Yesterday, I was playing WUCHANG: Fallen Feathers when I noticed a strange smell. I couldn’t figure out where it was coming from, so I decided to open my PC case. That’s when I saw it — the power cable of my NVIDIA GeForce RTX 5090 was burned, and smoke was coming out of it.
For those unaware, it’s well known how unreliable the 12V-2×6 power cables of the NVIDIA RTX-40 and RTX-50 series are. We’ve seen burned power cables for the RTX-40 series. However, this is the first time I’ve seen something like that on the RTX-50 series GPUs.
Here are some images that show the melted power connector and cable.
Now, since I’m well aware of the power cable issues, I always make sure the power cable is fully seated in the GPU. Not only that, but to avoid any accidents, I use different power cables for each and every GPU I test. For the NVIDIA RTX 4090, we use its official one. For the RTX 5080, we use the one it came with. And for the RTX 5090, we use its own. In fact, I don’t remove the power cable from the RTX 5090 whenever I remove the card from my PC case.
Still don’t believe me? Here’s a picture I took before removing the power cable. You can clearly see that it was fully plugged in. No matter how hard you push it, it can’t go in any further.
One of our readers noticed a “line” or “gap” in this image. That was as much as the cable could go in on our specific RTX 5090 Founder’s Edition. You can also clearly spot this “line” or “gap” at the connector in the video at the 0:12 mark. In the video, we had the RTX 5090 running for over 20 minutes at 100% usage. No burn, no smoke issues. Temps were fine (below 78 Celsius with auto fan), and power draw was correct.
Assuming the cable wasn’t fully seated, the GPU should be smoking in the video, not working perfectly. In other words, you can’t say “I see that line, it wasn’t fully seated” but then say “Well, I see the line in the video and it wasn’t fully seated, but it did not burn and it works fine“. If you can’t 100% replicate what you say, then your assumption is not correct. That’s how things work. That’s why companies always ask you for the steps to reproduce your problem, whether it’s a hardware failure or a software issue.
What’s even funnier here is that the top row of connectors didn’t get damaged at all. See that “gap” at the top in the picture? Yep, those top connectors didn’t burn. If the cable wasn’t fully seated, those would be the connectors that would get melted. Instead, the ones that burned were the bottom ones. And if you look closely, the bottom of the connector actually has more contact than the top.
Again, I’m fully aware of the melted power cable issues of the RTX-40 series. That’s why I always make sure to push the power cable in as firmly as possible. But if you look at the pictures of the burned cable, it seems like the bottom row of connectors wasn’t fully plugged in. Not the top. That’s the only thing that makes sense. So somehow, even though I pushed the cable in really hard, it still came loose. But here’s the weird part — if that’s true, why didn’t the middle connector at the bottom row melt too? It doesn’t make sense.
What’s really strange is that I didn’t experience any crashes or problems while playing. The only reason I found out something was wrong was because of the weird smell. Just to test things out, I removed the RTX 5090, then plugged it back in and ran WUCHANG for 20 minutes while keeping an eye on the GPU. I also recorded a video of the test. Surprisingly, the PC stayed stable, and there was no smoke coming from the power connector. Everything seemed to run fine — at least during that stress test.
The bottom line is this: the RTX-50 series still has problems with burned or melted power cables. And if it happened to me, it could happen to anyone. Don’t get me wrong, I’m not saying I’m the best of the best. I’m not. But I’ve been testing graphics cards for over 15 years. I also knew about the power cable issues with the RTX-40 series. I always made sure the cable was fully plugged in — and it still happened to me!
In case I didn’t make it clear in my previous paragraph, this COULD be a user error. This isn’t a “Ha, gotcha NVIDIA” article. Not at all. We’ve had this GPU for over six months and have used it to benchmark many games. 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. Still, the fact that it could come loose at all, even after I made sure to press it in firmly, is a bit concerning. It’s also inexcusable for this GPU not to have a safety mechanism for when a cable is not fully seated.
We’ve already informed NVIDIA about this, and they are looking into it. A power cable coming loose by itself on a $2K-3K GPU is unacceptable. What’s worse is that a GPU this expensive doesn’t have a safety system to stop it from pulling power if the cable isn’t fully connected. That shouldn’t be happening. Again, this is a premium GPU we’re talking about. Finally, until we get a replacement, we may not have any benchmarks for the RTX 5090.
