Streamer's RTX 4090 catches fire while live on Twitch, gamer asks viewers "WTF do I do?"

Jimmy2x

Posts: 255   +29
Staff
A hot potato: The melting 600W power solution for the RTX 4000 and 5000 series is anything but a new problem, and anyone in the PC hardware space has become all too familiar with images of the aftermath. Some tech channels and testing specialists have even created videos intentionally demonstrating the burning scenarios. On Friday, one unlucky gamer managed to capture the hot event live on stream using nothing but her camera and a game of Marvel Rivals.

Before we go one step further, let's drop a friendly TechSpot Public Service Announcement: if you're using your PC and your PC catches on fire, you don't need to ask anyone if it's bad. Just shut down your PC. Immediately. Really, nothing is going to get better from that point forward.

A streamer by the name of jessick was in the middle of a Marvel Rivals match when she suddenly realized her game was going up in smoke. As seen in the video below, the shielding on one of her RTX 4090's 16-pin power connector wires unexpectedly began to melt and burn away. And like any good streamer looking to capitalize on content regardless of the cost, jessick didn't take immediate action. She instead grabbed her camera, streamed it, and asked her viewers "wtf do I do?"

While there's plenty of information to make a case against the 600W power connector's design, one of the big questions is always "who is at fault?" Is it the less-than-desirable design of the 12VHPWR and 12V-2x6 connector? The quality of the connector at the GPU, PSU, or cable? The user or integrator's installation of the cable itself?

Based on the video, we see the cable connected to an RTX 4090 MSI Gaming X Trio. The card's specs identify a 450W total graphics power (TGP) and call for at least an 850W PSU. We can't see the PSU, but what we can see is the cable itself. In the video, the cable can be seen looping somewhat tightly around the underside of the card.

While this makes for very tidy cable management, it may create some connectivity concerns based on input from PSU manufacturers such as Seasonic and Corsair. Cable bends that are too extreme can cause loose or incomplete connections, creating a scenario that could induce catastrophic failure.

According to Seasonic and other industry experts, users should avoid sharp bends and ensure at least 35-40mm of straight, unbent wire is left at the connector before initiating the bend. A better solution, in many cases, is to use a properly designed 90-degree cable from a reputable or specified vendor.

Is there enough evidence in the video to really say who is at fault? No. It could be jessick, MSI, the designers and engineers from PCI-SIG, or the actual cable manufacturer. What we can say is this: if your computer bursts into flames, do the right thing and take steps to contain the issue instead of first leveraging it for more clicks. Trust us, fire is bad.

Permalink to story:

 
600 watts was always optimistic. So I'm going to give nVidia the benifit of the doubt for a moment just to make a point. Something I gave personally seen in the construction industry is electrical cable sold as one thing and it ends up being another, but it's a known issue and anytime we open a new spool of cable we have to verify that it is, infact, what we bought. We have a guy dedicated to inspecting and testing materials as they arrive on site as you can't count as the guys doing the work to test it them selves. While I don't have exact numbers, I hear about bad wire showing up about once a month and that's just me hearing about it.

That said, while the cable *should* work perfectly(I still don't believe 600 watts) if everything is to spec, these guys could get a bad batch of cable and that ends up getting sent to a consumer causing the fire. The counterfit wire problem is big enough that nVidia should have accounted for it in their spec. They should have lowered the rating or over built the connector
 
I wonder how long will They deny It's a poor design and blame everything other.
I'm pretty sure classic connectors happen to fail, too, but no mention in a public dispute, so It must happen in much lower numbers. Any half-witted engineer would already pointed that out. How much nVidia is paying to supress common sense in todays media? We know They pay Bloomberg.
 
That said, while the cable *should* work perfectly(I still don't believe 600 watts) if everything is to spec, these guys could get a bad batch of cable and that ends up getting sent to a consumer causing the fire. The counterfit wire problem is big enough that nVidia should have accounted for it in their spec. They should have lowered the rating or over built the connector
I don't know why they didn't just over-engineer it a bit for futureproofing. Thicker and fewer cables for high-power loads sounds like it should've been the direction they should've gone.

Sure, a bit more copper per cable, but man, after the 40 series was having problems, they should've been smart enough to rethink their "fragile" plug (knowing the next gen of cards were going to draw even more power).
 
I wonder how long will They deny It's a poor design and blame everything other.
I'm pretty sure classic connectors happen to fail, too, but no mention in a public dispute, so It must happen in much lower numbers. Any half-witted engineer would already pointed that out. How much nVidia is paying to supress common sense in todays media? We know They pay Bloomberg.

Classic connections like the PCI-E 8 pins or even 6 pins are not prone to catching fire. This is just a failed design and fire's are waiting to happen. The PCI-E 8 pin can take 25 removals and insertions, this seems so crap it will start doing this after less then 3 installations.

