 

[gingerbee]

Just curious.. out all of these failures..

How many were using Seasonic PSUs?

Just saying, the cable and plug look pretty cheap to my 4 eyes on this big screen.

Derbauer says in his video that it's a quality cable from MODDIY(never runs after-market cables so can't speak on this). He runs his 5090 doing a test with his build using a Corsair PSU and cable, getting the same results as the burnt cable. I don't see how the cable itself could cause power to flow through one wire unless the cables themself were damaged. very strange cause what I know about current/power is it should flow through all the cables equally, I am not sure how this is even possible, clearly it is.

The burnt cable was using an Asus Loki 1000watt PSU with a MODDIY cable.

Derbauer used a Corsair 1600 with a Corsair cable, which had a similar result.

 

[freeagent]

The cable that came with my Vertex GX1000 looks nicer than the cables that I have seen recently go up, that is why I asked.

3rd party stuff doesn't do warranty on anything but cables right?

That is too much of a gamble for me.

 

[gingerbee]

I have no clue about the warranty on 3rd party/after-market cables, you right way too much of a gamble, but it doesn't seem to be limited to after-market cables, kind of wish it was that simple.

 

[Shadowarez]

if they follow the spec laid out by pcie sig /nderpia and it still burns who's at fault

 

[gingerbee]

if they follow the spec laid out by pcie sig /nderpia and it still burns who's at fault

isn't PCIe sig just specs for signalling/communication/power between board and card through that connection, not secondary power delivery. ???

 

[Shadowarez]

I'd sworn they'd have to present this spec to them to get it ratified as a standard maybe I got them mixed up with a different certification body

 

[FreeKnight]

isn't PCIe sig just specs for signalling/communication/power between board and card through that connection, not secondary power delivery. ???

PCIE-Sig is the standards organization both for the board connector and the power delivery (either via board or the psu connection) IIRC. That being said, Nvidia likely has a lot of personnel on the standards board so there was likely quite a bit of cross talk between the two.

Much like if you look at the Canadian Standards Association, the 'board' for various codes is often got a lot of industry personnel on it (along with some academics, occasional NGO personnel etc). I know for CSA Z662 the pipeline standard, it's got tons of TC Energy, Enbridge and other Pipeline staff on the various boards. It's not necessarily malicious as these people often have deep bodies of knowledge and can push for improvements in the code.

 

[gingerbee]

I can find the pages for wire specs but nothing about power delivery specs for the GPU PCB side. I am not sure why they would set those, but I don't have an account with the site, so I can't be sure. I can understand how they would set wire specs and the PCIE slot power specs but not independent board partners as they have no clue what the board needs/wants. but again this is all conjecture. unless Nvidia is pushing the PCIE5/6.0 specs to their limits, I don't see why any standardization by PCIE SIG would allow this to happen. also others have tried to repeat the test only to fail, it's all a big WTF is Happening lol, so many questions with no answers.

When you watch AHO video it does make me think this may be some PCIE SIG specs because Asus's design doesn't fix the problem it makes it safer, but it's not a fix

time will tell I guess once more people get into this in more detail, if they do.

 

[clshades]

20+ amps through a single wire is similar to a 40 gal electric hot water tank so that's NOT good. There's obviously a lot of resistance on those two wires and the wire is suffering for it.