John is the founder and Editor in Chief at DSOGaming. He is a PC gaming fan and highly supports the modding and indie communities. Before creating DSOGaming, John worked on numerous gaming websites. While he is a die-hard PC gamer, his gaming roots can be found on consoles. John loved – and still does – the 16-bit consoles, and considers SNES to be one of the best consoles. Still, the PC platform won him over consoles. That was mainly due to 3DFX and its iconic dedicated 3D accelerator graphics card, Voodoo 2. John has also written a higher degree thesis on the “The Evolution of PC graphics cards.”
Contact: Email
OMG here we go… 🍿
So does this affect your feelings for Nvidia or the X090s? How so, how much?
It does not really. I assume that NVIDIA would replace this for everyone that would suffer from it (not just us). Still, NVIDIA should have used a safety mechanism for the cable. If, for whatever reason, the cable gets loose, the card should not be drawing power.
There’s a reason I paid the ASUS tax this year and got the Astral with per pin power monitoring. The connector is crap. Also had to replace my Nvidia provided adapter a few weeks after getting it cuz of the load imbalance that I could see thanks to the Astral. The load imbalance existed from day one, mind you.
Same here, got the 5090 Astral LC OC and it has been balanced across all pins for from day 1, 5 months of usage, using the original psu cable from an ATX 3.1 psu. I have pin readout on a gadget on my 2nd monitor (along with normal stats like cpu usage, temp etc) so always have view of it while gaming, and it's solid.
As you just noted this connector is on nearly all ATX 3.x power supplies. That is because it is an Intel design and part of Intel's ATX 3.x power supply standards.
Originally Intel wanted ATX 3 power supplies to have just 2 12 volt 12VHPWR connectors, one for the GPU and the other for the motherboard with all the voltage conversion and regulation on the motherboard side but all the motherboard manufacturers balked. However that is how it's been done for years on server motherboards but they use 24V supplies which lowers the amperage per wire in half even though the total watts remains the same.
Actually at day one they had load balancing (the 3090's that had the connector was split into segments that were load balanced to prevent this exact issue), it's nvidia who instead wanted to push another dollar to the bottom line by bunching together all the 12V leads into one big chunk that can't be balanced in the 40/50 series… and the 50 series have even higher power draw meaning solutions like the 30 series would made sense but no… lets keep the flawed design that kept melting 40 series connectors as it helps our bottom line by playing with potential fire-hazards.
I hope nvidia gets its fingers burned in court and will be forced to replace all their flawed cards – Or they will continue to deliver sloppy overly cost saving engineering.
I would have picked the astral too as it would warn if it went to bad but still just a band-aid for nvidias underlying issue
Fully plugged in you say…
https://uploads.disquscdn.com/images/3f09601a266cc6d41df33f3bd16ef4c699ede9abe8d1d03240845a582db88a22.png
How do they say that these days? Press X to doubt? I can CLEARLY see a gap between the pin and the connector.
That was as much as the cable could go in on our specific RTX 5090. You can also spot this "line" or "gap" at the connector in the video at the 0:12 mark. In the video, we had the RTX 5090 running for over 20 minutes at 100% usage. No burn, no smoke issues. Temps were fine (below 78 celcius with auto fan) and power draw was correct. Assuming the cable wasn't fully seated, the GPU should be smoking in the video, not working perfectly.
In other words, you can't say "Oh, I see a line so it's not fully seated so it burned" but then have a video which disproves this and say "Well, it just didn't happen in the video". This is simply moving the goalposts. By that logic, a devine power came and made the cable burn 😛 . Plus, the connectors that were burned were not those at the top (where you see the "gap"). The bottom ones were burned (and you can see in the image that the bottom of the connector has more contact than the top).
What you can say is that the unit was most likely defective. For some reason, it had that "line" even when the cable was fully seated. If that was the case, then it was only a matter of time before a meltdown occurred. So, it most likely was inevitable.
Still, as I said multiple times in the article, this could also be a user error. It happened so we have an article about it. Simple, really.
So about the "the bottom ones were burned even though they had more contact". If you look up a pinout diagram for this connector, you'll notice that the top connectors (the ones next to the locking clip) are all ground connectors, while the bottom connectors are the ones that carry the actual electricity. So it's entirely logical that it's the bottom connectors that burn.
