der8auer finds major problem with RTX 5090 power connector

[gingerbee]

well, I am not sure this is the source of the problem I could be wrong, but it is not good, with so many power and ground lines on the 12VHPWR it still doesn't explain why only certain lines are running so much current, I mean the crap connection could potentially increase resistance on that line but that doesn't explain why that one line would be hit with so much current when there are many power and ground lines it should be spread evenly. to my understanding anyway.

 

[ZZLEE]

Tiny Tom Logan

Asus has a gadget program that monitors power on there video cards .

 

[gingerbee]

Tiny Tom Logan

Asus has a gadget program that monitors power on there video cards .

Asus has power sensing per wire on its card, AHO talks about it, it can help but it's still not a fix for this problem it just prevents you from burning your house down, if you catch it.

 

[sswilson]

Tiny Tom Logan

Asus has a gadget program that monitors power on there video cards .

ASUS saved our bacon - We had 12VHPWR/12V-2x6 cable issues

ASUS' power detector helped us prevent a major GPU failure. Our expereince with 12VHPWR/12V-2x6

overclock3d.net

Fairly short text version discussing the issue they had with their cable.

 

[Izerous]

A followup from der8auer, he actually cuts a bunch of the wires to show off some stuff. I like his Summary/Conclusion

Didn't see @lowfat post this in the review thread first.

 

[moocow]

I hope AMD keep using the regular 2x8 or 3x8 pin setup. Sapphire Nitro 9070XT leak screenshot is using a 12V 2x6 instead but I suspect it's an exception.

SAPPHIRE Radeon RX 9070 XT NITRO+ pictures reveal hidden (16-pin?) power connector design - VideoCardz.com

SAPPHIRE set to launch NITRO+ RX 9070 XT card with unique design. Mesh and gold. Finally, some new photos of the NITRO+, Sapphire’s flagship RDNA4 design. This card was included in the family shot released by AMD at CES, promoting the upcoming RDNA4 architecture. Many focused on this NITRO...

videocardz.com

 

[CMetaphor]

Just a quick weigh-in: its 100% true that all the wires should be equal in length, resistance, etc so the problem can't be there. But what about the connector itself?

With so many wires in one spot coming from any angle to the connector, it's possible that the connection for the "melty" wire isn't getting full contact inside the connector. So while all the pins are providing the same power and current, in that one tiny spot inside the connector, the gauge effectively drops to something far too small.

Tiny gauge wire + lots of current = heat. Lots of it. So that's my theory on what's going on here. I haven't watched all the videos above, but cutting up the wires isn't going to find anything. Let's see someone take apart the connector carefully, remove all the plastic, then put it under some magnification to see what happens when the connection is made and flexed around.

Just my 0.02$. I don't have much sense left in the noggin, but these we're mine haha

 

[Marzipan]

how did the QC department not catch this fatal flaw? it's almost beyond comprehension and should have easily come up as an issue with any stress testing or during long term operation tests. >.<

 

[sswilson]

ASUS saved our bacon - We had 12VHPWR/12V-2x6 cable issues

ASUS' power detector helped us prevent a major GPU failure. Our expereince with 12VHPWR/12V-2x6

overclock3d.net

Fairly short text version discussing the issue they had with their cable.

This was my take-away....

What’s clear to us now is that wear and tear can be problematic with today’s 12V-2×6 power connectors. The 12v-2×6/12VHPWR cable standard is designed to handle up to around 660W, and the cable is rated for 600W. This gives this standard a safety factor of 1.1. That’s pitifully small. That means that factors like general wear and tear can cause problems.


For older 8-pin power connectors, their safety factor was 1.92. The cable was rated for 150W but could handle almost double that. That’s why most PC builders have never heard of anyone having issues with traditional 8-pin connectors.

To me those tolerances are way too tight, and it makes sense to me that minor flaws in the mating of connector pins could easily push individual cable current flow beyond what they're rated for.

 

[Izerous]

To me those tolerances are way too tight, and it makes sense to me that minor flaws in the mating of connector pins could easily push individual cable current flow beyond what they're rated for.

A 12VHP with ANY wire cut/damaged/poor contact and assuming perfect load balancing and it is already out of spec. 600W across 5 wires is 120W on a wire only speced to 110W / wire.

Those are not just tight but basically non-existant on a GPU that can actually pull 600W. Fine on a lower end card but not something pushing the limits. That doesn't even take into account boost spikes and such that can go beyond the TDP of a card.