NV 12VHPWR issues revisited

KompuKare · Jan 29, 2025

Ignoring how bad a connector this is, how much overclock headroom can those 550W+ 5090 possibly have?

Maybe an overpriced overclocker's motherboard can give an out-of-spec 100W through the PCIe slots. Plus a risky 700W through the single connector. For a max total of 800W? Although more power might not be 5090's thing anyhow.

Are any aftermarket models going to have more than one connector, or would Nvidia prevent that?

Heartbreaker · Jan 29, 2025

DrMrLordX said:
Even NV just admitted it was a connector problem. Oops!

A problem, in that the sense pins were too long compared to power pins, letting people get away not seating it properly.

That's really the change, that makes a difference. Shorter sense pins. It's all about making sure its properly seated. Even the supposedly problematic cables are identical:

12VHPWR and 12V-2x6 Compared | CORSAIR

With the introduction of the RTX 40-Series, the PCIe 5.0 12VHPWR GPU power connector made its debut with the promise to deliver up to 600W of power through a single connector. As with any new standard, things are likely to evolve quickly and we’re now seeing the introduction of a new connector...

www.corsair.com

Heartbreaker · Jan 29, 2025

KompuKare said:
Ignoring how bad a connector this is, how much overclock headroom can those 550W+ 5090 possibly have?

Maybe an overpriced overclocker's motherboard can give an out-of-spec 100W through the PCIe slots. Plus a risky 700W through the single connector. For a max total of 800W? Although more power might not be 5090's thing anyhow.

Are any aftermarket models going to have more than one connector, or would Nvidia prevent that?

NVidia typically limits how much OC you get, even with partner cards. You usually have to resort to Volt Modding to to really draw the big power. But unless you are doing it for social media views that seems like more risk than it's worth.

KompuKare · Jan 29, 2025

Heartbreaker said:
NVidia typically limits how much OC you get, even with partner cards. You usually have to resort to Volt Modding to to really draw the big power. But unless you are doing it for social media views that seems like more risk than it's worth.

Sure. But then I meant any attempt to push 700W or 800W through those cards is suicide run territory anyhow IMO. Even on water. Certainly not something anyone tight (or sane) with cash would do!

DrMrLordX · Jan 30, 2025

Heartbreaker said:
A problem, in that the sense pins were too long compared to power pins, letting people get away not seating it properly.

Do we have to go back to all the properly-seated connectors that burned out anyway?

Heartbreaker · Jan 30, 2025

DrMrLordX said:
Do we have to go back to all the properly-seated connectors that burned out anyway?

If you see something else that was significantly changed in the update, point it out, because this really appears to be about the sense pins.

The only thing that change does, is force people to plug in their connectors properly.

And what beyond that did third parties add?

Yellow on the connectors so it's more obvious, that you didn't plug it in properly...

An indicator light to so you know when it's plugged in properly...

I know it might be mind boggling, but partially plugged in connectors can't carry their rated power safely...

coercitiv · Jan 30, 2025

Heartbreaker said:
If you see something else that was significantly changed in the update, point it out, because this really appears to be about the sense pins.

One minor observation here would be that they appear to have changed dimension tolerance for elements in the PCB header side view. (essentially everything related to power pins and sense pins)

Another interesting aspect, though again related to sense pins:

Doubling-down on this shift to safety, how the sense pins are interpreted has also been changed. Now if both sense pins are unconnected (open), the ATX 3.1 standard dictates that devices are not to be supplied any power at all. The basis behind this change being that in scenarios where connectors might be incorrectly attached or completely absent, this configuration prevents a PSU from sending power down the line, effectively avoiding potential connector damage or melting.

Suffice it to say, that is a notable change over the previous specification, where two open sense pins was used to trigger 150W mode. Intended as a compatibility measure with 150W 8-pin PCIe power plugs, using open pins for 150W mode meant that PSUs lacked a good (and mandated) way to tell if there is a reliable connection between the PSU and video card. So this fixes that issue by making the null case to not send any power at all; the sense pins on the PSU side must sense something before the PSU will send down power. Which in the case of the newly displaced 150W mode, now requires the two sense pins to be shorted together.