At any rate, the burning of this connector is related to a horizontal difference in contact, not a vertical difference. It's essentially caused by some of the pins on one side of the bottom row making less contact than the pins on the other side of that row, leading to more current passing through the pins that make more contact, overloading them. And if you look at the way the connector is plugged in, you can see a slight difference in how deeply the left and right sides are plugged in.
So what happened was that one side had slightly less contact, leading to the other side getting overloaded and burning, after which the other side got more contact and burned as well. To be clear: I don't like using the term "user error", because I am of the opinion that a connector on a 2000 EUR consumer product should not be this susceptible to user error. But in this case, I think the term is warranted. Nvidia's official documentation says that there shouldn't be a gap this big between the connector and the port. It is possible for there to be a small-ish gap, but this gap looks massive.
One last thing: you said it didn't smoke the second time. the smoke is the plastic burning up. Pretty much all of the plastic on these pins has already been melted, so there isn't that much more to burn up anymore.
GG about the top connectors not carrying electricity, I wasn't aware of it.
There is still more plastic to burn xD.
The middle pin is intact at the bottom row of the cable, and it has the plastic. Then we have the plastic that is stuck inside the bottom pins of the GPU power connector (which should burn when you have not fully seated the cable).
As I said, it could be a User Error, I'm not excluding it. The thing is I'm always forcing in the cable as much as possible. Next time, I might break the power connector and people will say <<Well, you should not force it as much as you did>>. Decide people, which is which xD.
I have similiar experience on my 5070 ti, I used a lot of force, close to breaking the connector and still wasn`t sure if it was plugged all the way in. Really bad connector design, old ones worked flawlessly.
That's not how electricity works, power flows through both the positive wire and the ground wire. However the hot side tends to have varying resistance while the ground side resistance is close to the same. It's the resistance in the wire/circuits that cause the heating. Most of the time only one or two of the connections burns because those wires were carrying more current than the others were. If the load was balanced across all 6 wire pairs then this wouldn't happen but because the wire resistances (and connector resistances) are all different one or two wires carries more amps. In this case 600W / 6 pairs = 100W per pair …. 100W / 12V = 8.33 Amps per pair. However one wire pair can be pulling only 4 amps while the other is pulling 12.66 amps. Now you 100W rated pin is drawing 152W or 150% of it's rating while the other is carrying 48 watts or 48% of it's rating even though the two added together still equals 200W which is the rating of 2 wire pairs.
Don't feed the trolls. This shouldnt happen at all. Even if it wasnt properly seated, it should protect itself. Thats total BS feeding us since 3090. Thats too many years now to count for sitting on their hands.
PC accessories and parts are installed by anybody, and it should work and not burn down connectors. They might try to say otherwise but thats total bs.
Derbaur also had that problem, and he too knows how to plug them in. So i dunno why people even bring this BS to you. A person that does this for a living. The cable is bad, end of story.
Fully plugged in you say…
https://uploads.disquscdn.com/images/3f09601a266cc6d41df33f3bd16ef4c699ede9abe8d1d03240845a582db88a22.png
How do they say that these days? Press X to doubt? I can CLEARLY see a gap between the pin and the connector.
"a bit bend" , "not fully plugged in" yea yea yea. It's a trash design. End of story. A good connecter would not burn even if it was bent or inserted properly. Just keep buying more you fktards!
It's good that you took a picture of it right before because otherwise people who actually read this article will realize that it indeed is a user error problem
"plugged in"
In the video at the 0:12 mark, you can see the same "line". In that video, the GPU has been working at 100% usage for over 20 minutes. No smoke or burn issues.
You can't say "Well, it was pure luck that everything was working even though the cable wasn't fully seated". That's not how electronics and plastics work. Plus, that top line of connectors was not affected at all. The ones that burned were the bottom ones (which due to how the cable is inserted in the GPU should have more contact than the top connectors).
Still, as I said multiple times in the article, this could also be a user error. It happened so we have an article about it. Simple, really.
Is it possible that becaues the top line isn't fully seated that the bottom line could also not be fully seated due to bad angle? The gap looks so big that the bottom should have a gap too. I remember that JayZ and Debauer both showed that the cable not being fully seated could take some time before it started to smoke and they were being pretty agressive with how much of a gap there was. Anyways, pretty weird.
