News Veteran reviewer’s RTX 5090 Founders Edition GPU falls victim to 16-pin meltdown — user admits it could have been user error despite having tested...

[Admin]

DSOGaming's EIC suffered a melted power connector on his RTX 5090 out of nowhere. Nevertheless, the GPU continued to function normally for over 20 minutes at full load, despite the issue.

Veteran reviewer’s RTX 5090 Founders Edition GPU falls victim to 16-pin meltdown — user admits it could have been user error despite having tested... : Read more

 

[Pierce2623]

There will be a comment somewhere on here blaming it on user error lol.

 

[JayGau]

There will be a comment somewhere on here blaming it on user error lol.

You mean, like the user himself?

"Despite the many issues, Papadopoulos admits that the melted power connector could have been a user error. “I’m really testing eight different GPUs for each and every game that comes out (I’m not simulating the results). So, it’s possible the cable came loose while removing the GPU from the case. After dozens of tests, this can happen,” said the EIC."

You should read the articles before commenting.

 

[ezst036]

There will be a comment somewhere on here blaming it on user error lol.

The guy did say he might have had a role to play.

Still, a well engineered product would prevent user error, especially of the catastropic type. These are not well engineered products, but it don't matter. Saintly Nvidia will still be revered anyways.

 

[Alex/AT]

It's not a bug, it's a feature.

 

[Alex/AT]

The guy did say he might have had a role to play.

Still, a well engineered product would prevent user error, especially of the catastropic type. These are not well engineered products, but it don't matter. Saintly Nvidia will still be revered anyways.

nV number-crunching chips DO seem to be well-engineered. They are positioned for non-GPU market initially after all. But the whole rest that forms the consumer GPU product is, well... erm... Starting from the risky melty connectors and going to the always flakey drivers with lots of issues unfixed for years and totally outdated messy configuration interface that they are trying to replace with another incomplete mess.
[that's the experience I have from having 4090 for 2 years... yes, the chip is great... the rest does not feel good]

 

[tamalero]

You mean, like the user himself?

"Despite the many issues, Papadopoulos admits that the melted power connector could have been a user error. “I’m really testing eight different GPUs for each and every game that comes out (I’m not simulating the results). So, it’s possible the cable came loose while removing the GPU from the case. After dozens of tests, this can happen,” said the EIC."

You should read the articles before commenting.

Arent the connectors graded for like 50 full connects and disconnects before starting to deform?

 

[DingusDog]

There will be a comment somewhere on here blaming it on user error lol.

TL;DR is obvious.

 

[Amdlova]

The old 6 pin conector have tested 9.999 graphics card from 200x to 2025 without issues =)
Yes it's a use error to accept this connector

 

[Alvar "Miles" Udell]

It's not user error when the wires can break inside the sleeves and leave all the current to be drawn from as little as one wire, which will cause a meltdown even if it's fully inserted. It's a faulty design with insufficient safety margins.

 

[hollywoodrose]

It's not user error when the wires can break inside the sleeves and leave all the current to be drawn from as little as one wire, which will cause a meltdown even if it's fully inserted. It's a faulty design with insufficient safety margins.

Thank you! I don’t understand why many people don’t understand this. Nvidia has produced a GPU without basic safety measures. The GPU should have shunt resistors for all wires, as well as component that monitors how much power is coming in on each wire. Power should be distributed evenly and there should never be a case where anything is melting or catching on fire. When these cables run out of spec - which can happen if power runs unevenly - they can become damaged. Unplugging a the cable can also cause metal to break if it’s already been damaged by running out of spec. Then all it takes is for one wire to fail. After that the rest of the wires will be forced to run out of spec permanently and then you have meltdown.

 

[valthuer]

Approximately two months ago, my PC would restart while playing certain games, like Counterstrike 2.0 and S.T.A.L.K.E.R. 2: Heart of Chornobyl.

Even though i had zero problems with the 3dMark stress tests - my GPU passed them with flying colours - the problem persisted.

Until my screen went black and never turned on again.

