PCIe 12VHPWR and 12V-2x6 power connector issues

[arandomguy]

The flip side of this is wouldn't the new standard have a much higher margin of safety for the lower power cards?

You do get a clean single connector for the entire line as well, instead of ranging from 1x6 pin to now potentially 4x8pins.

My understanding also is that Dell was one of the proponents of this. So perhaps this decision also needs to viewed through the lens of of OEMs doing large scale prebuilts.

I'm also wondering if the sense pins have practical benefits.

 

[Albuquerque]

Well, 50 amps through 12 pins is a bit over 4A per pin. USB-C can carry 100W max, and uses a grand total of two power pins... Yoy ever stop to look at how tiny USB-C pins are?

I think we're OK.

 

[Kaotik]

Well, 50 amps through 12 pins is a bit over 4A per pin. USB-C can carry 100W max, and uses a grand total of two power pins... Yoy ever stop to look at how tiny USB-C pins are?

I think we're OK.

It's not 50 amps through 12 pins, it's 50 amps through 6 pins, the other 6 are grounds.
And it gets worse, according to Der8auer he found in his testing one cable carrying 23A when the cable (16AWG) is rated for 10A/13A

As for USB-C, they use high voltage and relatively low amps (5A max)

 

[DegustatoR]

according to Der8auer he found in his testing one cable carrying 23A when the cable (16AWG) is rated for 10A/13A

I wonder how that happened considering that all cables there are supposed to connect to one point on both sides of the cable.

 

[Kaotik]

I wonder how that happened considering that all cables there are supposed to connect to one point on both sides of the cable.

Most likely there's some issue with either the cards or PSUs side of things, but these kinds of issues explain why cables melt

 

[IQandHDR]

I wonder how that happened considering that all cables there are supposed to connect to one point on both sides of the cable.

He was talking about it was a no-no to mix gold/tin/other metals in the connectors as the cable in question apperantly did.
Also he mentions that the FE cards does not have Ampere sensors.

 

[DavidGraham]

HardwareLuxx can't replicate these results. Either on 4090 or 5090.

https://twitter.com/x/status/1889311313135841697

 

[DegustatoR]

Most likely there's some issue with either the cards or PSUs side of things, but these kinds of issues explain why cables melt

There can be no issues leading to such result on either. The wires are crossed in the cable connectors. The only way to get higher amperage over a couple of wires would be by physically modifying the connectors.

 

[Kaotik]

There can be no issues leading to such result on either. The wires are crossed in the cable connectors. The only way to get higher amperage over a couple of wires would be by physically modifying the connectors.

So you're suggesting Der8auer on purpose modified the cable to get these results? That's quite an accusation.
You're also disregarding all possibilities of manufacturing issues and other malfunctions. The fact is we can all see his measurements in one

 

[Kaotik]

HardwareLuxx can't replicate these results. Either on 4090 or 5090.

https://twitter.com/x/status/1889311313135841697

It's not suggested that it's normal behaviour, it just shows how some sort of issues can lead to cables melting

 

[homerdog]

So you're suggesting Der8auer on purpose modified the cable to get these results? That's quite an accusation.
You're also disregarding all possibilities of manufacturing issues and other malfunctions. The fact is we can all see his measurements in one

Buildzoid is in on the conspiracy as well

Pretty good explanation of why the 3090 didn't have this problem.

 

[Cappuccino]

Asking for somebody to provide a pdf of a contract between Nvidia and an AIB manufacturer is the definition of complete ridiculousness.

Yeah I know nobody can show this, because we have no earthly idea how these guys interact. That’s my entire point…

 

[iroboto]

Just got mine. Gigabyte gooo.

Managed to get the one at the price point I wanted: 5080 FE price. 1450 CAD.

What a rush. I live 15-17 minutes from Brampton CC, but only 2 stop lights and a highway inbetween. Managed to get the last one of 3. The dude came in 30 seconds right after me.
My heart is still thumping, more fun than playing video games. Who knew driving around the city chasing down video cards would be so euphoric.

 

[DegustatoR]

It's fascinating that you could cut 5/6 of the load bearing conductors in a 12V-2x6 (H++) cable and you wouldn't know anything is wrong until you smelled the melting plastic.

Yeah cause you'd cut them in your sleep so you wouldn't know that you have actually done something wrong.

 

[homerdog]

Yeah cause you'd cut them in your sleep so you wouldn't know that you have actually done something wrong.

To word it differently, 5/6 of the pins could be making poor/no contact and you wouldn't know anything was wrong. a 3090Ti would not turn on in that situation.

 

[DegustatoR]

To word it differently, 5/6 of the pins could be making poor/no contact and you wouldn't know anything was wrong. a 3090Ti would not turn on in that situation.

That's brand new information! We've only spent several months on that topic during Lovelace launch so it is definitely time to return to that now because there's one plug which has melted somewhere.
Also this is completely different from cutting off wires in the cable. With a bad contact you don't need to cut anything.

 

[iroboto]

lol. Oh I can see why this narrative is playing out.

The store asked me if I wanted to pay for warranty incase I effed up my 12v. Now I have to think about PSU being sufficient. I think a 750Gold should be ok. But these are big cards. A bit worrisome to know people’s connectors are melting.

is there software to test your card before you decide to really work it out in a game ??

 

[homerdog]

I would urge anyone to watch the der8auer and buildzoid videos and draw their own conclusions. Buildzoid even has nice wiring diagrams that explain what is different with the 4090/5090 vs older 12VHPWR cards and why those changes are problematic.

It is still unclear what is wrong with der8auer's setup. It's probably not the case that he is just an idiot or he built his test bench in his sleep. Hoping he does some further investigation.

 

[DegustatoR]

A bit worrisome to know people’s connectors are melting.

Just plug it in completely and preferably don't bend the cable too much.
We're at a third year of people worldwide using these cables without any outstanding amounts of issues and that's with the old 12VHPWR sockets.

 

[Florin]

I would urge anyone to watch the der8auer and buildzoid videos and draw their own conclusions. Buildzoid even has nice wiring diagrams that explain what is different with the 4090/5090 vs older 12VHPWR cards and why those changes are problematic.

It is still unclear what is wrong with der8auer's setup. It's probably not the case that he is just an idiot or he built his test bench in his sleep. Hoping he does some further investigation.

Yeah I do find Buildzoid's claims worrying. It does not look like a very foolproof design.

On the other hand, Andreas Schilling is a legend, and his measurements on the power distribution across the 6 12V wires look relatively solid (and he also has one for the same 5090 Aorus Master I'm using).
I guess it's something of a relief that it does not use the Nvidia reference PCB. I can see at least 2 shunts/power delivery channels on teardowns.

Overall, I remain of 2 minds on this issue. I've used 4070Ti/4080 with both Nvidia branded and OEM 12vhpwr adapters. My 5090 is plugged into a Corsair HX1000i which has 12v2x6 support.
Never had any problems (knock on wood). I reckon 4090 and 5090 owners do need to be paying close attention. Cards below that, well, far less of a concern.