Exactly. Having an improperly seated cable most likely won't immediately cause a problem, but it will eventually.
sh!eeet.
*runs to check the cable asap*
Congrats for buying $3K bricks that you cant even use as one … j/k , condolences John, I knew as PC Gamer myself, one broken parts that interrupts what should have been perfect chilling afternoon after work is suck as always
Fool me once, shame on you…
same sh*t different day
It's good that you took a picture of it right before because people who actually read this article will realize that it indeed is a user error problem instead of blindly talking trash
https://uploads.disquscdn.com/images/3f09601a266cc6d41df33f3bd16ef4c699ede9abe8d1d03240845a582db88a22.png?w=600&h=412
oops double post
it's the gap isn't it?
So the lesson here is that if there's a gap in the connector no matter how hard you try to plug it in, you should consider the card defective immediately. I just looked at my 5090 and that gap isn't there at all. Not blaming you at all as this shouldn't be a thing, but with the history of this connector I would not accept any kind of gap. That goes for the PSU side of the connector as well.
Not fully plugged in.. limp wrist error unfortunately. Had a 4090 and now have a 5090. No burn but I do check and change or interchange cables after a couple of months I buy from moddiy
I don't check or change anything ever, I just watch pin amps using asus software, and no issues.
Buys a high end product from an anti-consumer company.
Expects the privilege of being treated differently for buying high-end.
This sums up NVIDIA's hardcore fans
Wait a minute… So you're telling me NO MOAR 8K BENCHMARKS for a while???! 😅🤣
Time to spam you with Nude Mods xD
Orrrr RTX remix mods 😀 i love these! I never find them on my own…
Asta vrasta kai xesta Papadopoule!
All in all the 5090 performed really well during the tests especially in the midst of copy and pasting the same template for our conclusion sections.
serves you right for buying overpriced GPU's.
and making Nvidia even more greedy.
I fully agree that the 5090 is priced ridiculously high (thank the AI craze for that) but this is a PC gaming site and part of what sites like these do is bench GPUs for game articles. Are you implying that DSOG should be the only PC gaming site on the internet that refuses to bench the 5090 on principle? That would just cut down on DSOG traffic. Why would any site owner want to cut down on traffic to their site?
LOL. Even if NVIDIA declined to send us a replacement (which it did not, we’ll have a replacement in a week or two), we’d buy one. We can afford to buy 5090s (well, as long as they are not out of stock 😛 ).
And there's the problem. Oh my 5090 burned down, well, I'll just buy a new one. This poor connector design is not costing them money, it's making them more money. Why would they ever fix the problem?
This, and people making excuses for the poor engineering. I hope the gained money will burn up after someone actually challenged that through court, faulty design in many countrys with consumer protection worthy of its name equal prolonged warranty for that isolated fault and perhaps even a mandatory recall for both 40 & 50 series that feature this faulty connector that nvidia blames the user for – All while counting the money gained from the cheap engineering
There is a big difference between a tech website like us (which should cover most GPUs) and an individual PC gamer. As I’ve said before, we mostly buy the PC hardware parts we use. If an individual encounters this issue, they should contact the AIB partner about it. Normally, they should get a replacement. If they don’t, it’s up to them to do whatever they believe with their money. I won’t ever suggest them to “Buy a new one”. Instead, we’ll try to get in touch with the AIB and convince them that they should replace it.
I also said multiple times that they need to fix the design so that the GPU does not draw power when a cable is not fully seated. And then you see people saying, “Oh, it serves you well. You should have fully seated it”. Decide, people — which is which? You can’t complain about EVERYTHING 😛 .
It's a strange thing to me the disconnect some people have with logic. They log onto this site knowing that it is a PC gaming site and expect articles about PC games with benches but seem to be unaware that this is a PC gaming site with articles about gaming with benchmarks.
They think you are buying the 5090 just for personal use and not also for articles about gaming with benchmarks like every other PC gaming site on the internet.
And there's the problem. Oh my 5090 burned down, well, I'll just buy a new one. This poor connector design is not costing them money, it's making them more money. Why would they ever fix the problem?
