msi 5090 "Bite The Dust" 12HPWR melting

[DarkOmenX]

In my case bought an overpriced MSI RTX 5090 Liquid SOC in May 2025 to replace my MSI RTX 4090 Liquid X, which I’d had for about three years.
Back when the melting 12HPWR connector issue was happening, I got a 1200W Thermaltake Toughpower GF3 PSU with a dedicated 12HPWR cable to protect the card, and it worked perfectly until yesterday.
Out of nowhere no warnings, no smell ,everything seemed fine while I was playing Cyberpunk 2077, then suddenly, black screen and the PC froze.
After restarting, Windows booted normally, but as soon as I tried to resume the game, the same thing happened.
Feeling confused, I shut the PC down and looked at the card through the glass panel, but everything looked normal.
On the third boot into Windows, I tried running Warhammer 40K: Space Marine 2, and again, right when the game started, black screen and freeze.
I realized something was going terribly wrong, so I shut down the PC and started inspecting the card again. It seemed normal, so I decided to unplug the 12VHPWR connector, which came out without any obvious signs of melting. However, when I looked at the pins, I noticed one was dark or black, and the surrounding area showed signs of burning or melting. On the GPU side, the same port was also dark, but not burned more like it had residue from the connector.
The inside actually looked fairly good, considering what happened. Then I checked the PSU side, and the situation was very different the 12VHPWR connector showed significant melting and deformation.
That’s when reality hit me. What I had read and seen happening to others, hoping it wouldn’t happen to me, had just happened. No warnings, no burnt plastic smell, nothing out of the ordinary .
I didn’t turn on the PC again and got in touch with the retail shop that sold me the card. After some back-and-forth discussion, they asked me to bring in the 5090 and the PSU. When I arrived, they seemed quite surprised and told me I’m the first person to report this melting issue. Like me, they were aware of it but were hoping it wouldn’t happen though I guess there’s a first time for everything.
They said they need to have it inspected and follow some RMA procedures before they can actually replace the card.
I reinstalled my RTX 4090 Liquid X and bought a brand new PSU, the MSI MEG Ai1300P, and now the PC is working fine again. I have no idea how long it will take for the 5090 case to be resolved, but I’ll report a follow-up when I know more.

 

[Svet]

That looks more like a PSU issue rather than GPU issue.

 

[DarkOmenX]

That looks more like a PSU issue rather than GPU issue.

Well hard to say, it is 12VHPWR issue for sure , but the GPU is the one pulling the power and the PSU was more then enough to handle any load of 5090 .this shouldn't ever happen .
Maybe the GPU survived but I'm not the one who is going to test it is responsibility of the Retailer however only because the GPU side seems fine doesn't mean it is and in 12HPVR melting both sides are effect GPU and PSU only difference is in what extent.

 

[Svet]

but the GPU is the one pulling the power and the PSU

The GPU will be pulling the same power from any PSU

and the PSU was more then enough to handle any load of 5090

Results shows opposite.

it is 12VHPWR issue for sure

Yes, that could be too, but this is also part of the PSU.
[or both]

Well hard to say

It happens on the PSU's side of the power connector, so clearly the issue is there.

 

[DarkOmenX]

The GPU will be pulling the same power from any PSU



Results shows opposite.



Yes, that could be too, but this is also part of the PSU.
[or both]



It happens on the PSU's side of the power connector, so clearly the issue is there.

By that definition it will always happen on the PSU side, as the PSU is the one providing the power. Did you research the 12VHPWR melting problem? Here, it could be that the protection kicked in and maybe saved the card, though I can’t confirm this without risking the new PSU. You say 1200W isn’t enough for the 5090? The technical specifications say otherwise, and I had plenty of power headroom, so I completely disagree.
This PSU powered a 4090 for three years and has been running the 5090 since May 2025.
The 12VHPWR melting issue is ongoing with the 5090, 4090, 5080, and 5070 Ti regardless the brand of PSU or VGA

 

[DarkOmenX]

I believe what DerBauer found is exactly what’s going on check out the video.

 

[Svet]

You say 1200W isn’t enough for the 5090?

I didn't means this.
On theory it should be plenty of enough power of course.
I don't doubt if the power is not enough or PSU's have lag of power.
But I suspect that something is wrong in the power delivery of the PSU itself (or this particular PSU's rail) / or/and it's cable.

Did you research the 12VHPWR melting problem?

Yes

I believe what DerBauer found is exactly what’s going on check out the video.

Yes, I know this video.

The 12VHPWR melting issue is ongoing with the 5090, 4090, 5080, and 5070 Ti regardless the brand of PSU or VGA

For 5090 & 5080 mostly, however I think you have a different issue because the melting happens on the PSU's side,
and not the typical melting of the Power cable (GPU side) and GPU's power pins melting on the GPU side.

