News Nvidia's RTX 5090 power cables may be doomed to burn

[YSCCC]

Just because planes can technically take off with half of their engines disabled doesn't mean they should since that leaves no reserve power in case of bird strike or other engine failure during take-off.

Just because a good HPWR cable should be fine (stay within thermal specs) with half as many wires/pins doesn't mean you should intentionally run it that way either since that leaves no room for externalities.

Just because on paper the cable can handle the current doesn't mean that they should not monitor the current per pin for safety and product protection🙄

 

[InvalidError]

Just because on paper the cable can handle the current doesn't mean that they should not monitor the current per pin for safety and product protection🙄

And the pointless argument circle goes around yet again. There is no actual safety issue, good cables are balancing current perfectly fine, additional "product protection" is an unnecessary expense.

I am 100% confident that Nvidia did the math on how much out-of-pocket cost they may have to cover for melted connectors and concluded they were perfectly fine with it. They have a fiduciary duty to optimize value for the shareholders and wouldn't melt connectors if the projected frequency and costs wiped out the savings.

 

[YSCCC]

And the pointless argument circle goes around yet again. There is no actual safety issue, good cables are balancing current perfectly fine, additional "product protection" is an unnecessary expense.

I am 100% confident that Nvidia did the math on how much out-of-pocket cost they may have to cover for melted connectors and concluded they were perfectly fine with it. They have a fiduciary duty to optimize value for the shareholders and wouldn't melt connectors if the projected frequency and costs wiped out the savings.

Of course they did the calculation, they calculated that if they put the balancing circuit or protection in they will have a lot of "the card is not operating properly" bad press coz the card is stopping to pull full power due to the lack of overhead and poor design of the 12V 2X6, so they ditched it and shove it to user error if anything happens, blind fans will just blame the user or cable or PSU or whatever but not Nvidia, and here you go.

Your arguement is ciruling around the faith in Nvidia team, which, Boeing have taught us that being the trillion dollar company, even if life in hand they still do the same stupid thing to maximize the stock price/profit.

When they can easily add extra $50 to cover the cost of the protection circuit and ppl will still buy those, that arguement they don't need it but need profit is making zero sense. Plus that doesn't make them a responsible company anyway.

 

[InvalidError]

Jayz's $0.02 just posted a video about 100 plug-unplug cycles on both leaf-spring and split double-dimple cables. No problems with either, current distribution actually becomes more even after the first 40 cycles instead of worse as the hypothesis behind the exercise was.

View: https://www.youtube.com/watch?v=lAdLOf5of8Y

 

[YSCCC]

Jayz's $0.02 just posted a video about 100 plug-unplug cycles on both leaf-spring and split double-dimple cables. No problems with either, current distribution actually becomes more even after the first 40 cycles instead of worse as the hypothesis behind the exercise was.

View: https://www.youtube.com/watch?v=lAdLOf5of8Y

So this just means that the theory of it being the plug cycles is leading to the melting doesn’t hold, it’s the whole plug design causing the melting issues, small undetected manufacturing defect and lack of load balancing with lack of overhead is likely the issue

 

[InvalidError]

So this just means that the theory of it being the plug cycles is leading to the melting doesn’t hold, it’s the whole plug design causing the melting issues, small undetected manufacturing defect and lack of load balancing with lack of overhead is likely the issue

Small defects don't cause two wires to end up carrying all of the current. Small defects and normal manufacturing variability cause mostly negligible variance like what techtubers see whenever they try new cables and measure between plug cycles.

 

[YSCCC]

It's bad plug design and the spec, nothing more, period.

 

[Moonstick2]

In countries like Canada that take licensed engineers' role in preserving public safety seriously, licensed engineers have to notify authorities about any obvious safety concerns regardless of their relationship with any standards or companies.

If HPWR was deemed unsafe by licensed engineers...If safety margins were only 10%, cable...

Lets not confuse things here. There's safety when referring to danger of injury or serious property damage, and there's safety (in my country at least) used in far more general terms, like "it's an hour's drive, but lets leave quarter of an hour earlier just to be safe". Nobody (here) would think that means "otherwise we might die".

As has been pointed out, these things aren't going to burn houses down. "Safety margin" (or "Safety factor" as used on the Wikipedia page) doesn't always mean "it's physically dangerous to exceed this". In this case it means "it's likely to fail if it exceeds this". 12VHPWR was paper specified with a far lower factor than was typical (as if they started with the required wattage + cable design and worked backwards) and appears to be colliding with reality.

But because all that happens is a melted connector and a defunct $2000 graphics card, the authorities aren't really going to be that interested. Nobody's going to die, there's no danger to public safety.

 

[Moonstick2]

I am 100% confident that Nvidia did the math on how much out-of-pocket cost they may have to cover for melted connectors and concluded they were perfectly fine with it.

Thing is, they were still putting in shunts to treat the incoming power lines separately on the 3090, after 20 years experience making graphics cards.

Then, apparently, while designing the 4090, that's when they suddenly worked out "hey, we don't need to do this" and effectively just twisted all six power lines together at the end. That it happened at the same time as introducing a new connector standard with a safety margin of 1.1 vs the previous 1.68 while pushing the power draw higher than ever before, and that graphics card connectors started melting left right and centre, was just an unhappy coincidence. Doubly so when it immediately happened on the next generation too.

I'd be interested to know if they've removed all the shunt stuff from their other cards too, like the 4060, or the H100.