I would love to get some insight from Nvidia engineers on what happened here.
The 3 series and before were overbuilt in their power delivery system. How did we get from that to the 1 shunt resistor, incredibly dangerous fire-hazard design of the 5 series? [1] Even after the obvious problems with 4 series, Nvidia actually doubled down and made it _worse_!
The level of incompetence here is actually astounding to me. Nvidia has some of the top, most well paid EE people in the industry. How the heck did this happen?
Derbauer and Buildzoid on YouTube made nice informative videos on subject, and no it is not "simply a user error". So glad I went with 7900 XTX - should be all set for a couple of years.
This same thing happening on the 40 series cards was good enough vindication for me not 'upgrading' to that at the time. I'd rather not burn my house down with my beloved inside.
Can't believe the same is happening again.
There is a real problem with the connector design somewhere: der8auer tested with his own RTX 5090FE and saw two of the cable strands reach concerning temperatures (>150ºC).
Video timestamp: https://youtu.be/Ndmoi1s0ZaY?t=883
Is NVIDIA breaching any consumer safety laws by pumping twice the rated current through a 24ish gauge wire? Perhaps by violating their UL certification?
Step the GPU voltage up to 48V. (anyway you make a new connector that's not compatible with existing PSUs. Why not actually fix a problem at the same time, once and for all! [48V should be enough for anybody, right?])
It's strange how Nvidia just doubled down on a flawed design for no apparent reason. It doesn't even do anything, the adapter is so short you still have the same mess of cables in the front of the case as before.
This shit is so fucking dumb. Sorry for the unhinged rant, but it's ridiculous how bad every single connector involved with building a PC is in 2025.
I'm just a software guy, so maybe some hardware engineer can chime in (and I'd love to find out exactly what I'm missing and why it might be harder than it seems), but why on earth can everything not just be easily accessible and click nicely into place?
I'm paying multiple hundred dollars for most of these parts, and multiple thousands for some now that GPUs just get more and more expensive by the year, and the connector quality just gets worse and worse. How much more per unit can proper connectors possibly cost?
I still have to sit there stressing out because I have no idea if the PSU<->Mobo power connector is seated properly, I have no idea if the GPU 12VHPWR cable is seated properly, I'm tearing skin off my fingers trying to get the PSU side of the power cables in because they're all seated so closely together, have a microscopic amount of plastic to grip onto making it impossible to get any leverage, and need so much force to seat properly, again with no fucking click. I have no idea if any of the front panel pins are seated properly, I can't even reach half of them even in a full ATX case, fuck me if I want anything smaller, and no matter what order you assemble everything in, something is going to block off access to something else.
I'm sure if you work in a PC shop and deal with this 15 times a day you'll have strategies for dealing with it all, but most of us build a PC once every 3 years if that. It feels like as an average user you have zero chance to build any intuition about how any of it works, and it's infuriating that the hardware engineers seem to put in fuck all effort to help their customers assemble their expensive parts without breaking them, or in this case, having them catch fire because something is off by a millimetre.
This space feels ripe for a radical re-design.
If we’re going to keep up these kilowatt scale cards, we’re just going to need higher voltage rails on PSUs. I had a bunch of similar dumb power connector problems when my 4090 was new.
Discussion (22 points, 1 day ago, 10 comments) https://news.ycombinator.com/item?id=42996057
At this point, I'm waiting for the first RTX generation that just comes with its own separate PSU and wall plug cable.
You only paid $2000 for it, what did you expect.
There are plastics that can deal with high temperatures.[1] They're heavily used in automotive applications. They're not often seen inside computers.
Still, 50 amps inside a consumer computer is excessive. At some point it's time to go to a higher voltage and get the current down.
Someone remind me again why GPUs need 600 watts? I never liked the concept of having to plug a power cable into a GPU, but these new connectors are just terrible...
der8auer got his hands on the actual card, cable and PSU: https://www.youtube.com/watch?v=Ndmoi1s0ZaY (I'm assuming the content is identical to the German https://www.youtube.com/watch?v=puQ3ayJKWds - I haven't watched the English one)
Notable is that on the PSU connector side, 5 pins show heat damage. That means at minimum those 5 must have been carrying some current; i.e. only one of the 6 connections could have failed completely open.
On the PSU one of the ground pins was melted into the PSU connector, this should allow verifying if the plug was fully inserted by a lab disassembling and cross-sectioning it.
The extended title and subtitle say a lot.
> -- Uneven current distribution likely the culprit
> One wire was spotted carrying 22A, more than double the max spec.
Idk why we haven't migrated to better power standards for PC/computers.
I loathe those molex connectors, especially the 24-pin mobo one; it's just ridiculously massive/overcomplicated.