 

[DarkOmenX]

Whatever the case, I’m now out $3,800 USD and $350 USD for the new PSU, plus about $280 for the old PSU. No other components in the PC were damaged, and the system had been running under the same conditions for months. The VGA, like everything else, is under warranty, so I expect the retailer and MSI to honor it. I know I didn’t do anything wrong all connectors were properly seated. I’m an electronics engineer working for a company that maintains millions of dollars’ worth of equipment. This should not have happened, especially since the power supply was connected to a protected circuit, meaning no external overcurrent or overvoltage could have affected the connected systems.
Right now, I'm waiting to hear what has been decided, and I believe it might take some time.
Good thing I didn't sell the 4090 Liquid X

 

[Svet]

Got it, let me know how it's gone and what they will find out.

 

[DarkOmenX]

Got it, let me know how it's gone and what they will find out.

Good News/ Bad News
Good News:
The card survived the ordeal, and extensive testing on benchmark loops for 12 hours without a single crash was done; I had the opportunity to witness its operation. I also had them run the performance tuning from Nvidia's app to see if the card would complete the cycle or crash. It completed the cycle with +125 on Core and 200 on Memory without any crashes.
At this point, the retailer has no reason to issue an RMA or warranty exchange since the card is functioning as intended, and "even better" once the performance tuner finishes the cycle and the result is presented.
It is evident that the Thermaltake Tough Power GF3 1200W PSU protection circuit prevented damage to the card
I agreed with this assessment and got my card back.
The retailer suggested that I contact Thermaltake and present the case. Since the unit comes with a 10-year warranty, it should be replaced with the same or a similar model. As this particular PSU was not purchased from their store and is now 3 years old, they advised that the best way to seek compensation is to reach out to Thermaltake support directly.
Bad News:
I personally don't trust this card at all. I'm afraid that what happened before may likely happen again. While there's a chance it won't, I'm highly skeptical and worried it will damage another PSU my new MSI Ai 1300P. If another incident occurs, the card may not survive, and once the warranty expires, I'll be left with a loss of $3,800 by MSI, Nvidia, Intel, and PCISIG with no alternative but to "suck it up." Just knowing this makes me extremely angry and disappointed.

 

[Alan J T]

I personally don't trust this card at all. I'm afraid that what happened before may likely happen again.

Looks ike a Cabel fail, one of the Cripms has moved too far back in the Plug
You should not be able to see the Cripm at all
With that one bad connection I can see time did the rest of the damage till it failed complety

 

[DarkOmenX]

Looks ike a Cabel fail, one of the Cripms has moved too far back in the Plug
You should not be able to see the Cripm at all
With that one bad connection I can see time did the rest of the damage till it failed complety
View attachment 207539

My take is that the overheating occurred on a single rail, causing the insulation to melt and break away. As it cooled down, the insulation pulled back. There was never any pull or tension on that plug; however, it appears as though it was forcibly pulled. The wire has also thickened, possibly due to the melting shroud now expanding inside the cable. After cooling down, the entire cable in the affected area is very stiff, which also suggests that further melting may have happened. You can clearly see the expansion of the cable in the connector area.
If Thermaltke would like to pay for shipment Im willing to ship it and let them inspect it and get some conclusion out of it. By now the connector is still attached and in the condition it was when I discovered the issue.

 

[DarkOmenX]

What I find intriguing is that the protection circuit only affected the VGA, shutting it down instantly and causing the PC to freeze, but it did not shut down the entire system.

 

[Svet]

As expected results, no surprise.

and bought a brand new PSU, the MSI MEG Ai1300P

That's a good choice.
FYI:

MSI BEST PSU FOR 5090

The RTX 50 Series GPUs have launched, and the RTX 5090 in particular draws up to 575W of power. That kind of power demand means your power supply unit needs to be ATX 3.1 ready.

www.msi.com

 

[DarkOmenX]

yeah , my 4090 Liquid X is still on its original 600W Bios and I saw this card drawing 610W in peak the PSU was and is sufficient.

Toughpower GF3 1200W Gold - TT Premium Edition

Designed with a native PCIe 12+4pin connector, the Toughpower GF3 series is compatible with Intel ‘s ATX 3.0 specification and is built for next-gen PCIe 5.0 graphics cards. The wattage marked on 12VHPWR’s connector is the recommended wattage. (Ref. Intel design guideline) The native 16 PIN...

www.thermaltake.com

 

[DarkOmenX]

I'm almost 99% sure the 5090 will kill itself and the new PSU in due time I just hope it will be sooner than later, so I can finally have the Warranty engaged and the card replaced or a lawsuit will follow.

 

[agrebinidhal]

This is clearly a design flow, i can't understand how some of the best electrical engineers in the industry designed something that lacks basic protection and monitoring features like that, by the way your 4090 is using the old or the new connector because i just got a 4090 suprimx with an msi a1250gl psu ensured the cable is fully seated and no visible gaps from either ends, no pressure on the cable from the side panel, i even added a sponge support underneath the cable and 2X140mm to keep it cool but i'm still worried