Last bit I remembered to check is related to the quality of the cable plug connection. At some points during the 12HPWR connector's life there appeared a recommendation on Intel's ATX 3 documentation to use 4 spring connections over the 3 dimple version. With the update to ATX 3.1 the specification for 12V-2x6 revert this recommendation and allow multiple connection types but introduce clear specifications for the quality of the connection:

Links:

Will my ATX v3.0 PSU or my GPU be compatible with 12V-2x6? Do I need a new cable/PSU/GPU? Everything you need to know! - Hardware Busters

Hardware Busters - Will my ATX v3.0 PSU or my GPU be compatible with 12V-2x6? Do I need a new cable/PSU/GPU? Everything you need to know! - PSUs

hwbusters.com

The XPG Core Reactor II VE 850W PSU Review: Our First ATX 3.1 Power Supply

www.anandtech.com

https://edc.intel.com/content/www/us/en/design/ipla/software-development-platforms/client/platforms/alder-lake-desktop/atx-version-3-0-multi-rail-desktop-platform-power-supply-design-guide/2.1a/pci-express-pcie-add-in-card-connectors-recommended/

Heartbreaker · Friday at 8:39 AM

coercitiv said:
One minor observation here would be that they appear to have changed dimension tolerance for elements in the PCB header side view. (essentially everything related to power pins and sense pins)

View attachment 115930 View attachment 115931

Another interesting aspect, though again related to sense pins:

Last bit I remembered to check is related to the quality of the cable plug connection. At some points during the 12HPWR connector's life there appeared a recommendation on Intel's ATX 3 documentation to use 4 spring connections over the 3 dimple version. With the update to ATX 3.1 the specification for 12V-2x6 revert this recommendation and allow multiple connection types but introduce clear specifications for the quality of the connection:

View attachment 115932

Links:

Will my ATX v3.0 PSU or my GPU be compatible with 12V-2x6? Do I need a new cable/PSU/GPU? Everything you need to know! - Hardware Busters

Hardware Busters - Will my ATX v3.0 PSU or my GPU be compatible with 12V-2x6? Do I need a new cable/PSU/GPU? Everything you need to know! - PSUs

hwbusters.com

The XPG Core Reactor II VE 850W PSU Review: Our First ATX 3.1 Power Supply

www.anandtech.com

https://edc.intel.com/content/www/us/en/design/ipla/software-development-platforms/client/platforms/alder-lake-desktop/atx-version-3-0-multi-rail-desktop-platform-power-supply-design-guide/2.1a/pci-express-pcie-add-in-card-connectors-recommended/

While there is more detail, It's still really about the sense pins. Remove the sense pin related changes and there really isn't anything substantially different.

First time around they missed the opportunity to use sense pins as a safety feature.

It's clear they just assumed (bad assumption) that connectors would always be plugged in correctly, and then the sense pins would be used to determine how much power the connector was configured to supply. That's all they were for.

Back then: Sense pins were so long you could be losing contact with power and still have a sense pin connection, and even with no sense pins it could still draw some power.

Now: No sense pins detected, no power. Plus they have shortened the pins and tightened the tolerances so they break connection early.

I expect this change will slash the issue by an order of magnitude, to the point where melted connectors happen less than they do for 8-pins (where it does also happen).

coercitiv · Friday at 10:25 AM

Heartbreaker said:
I expect this change will slash the issue by an order of magnitude, to the point where melted connectors happen less than they do for 8-pins (where it does also happen).

I hope so too, though we did get a jump in wattage with the 5090 ... so now we wait for validation

Dave3000 · Friday at 4:24 PM

I have an FE RTX 4090 that I purchased in June '23. It does not have the revised connector but the connector has not melted yet and I'm not sure if it ever will, and I recently did check the connector of the graphics card when I was replacing my primary NVMe SSD and it looks fine. I'm not using the included power adapter that came with my 4090 as I'm using the Seasonic 12VPWR 600W cable that I purchased separately that is compatible with my Seasonic 850W power supply as it provides a direct connection to the power supply from the graphics card and makes it less crowded in my PC. It's also easier to connect than using the included adapter that came with my 4090. At this point what are the chances of the 12VPWR connector of my graphics card melting in the future?

coercitiv · Saturday at 2:55 AM

Dave3000 said:
I'm not using the included power adapter that came with my 4090 as I'm using the Seasonic 12VPWR 600W cable that I purchased separately [...] At this point what are the chances of the 12VPWR connector of my graphics card melting in the future?

