Why are we not holding Nvidia accountable for putting fire starters in millions of homes?
Oh right, there are no adults left in the room anymore.
this post was submitted on 29 Aug 2026
86 points (97.8% liked)
PC Master Race
22260 readers
414 users here now
A community for PC Master Race.
Rules:
- No bigotry: Including racism, sexism, homophobia, transphobia, or xenophobia. Code of Conduct.
- Be respectful. Everyone should feel welcome here.
- No NSFW content.
- No Ads / Spamming.
- Be thoughtful and helpful: especially when new beginners have questions.
founded 3 years ago
MODERATORS
We are the adults in the room...
They would have been so much better off using a 2 wire power connection like they use for car subs than this. Just tighten down the wires.
Nv:"Oh we made a connector that can peak at 600w" Everyone: "what about any safety leeway?" Nv:"What's that? Oh, AI Didn't say we needed that"
It's not really the wattage that's an issue.
If it ran at 600 volts, it would only be 1 amp.
But at 12v, it's 50 amps of current.
Which is a lot.
And it's not massively peaky/bursty, so it needs to withstand that constantly.
Pretty sure something like 10-15mm² cross section of copper is needed for that, which is a thick cable.
But, this is also a problem that has been solved elsewhere. For whatever reason, manufacturers have settled in a terrible solution.
For whatever reason, manufacturers have settled in a terrible solution.
As I understand it, not "manufacturers," Nvidia specifically. They designed the connector then basically forced PSU makers to follow because they're 90% of the market.
“Inside an Alienware Aurora”
Well, unless they’ve changed, Alienware was kinda famous for basically making a sealed box with a stock CPU fan and crossing their fingers as the machine ran 100C at all times. I don’t know if that could lead to fire instead of just failure, though.
100C / 212F is the temperature water boils at. You’re going to need at least twice that to combust something thin and dry, like paper. The wave soldering used in assembly of these components exposes them to temperatures exceeding 250C / 480F.
Silicon degradation takes place at sustained temperatures in excess of 120C.
There are plenty of notebooks that, by design, can boost the CPU up against its thermal limit (usually 95C or 100C) so as not to leave potential performance on the table.
On the topic of the article, operating temperatures of components like the CPU and general cooling of the system are wholly irrelevant.
The issue is that the 12VHPWR specification does not include the means to balance electrical load across the 6 VCC wires that carry power. Electrical current follows the path of least resistance - if that happens to be one wire over the others it may overheat, melt that wire’s insulator, melt the adjacent wire’s insulator, and eventually melt an adjacent ground wire’s insulator. VCC comes into contact with GND, a ton of current gets dumped into the shorted circuit, and there is a chance of a combustion event in the milliseconds before the PSU’s Over Current / Short Circuit Protection measures can kick in and cut power.
EDIT: Hadn’t read the article yet. It seems that this particular case involved a failure on the FBVDDQ (VRAM) line, not the 12VHPWR connector. The facts that this happened to a 5090 in an Alienware system are largely irrelevant, the most probable point of failure is a power stage or MOSFET which can and do fail in any laptop or graphics card.
I say largely irrelevant because the 5090 does have a TON of VRAM, it is possible that Inno3D cheaped out on a power stage / FET that is not up to the current demands of that much memory. If that were the case, though, I would expect to be seeing / hearing about widespread issues with this particular card in GPU repair communities.
Kind of a feature these days. Shit amp distribution. They just cook a pin and woof! Plenty of tests showing it and plenty of dead 5090s every other day.
Not a smart card to buy unless you triple check connections and aren't gonna touch it. But even then, too risky for me. Especially at the ludicrous price.
This article wins the
- more questions after reading it
- most useless pictures that showcase exactly nothing (photographing the burnt GPU from below was to hard I guess)
- most annoying popups
awards
We didn't start the GPU fire!
Urgh and I'm sitting here downltairs grilling my RTX3060 upstairs via ssh ...
If it calms you, I think it'd be hard to start a fire with a 3060 even if you wanted to.
Wow, a burning card not because of 12vhpwr‽