 
See this is where gamers n streamers get mixed up. A gamer knows their pc very well. A streamer, well they can barely find the power button without help.

A gamer knows damn well they aren't running a 600w psu for a 4090. Most would say are you stupid.
A streamer would say close/good enough for a 600w. And in most cases they could be right but not with a powerful gpu.

Anything with since the rtx xx70 and higher I'd go with 800w or higher. With my 1070 I've been using 850w for like 6 or 7 yrs, before that was still running 800w psu.
Also dont forget that its not just the gpu but the whole system that needs power, some people do forget that.
 
Sure, a bit more copper per cable, but man, after the 40 series was having problems, they should've been smart enough to rethink their "fragile" plug (knowing the next gen of cards were going to draw even more power).
Sure.
They rethinked it.
From 450W to 600W.
Just so they could go with only one connector and no load balancing.
 
Anything with since the rtx xx70 and higher I'd go with 800w or higher.
Those numbers are recommendations for people using cheap noname bunch of wires instead of PSU.

I have run 6800xt, 7900xt, 9070xt with Ryzen 5800x/5700x3D/5800x3D using 660W PSU.
It works just fine even with 560W PSU.
80+ Platinum.

But ... cheap PSU and hungry GPU does not mix well.
 
Bet a years salary that the dimwit Twitch'er did something to cause it to happen to get da clicks, just like what is always the case.
 
Bet a years salary that the dimwit Twitch'er did something to cause it to happen to get da clicks, just like what is always the case.
Shes not using the stock power cables from the PSU, so yep. Right off the bat we have a failure.
I wonder how long will They deny It's a poor design and blame everything other.
I'm pretty sure classic connectors happen to fail, too, but no mention in a public dispute, so It must happen in much lower numbers. Any half-witted engineer would already pointed that out. How much nVidia is paying to supress common sense in todays media? We know They pay Bloomberg.
I wonder how long people will claim this connector is bad until they post some actual numbers showing the connector fails worse then the 8 pin did, then acknowledge how many of those failures are using either cable squids or non manufacturer cables or extensions for aesthetic reasons.
 
Miss hotty girl steamer asks: What to dooooo0o0, my house is buring guys? I see fire questionmarks?? whaat?
Cringe factor 67%... it is funny tho. Not her or the comment she made.. The game she is playing. Valorant burning a 4090 lol. It's not even a demanding game. Gotta love the cables. Nvidia is the biggest scammer company in the world atm. After Asrock. Even my X3D CPU has issues with their motherboards. Anyways. My 5080 cable is very hot, I'll probably have the same issue eventually. Should I make a vid too? AMAGAD!?! What to do boyz pls pls hellp me I see fire!!!!!!!!!!!!!!!!!!! (help!)
 
User issues aside, the connector and the wiresize alone are an issue in these cables.
I’ve done a summary calculation in an earlier Tech Spot article showing the marginal safety factors of this design.
 
Nvidia's Proprietary Fire Hazard
They are trying to run 600W through 6 wires. It gets unbalanced and instead of pulling 8.5A per wire, 1 wire ends up pulling 24A. This was tested and reviewed by Der8auer

It is a design issue. They call it a "16pin connector" to distract from there being only 12 pins for load. If they didn't, you'd realize that a 6pin PCIe power cable only pulls 75W for safety, 2 of those would pull 150W across 12 pins, and they are trying to pull 4x that thru their Proprietary Fire Hazard Power Connector.
 
Who should we sue if our house burns down due to a fire like this? Is the liability on Nvidia's spec or on the manufacturer of the device? Why has no one gotten to the bottom of this.

If everyone KNOWS THIS why would hardware reviewers not award the solution 0 starts/points/percent, due to the imminent risk of fire?

Have we started hearing of datacenters on fire? Sounds like the kind of thing that will also happen.

Long time since I heard of a class action lawsuit against big tech, and it's about time people woke TF up.
 
Sorry, but with this whole fiasco it is never the users fault. Why? Because expecting the end user to compensate for design flaws isn't good policy. It's been shown by numerous people that the contact for the small pins isn't sufficient even with the last revision. Without sufficient contact the other pins attempt to compensate for the GPU power draw and overheat. While not fully seating the plug will create a greater chance, this even happens with full and proper insertion of the plug. It's simply a bad design, and nothing more.
 
Those numbers are recommendations for people using cheap noname bunch of wires instead of PSU.

I have run 6800xt, 7900xt, 9070xt with Ryzen 5800x/5700x3D/5800x3D using 660W PSU.
It works just fine even with 560W PSU.
80+ Platinum.

But ... cheap PSU and hungry GPU does not mix well.
That's true. But you are using one of the highest kind of power supplies. You think any streamer knows the difference between bronze, gold or platinum. They dont. But gold is usually all you need in most cases.
 
Back