Of course it melted, every one of these things will melt in the end.
never had a melted connector with the 4090.. 5090 going strong.
Agreed. I consider myself lucky with mine and nothing more.
These connectors are the worst. Completely unacceptable design. So much power going through wires that are too thin. Then throw in unreliable contact point and not being monitored properly with no failsafe… Recipe for disaster.
I hope someone starts a class action lawsuit as this has happened to too many people. What is it going to take to fix this? Somebody's house burning down and someone dying in a fire? Nvidia needs to be held accountable and fix this.
Your law suit would have to be against Intel, it is their design and part of Intel's ATX 3.x power supply standards which can be found in part on Intel's website. Intel has forced both power supply manufacturers and GPUs to use this connector in order to get ATX 3.x certification.
From what I remember from Gamers Nexus videos, NVIDIA originally designed the 12-pin power connector, and what was added to the ATX 3 spec was based on what NVIDIA designed for the RTX 3090 Ti.
The original connector was designed and manufactured by Amphenol and has been in use for many years in even higher power applications (Minitek PWR). Nvidia did not design it, end of story.
The sense pin connector for ATX3/3.1 belong to Molex, who also isn't nvidia last I checked.
Perhaps you could google the words "molex PCIE CEM5 spec" to fix your memory issues.
Of course, if your memory comes from places like YouTube talking heads, you are already starting off with bad info, clearly.
Nvidia has no over voltage protections on the GPU end, so it will draw more power then each individual rail can handle. It's Nvidia being cheap. As usual
4 GB of VRAM are enough, just upscale. you should be happy I'm giving you these 3.5 GB of VRAM!
Who stuck the 12-pin connector and the stupid side-band connector with the sense pins together?
This is really good info thank you, What in your opinion is causing all these faults ?
This is how it looks "fully plugged in" .. connector melted due to limp wrist. https://uploads.disquscdn.com/images/7cc6821bec45cf579bb8b52633074e8a8a9b37e0773d25bc94cb8b68f200aa36.jpg
Show us the insides.
Yes sir, will do. Fully plugged in and no burns https://uploads.disquscdn.com/images/ad5ed91b5ac7016b09f9c4fc7b2159a6671b85a8343170f00855892f2ac922fc.jpg https://uploads.disquscdn.com/images/45ffae50e69e67afdedc66efad7f7c38e3756369ec366df33749d5c06399ee36.jpg
You are complaining to the wrong people. This is NOT a Nvidia design, it's an Intel design that is part of their ATX 3.x power supply standard. Pretty much all ATX 3.x power supplies have this exact same connector on them now so even if you change the GPU side then it will just burn up the power supply plug instead and in fact several of them have burned both the GPU and power supply connector
The problem is even if you fix the GPU side the problem is still there and the only way to fix it is a new connector on the ATX 3.x power supplies which would make all existing ATX 3.x power supplies obsolete.
You are complaining to the wrong people. This is NOT a Nvidia design, it's an Intel design that is part of their ATX 3.x power supply standard. Pretty much all ATX 3.x power supplies have this exact same connector on them now so even if you change the GPU side then it will just burn up the power supply plug instead and in fact several of them have burned both the GPU and power supply connector
The problem is even if you fix the GPU side the problem is still there and the only way to fix it is a new connector on the ATX 3.x power supplies which would make all existing ATX 3.x power supplies obsolete.
The fix is to NOT plug it in limp wrist. Ensure it is "fully inserted".
Nvidia designed their card without any way of load balance monitoring because it saves them few cents per card. Some partner cards have load monitoring and you instantly know if one cable in the connector is overloaded.
Sorry about your card. I guess you will have to use an AMD card or your 4090 for a bit?
The entire RTX 50 series should be using "3nm" (N3P) chips.
That was the original plan for Blackwell but the problem with 3nm is the heat density increases making them hard to cool the chip and they just burn up. If you look around you'll find that 3nm chips all pretty much top out at 150W TDP which is fine for mobile devices (smart phones Apple laptops, etc.) but obviously no where close to the 400+W needed by a high end GPU
This is also part of the reason AMD stopped trying to compete at the high end, they just couldn't make the dies big enough. For example a RTX 5080 is about 20% faster than a 9070XT but uses 14% fewer transistors because the overall design is more efficient needing fewer transistors for the same power. I think that is a major push for UDNA, making it more efficient and by efficient in this case needing fewer transistors for the same amount of power output
This connector and cable is a menace.