That's what I mentioned in the past, I think people who use 12VHPWR cables instead of adapters included with the cards are more likely to have a trouble-free experience. I don't think you'll have issues with your connector as long as the cable is correctly reinstalled after each time you modify your system. (which is a given considering you already know it's important)

DrMrLordX · Saturday at 3:03 AM

Dave3000 said:
I have an FE RTX 4090 that I purchased in June '23. It does not have the revised connector but the connector has not melted yet and I'm not sure if it ever will, and I recently did check the connector of the graphics card when I was replacing my primary NVMe SSD and it looks fine. I'm not using the included power adapter that came with my 4090 as I'm using the Seasonic 12VPWR 600W cable that I purchased separately that is compatible with my Seasonic 850W power supply as it provides a direct connection to the power supply from the graphics card and makes it less crowded in my PC. It's also easier to connect than using the included adapter that came with my 4090. At this point what are the chances of the 12VPWR connector of my graphics card melting in the future?

It's a crapshoot, though if you're using the 450W default profile on the FE then you have more room for error than people using a 600W profile.

igor_kavinski · Monday at 8:05 AM

https://videocardz.com/newz/hong-kong-media-pcm-discovers-melted-cables-and-psu-failure-after-testing-rtx-5090d-5080

Maybe this melting issue is a blessing in disguise? Best way to identify stupid people since most loyal Nvidia defenders keep claiming that it's user error even after hundreds (maybe more than a thousand now?) of meltdowns?

Oh man, my GPU cable melted.

Dude, how can you be SO stupid???

DrMrLordX · Monday at 11:20 AM

Apparently PCM's PSU and cables may have suffered damage while testing their 4090, and they've since chalked it up to "user error" ha ha.

CP5670 · Monday at 11:28 AM

Dave3000 said:
I have an FE RTX 4090 that I purchased in June '23. It does not have the revised connector but the connector has not melted yet and I'm not sure if it ever will, and I recently did check the connector of the graphics card when I was replacing my primary NVMe SSD and it looks fine. I'm not using the included power adapter that came with my 4090 as I'm using the Seasonic 12VPWR 600W cable that I purchased separately that is compatible with my Seasonic 850W power supply as it provides a direct connection to the power supply from the graphics card and makes it less crowded in my PC. It's also easier to connect than using the included adapter that came with my 4090. At this point what are the chances of the 12VPWR connector of my graphics card melting in the future?

The chances are ultimately tiny, especially with a cable that connects directly without the adapter. I have had a 4090 for over two years now with the Corsair cable and the limit set to 500W. You only hear about the problem cases.

There is zero power limit headroom on the 5090 without two of these connectors, which a few cards will probably have like the 4090 did. But on the 4090 the power limit doesn't matter much anyway, they all top out around 2.9-3.1ghz based on chip lottery.

MrTeal · Monday at 11:55 AM

Thanks for the translation, much clearer.

Interesting that their cable seems to have melted at both ends.

reb0rn · Monday at 12:36 PM

modular cable if one pin go bad, it will 80%+ melt at both ends .... so you not just kill GPU but also PSU that is if stupid 16 pin used on both side if its adapter then PSU side WILL NEVER melt as 16pin short all 12V pins as one so you can not overload 3-4 connectors on PSU side ever!!!
Just point how stupid those ppl at NV/intel are!

Heartbreaker · Monday at 5:24 PM

MrTeal said:
View attachment 116212
Thanks for the translation, much clearer.

Interesting that their cable seems to have melted at both ends.

I wonder if DeepSeek did that translation.

Note on VC page:

Update: PCM updated their post with new information. They now confirm they saw burn marks on RTX 4090, so presumably, the issue was spotted after testing 4090. It looks like user error after all.

coercitiv · Monday at 5:35 PM

I find it mildly amusing their conclusion is "user error" based on the GPU that has the burn marks.

reb0rn · Monday at 10:20 PM

Yeah I bet videocardz site as most are paid to shill for NV........ and hope next time they get again free card to review....

Ranulf · 2025-02-05T16:05:34-0500

4090 cable burns up under normal use apparently.

https://www.reddit.com/r/pcmasterrace/comments/1iieo8i/after_over_2_years_of_faithful_service_my_strix

Seba · 2025-02-06T02:10:03-0500

Clearly user error. The user bought a 4090.

coercitiv · 2025-02-06T02:15:51-0500

Props to the OP for their photo shoot setup:

NV 12VHPWR issues revisited

Golden Member

Diamond Member

Diamond Member

Golden Member

Lifer

Diamond Member

Diamond Member

Diamond Member

Diamond Member

Golden Member

Diamond Member

Lifer

Lifer

Lifer

Diamond Member

Diamond Member

Senior member

Diamond Member

Diamond Member

Senior member

Platinum Member

Golden Member

Diamond Member