Unlucky PC user finds AMD's cheaper 8-pin GPU connectors can melt, too

[Daniel Sims]

In brief: Nvidia's RTX 4090 and 5090 have become notorious for melting power cable connectors. While the two flagship GPUs have drawn renewed attention to the problem over the past several years, users should remember that lower-wattage cards and cables aren't immune, even with power monitoring devices, as one unlucky AMD Radeon RX 7900 XTX user recently discovered.

Redditor AwarenessHour9192 recently complained about a melted power connector on a PC graphics card, an issue commonly associated with Nvidia's top-of-the-line GPUs. However, this time the affected card was an AMD Radeon RX 7900 XTX.

The user reported black screens and other problems with his PC over the course of months and could not diagnose the issue, even after a workshop reinstalled Windows and ran benchmarks. Only after disassembling the tower did he discover burn marks on the GPU's power connector.

Nvidia's RTX 4090 and 5090 have frequently experienced the same problem over the past few years. While the company attempted to close the issue long ago by blaming user error, users have largely blamed its 16-pin power connectors and the 5090's near-600W power draw, with cases continuing to emerge. YouTube channel TecLab recently renewed awareness of the flaw by demonstrating two extreme ways to circumvent it.

First, they lowered a 5090's operating temperature by soldering its PCB directly to the power cables, bypassing the connector entirely. On another unit, they proved that the 16-pin connector might be unnecessary by replacing it with a triple 8-pin setup without sacrificing performance, which also lowered temperatures.

AwarenessHour's issue occurred with AMD's last-gen flagship GPU, which draws only 400W via an 8-pin connector. Furthermore, instead of plugging the power connector directly into the graphics card, he passed it through a Thermal Grizzly WireView power monitor, which is designed to prevent this exact problem. Luckily, the monitor prevented the damage from reaching the GPU, whereas many other cases have left severe burns on Nvidia graphics cards that now cost over $5,000.

While AwarenessHour has not yet determined what caused the meltdown, Thermal Grizzly fortunately replaced the power connector and the WireView. The incident shows that even GPUs with modest TDPs and 8-pin power connectors can suffer from melted connectors. Users should still take care when installing them and check for physical damage if they experience video output issues.

Permalink to story:

Unlucky PC user finds AMD's cheaper 8-pin GPU connectors can melt, too

 

[Rrrgg]

8 pin is the goat

 

[NumberNine]

The incident shows that even GPUs with modest TDPs and 8-pin power connectors can suffer from melted connectors.

I'm failing to see the modesty of a dual slot GPU with three fans using 400 watts, more than the rest of the average gaming PC combined.

 

[MasterMace]

Seems more like an issue with WireView than the standard design of a PCIe cable.

 

[PlayerTTT]

The 12Vx6 connector is bad and under spec'd, that is for sure. But if people had any doubt all these issues are from a new trend in cutting costs and bad GPU VRM design, here is another data point.

The fault is not the PSU, it is not the connector. If you want single rail PSUs (which we do), the cards need to limit current per pin properly.

 

[Squid Surprise]

So it’s not an inherent flaw with Nvidia? *gasp*

It effects a minuscule percentage of cards and is largely a non-issue… but people need the clicks I guess.

 

[Theinsanegamer]

I'm failing to see the modesty of a dual slot GPU with three fans using 400 watts, more than the rest of the average gaming PC combined.

Context. It's hard for some, you should look it up.

In the context of the melting plug hysteria, 400w is modest draw.

 

[poshflamingos]

I'm failing to see the modesty of a dual slot GPU with three fans using 400 watts, more than the rest of the average gaming PC combined.

Bro, WTF is happening to TechSpot writers? So many of them are writing the dumbest stuff lately.

Also, articles about a reddit post should be banned. How is this guy not ashamed of trying to pass random internet posts as "journalism" like this? Doesn't he have even a shred of pride or dignity for his profession? Why would any writer scrape the bottom of the barrel like this?

We criticize AI for taking human jobs, and rightfully so, but ask any AI if passing a reddit post off as journalism is a good idea and the AI will tell you you're being insane. Hard to argue with AI taking over tech journalism jobs like this.

Anyway, everyone, take good care of your modest 400W GPUs. As you can see from the experiences of this redditor, even simple low-power devices like the modest 7900 XTX can have power delivery issues.

 

[Theinsanegamer]

Bro, WTF is happening to TechSpot writers? So many of them are writing the dumbest stuff lately.

Also, articles about a reddit post should be banned. How is this guy not ashamed of trying to pass random internet posts as "journalism" like this? Doesn't he have even a shred of pride or dignity for his profession? Why would any writer scrape the bottom of the barrel like this?

We criticize AI for taking human jobs, and rightfully so, but ask any AI if passing a reddit post off as journalism is a good idea and the AI will tell you you're being insane. Hard to argue with AI taking over tech journalism jobs like this.

Anyway, everyone, take good care of your modest 400W GPUs. As you can see from the experiences of this redditor, even simple low-power devices like the modest 7900 XTX can have power delivery issues.

The downfall of the quality in tech articles reflects the continued decline of their viewership audience that write insane drivel like this.

 

[PlayerTTT]

So it’s not an inherent flaw with Nvidia? *gasp*

It effects a minuscule percentage of cards and is largely a non-issue… but people need the clicks I guess.

No this is industry wide. I think only Zotac was isolating circuits somewhat, most manufacturers don't even use fuses anymore. So when something goes wrong, you get a hole on your board.

If someone designs an AMD GPU without balancing or limiting the power per pin, it can also fail. Doesn't matter ASUS saying they monitor the power per pin, it is complete nonsense.

If you shunt all pins into one before splitting them onto VRM chips, when some of those pins have less resistance than others due to age, those will get most if not all the current. This is basic electronics you learn in school.

Where if you are load balancing properly, with some extra resistance you would have heat but only up to what the pins/connectors are rated for. More likely, the voltage drop on the pins would mean the GPU would start crashing before it melted something due to lower currents.