This reinforces my view of never using a GPU over 350 W.
This is clearly USER ERROR! You should, obviously, use two elephants to secure power plug. One for 5080. /jk
On more serious note @JohnDio:disqus IMHO You should send GPU to der8auer he has means to check it out.
Nvidia loves one time use graphics cards. 🤪 https://uploads.disquscdn.com/images/b8537e9995815bd269e824cf021e3d4cc6d1c6a9a1c7f16fe4826ef3c58f9ddf.jpg 2025 PC enthusiast gaming bring your own thermal camera. 😵💫
I really I`m not a fan of this connector on my 5070 ti. I`m not afraid it will melt, 5070 ti only used 300w, but the connector doesn`t even click or give any form of feedback when inserted fully, contrary to the old ones did so you were sure it`s all good.
I think this is the video where Actually Hardcore Overclocking on YouTube explains why the power connectors are melting on RTX 4090's and RTX 5090's, but they didn't on the RTX 3090 Ti (which used the same bad 12-pin design):
https://www.youtube.com/watch?v=kb5YzMoVQyw
@JohnDio:disqus The 5090 Astra does have monitoring capability, but thats all it is, it dosent shut down the card in the event of an imbalance on any of the 6x 12v wires.
And the upcoming wireview 2 pro from thermal grizzly is going to have per-pin sensing/monitoring, which if in some wacko universe I paid £2k for a gpu I would almost certainly buy and install.
Its the flawed design nvidia knew was poor engineering already at the 40 series yet kept it in the 50 series and instead turned to blaming the users for their sloppy design. The 3090 that had the connector had its 6 12v leads split in groups of two so they could be load balanced… the 40 series and 50 series both have the flawed "lets save a $ by bunching them all together so they can't be load balanced" design…
The 5090 will have more melted connectors than the 4090 for one reason – Even more power draw through the "cost optimized" connector, hope they get a court case so they will be forced to replace all the fire hazards they have out there
I think it's an engineering flaw in design too. I recall an article on a 4090 on a tech site where they measured the temp of each power wire. The load wasn't anything close to balanced. Some wires weren't carrying much load while one was carrying so much watts that it's temp was out of spec and close to the point of melting the cheap plastic sheathing.
imo if Nvidia is going to push the 12 pin connector then they need to beef up the gauge of the power wires and use a higher quality plastic sheathing. Problem is Nvidia doesn't like admitting to mistakes so they probably won't.
The beta testing never ends with Nvidia.
An Ultra Premium card without any security mechanisms is psychedelic absurdity.
The HW segment is a*s.
The drivers department is also a*s.
With 570 branch, the drivers went downhill in quality in performance.
The only drivers that have some promise are The Alpha Dev 590.26 drivers.
They manage to regain the lost performance, but it's still are producing lower scores, than 561.09, that were released in September 2024.
The 12V-2×6 plug is the new capacitor plague.
John you're probably not going to read this but I'm assuming you swap your card when testing games right? Those crap connectors are only rated for around 20 or so times of being pulled out before they start to fail.
Yea, Nvidia totally sucks for not fixing the issues. It should be fool proof, there should be NO WAY for that cable to burn. No matter what GPU testing u do, no matter how the cable is plugged. There should be a failsafe system. Not plugged? No working/no power/warning. Should be something. As it is.. on a card thats 2500-3500 euro… yeah. TOTAL FAIL.
I totally feel your pain John. I would be scared to do any benchmarking on these cards.
Premium or budget GPU it shouldn't happen.
If you don't mind me asking, which PSU was installed in your system when the RTX 5090 encountered the melting issue?
Corsair HX 1200W. The PSU is irrelevant to the issue, it had nothing to do with it really.
Moral of the story? Everyone knows about the issue, everyone is aware of the problem, yet people still keep buying Nvidia's trash. You want Nvidia to fix the problem? STOP buying their garbage! Once they start feeling a tingle in their wallets, they might actually do something about it. Until then, they happily sell you a replacement.
Nvidia needs a class action lawsuit against it to happen ASAP !
Oh God…I hope they replaced it for free…that's the least they could do!