Like, instead of the very rectangular 24-pin with a flat but wiiiiide cable, why don't we have a circular connector with a circular cable that is a lot more manageable for mini-itx builds.
I've believed for a while now that GPUs are eventually going to get so big they'd need to be external to satisfy power and cooling requirements, and this is just more proof of that IMO.
There's a meme image somewhere out there of a GPU crudely photoshopped to look like a split-type AC condenser unit, i.e the kind you mount outside your house. It's pretty much how I picture GPUs will end up if things keep going the way they are.
Why are people using 3rd party cables after the 40 series disaster?
I've designed PCBs using connectors with multiple parallel wires to carry high currents. It's hard to do, and very easy to mess up during later PCB changes. You have to match trace resistances from each pin to the destination. The resistance in the traces is probably higher than the resistance in the wires, so the wires don't help balance the current much.
It's easy to mess up during minor PCB changes because the netlist doesn't capture the important bit. An autorouter would happily connect the pins of the connector together and cause uneven current sharing.
This needs a name along the lines of The Red Ring of Death. What will we call it?
On a side note, there are no 5090's on amazon.com, except as part of a system. I suppose this is because any that become available are purchased and amazon doesn't display out of stock items in their search results? There are 5080's.
https://www.amazon.com/s?k=nvidia+rtx+5090+32gb+ddr7&i=elect...
For the current running through these wires I would have never imagined that the gauge they have and the pin spacing would never work. It's nuts.
I hope these connectors are transitional and limited to NVIDIA, and not something that the industry adopts moving forward.
Instead, a proper standard should be developed.
So after all the drama with the 4090, this is where we've landed? Another generation, another round of melted connectors.
I really hope this means that we reign back in the power on these high end cards. 600 what is just too much for a connector like this. 450watts seems much safer (though I wish the spec as a whole had better margins). Nvidia really just tried to pass the 5090 as a new generation by pumping more power through it and it shows.
This has been the situation since the last 4090's melting as well. Just check youtube channels for videos of the repairs daily from these folks. It's fun to watch, less if you actually bought one.
I wouldn't buy one after watching these and have, period. The rest are beta testers, now two generations in of failure and still lining up in tents.
FWIW, I had the same issue with my 3090 (though I believe that uses a slightly different port?). I was using a custom cable like this guy. Nvidia replaced it under warranty, and I went back to using the (ugly) provided adapter.
I hope that a brand that produce PSUs and GPUs develop a higher voltage rail and a card that goes with it as an open standard.
Wishful thinking, I know.
Especially because I don't even know if they can drift from the nVidia/AMD specs that much before being sanctioned or something.
Yeah, they will be more expensive, but I'd rather pay few bucks more and be safer/not to worry to burn my house down.
They should pigtail the GPU straight into the AC wall voltage and build in a dedicated power supply to convert to DC.
The thing takes enormous power. Some people had trouble with the 4090s too but I haven’t and I run a shit ton of them.
Why can’t they just use a cable+socket similar to PSU - wall socket? It’s not even multiple-kilowatts range.
Bring back screw in connectors.
Why don't they just put two connectors on the 90-series cards?
150 deg? You can nearly bake a pizza with that.
What’s wild is that this is a company clearly capable of designing highly complex things with numerous tradeoffs, challenges and unknowns. And then the fuckin cable is the issue. Repeatedly.
So when will PCs migrate to 48V? :)
Recall in 3, 2, 1...
Frankly I'd be fine using barrel connectors to power my 5090, heck even consider just directly soldering the connection.
All these incidents are from aftermarket cables.
People are letting their hatred of Nvidia blind them to what happened here, a customer upgraded from a 4090FE to 5090FE, they were using a 3rd party ASUS 12VHPWR cable and didn't realize the 5090FE actually uses 12V-2x6 not 12VHPWR, while the port is the same the pins are different lengths.
End of the day PC power cabling is such a shambles with things that look standard but are not so you should only ever really use the cables that came with the product if you don't want to risk issues, especially with this specific ports poor history with 3rd party cables.
12vhpwr has almost no safety margin. Any minor problem with it rapidly becomes major. 600W is scary, with reports of 800W spikes.
12V2x6 is particularly problematic because any imbalance, such as a bad connection of a single pin, will quickly push things over spec. For example, at 600W, 8.3A are carried on each pin in the connector. Molex Micro-Fit 3.0 connectors are typically rated to 8.5A -- That's almost no margin. If a single connection is bad, current per connector goes to 10A and we are over spec. And this if things are mated correctly. 8.5A-10A over a partially mated pin will rapidly heat up to the point of melting solder. Hell, the 16 gauge wire typically used is pushing it for 12V/8.5A/100W -- that's rated to 10A. Really would like to see more safety margin with 14 gauge wire.
In short, 12V2x6 has very little safety margin. Treat it with respect if you care for your hardware.