I removed my RTX 5090 MSI yellow-tipped connector (which was firmly put in my GPU) and i realised it had been burnt:



I used to think all of the 5090 burnt connector cases were user errors.

Well, joke's on me now.

Serves me right, i guess.

 

[alceryes]

The guy did say he might have had a role to play.

Still, a well engineered product would prevent user error, especially of the catastropic type. These are not well engineered products, but it don't matter. Saintly Nvidia will still be revered anyways.

Exactly this. I've said it a hundred times.
If a product is so poorly designed, that it lends itself to being used improperly, then the design itself is at fault. This is ME 101.

 

[SedrickGates]

User error does not mean user's fault.
An error should result on the GPU not starting or running in degraded mode and have some way of indicating it.
A fault would be to do something that should not be done consciously, like changing the power resistors to get more amps in the GPU for OC. This is a fault, not an error.

Nvidia is sole responsible even if user do not plug it properly. That new connector is a mistake, desktops have the room to accommodate multiple 6 or 8 pins connectors. Or that should have designed beefier cable/pins and/or a safety system instead :-(

 

[Hella_D]

Just go back to the well-proven 8-Pin ATX design must still be perfectly fine, the AMD cards still use it...... DONT FIX WHAT ISNT BROKEN!!!

 

[Alvar "Miles" Udell]

Just go back to the well-proven 8-Pin ATX design must still be perfectly fine, the AMD cards still use it...... DONT FIX WHAT ISNT BROKEN!!!

No, the new spec is needed due to increased power loads, but the design needs to be tweaked to either incorporate thicker cables and monitoring capabilities, or, in my opinion, changed to a single, thicker copper wire and locking coax/BNC type connector.

 

[SedrickGates]

No, the new spec is needed due to increased power loads, but the design needs to be tweaked to either incorporate thicker cables and monitoring capabilities, or, in my opinion, changed to a single, thicker copper wire and locking coax/BNC type connector.

While new specs are just not cutting it, 8 pins ATX seems like the most stable version of them all.
From the 12 pin super connector, the amp per wire has increased while section has not changed. This new connector is by design pushing the limits close to failure One :-(
A better connector is indeed needed and I think, like you wrote, larger copper section to accommodate more amps. Another way would be to get more volts and have a new on-board regulation. Less amps, more volts on the wires is better. That would however require a new ATX Standard :-(

 

[spongiemaster]

Arent the connectors graded for like 50 full connects and disconnects before starting to deform?

No, industry standard is typically around 25 which for anyone that doesn't test hardware for a living is a very large number.

 

[spongiemaster]

Just go back to the well-proven 8-Pin ATX design must still be perfectly fine, the AMD cards still use it...... DONT FIX WHAT ISNT BROKEN!!!

AMD cards don't pull enough power to need them yet. 7900XTX max power draw is around 350W, while the 4090, where all the problems started is over 450W, and the 5090 is over 600W. This connector is not having issues with cards in the 350W range. The original 12 pin connector on the 3090 didn't have the melting issue and it peaks right where the 7900XTX does, 4070 and 5070, which are much higher volume selling cards are not seeing reports of this issue. The problem is not the connector, but how much power Nvidia is asking it to safely transfer. Nvidia needs to lower the spec to about 350-400W. Two of these cables for a 5090 is still a lot better than 4 or more of the old 8 pin.

 

[Amdlova]

AMD cards don't pull enough power to need them yet. 7900XTX max power draw is around 350W, while the 4090, where all the problems started is over 450W, and the 5090 is over 600W. This connector is not having issues with cards in the 350W range. The original 12 pin connector on the 3090 didn't have the melting issue and it peaks right where the 7900XTX does, 4070 and 5070, which are much higher volume selling cards are not seeing reports of this issue. The problem is not the connector, but how much power Nvidia is asking it to safely transfer. Nvidia needs to lower the spec to about 350-400W. Two of these cables for a 5090 is still a lot better than 4 or more of the old 8 pin.

I have a modded ps3 psu who gives 40amp for a short time and 32 amp continously on a 8 pin cable. Still working almost ten years of abuse.

Connector 150w / abusing 480w / 384w soft use