News RTX 4090 Owner Hits Nvidia With Lawsuit Over Melting 16-pin Connector

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

InvalidError

Titan
Moderator
But as you said, the suit was filed in California. Orwellian arguments aren't merely acceptable there, they're welcomed. Ignorance is strength. "Your honor, my client didn't bother to read through the document that was right there in front of him. That's why we're entitled to go through all of Nvidia's internal documents."
Yeah, California pushes warning labels way beyond stupid. May as well attach a label saying "may cause death, cancer, choking hazard, etc." to everything.
 

TJ Hooker

Titan
Ambassador
On a side note, we can tell who the hard core nVidia reps are because there are multiple failure modes, only one of which is the Steve Jobs "your not holding it right".


They can burn out even when fully inserted. Not that this would matter to some folks, they'd just assume everyone else is lying.

The lawsuit itself for those interested.

https://dockets.justia.com/docket/california/candce/5:2022cv07090/403711
Realistically, what are the odds that person (or anyone) would admit to installing it improperly, even if that was the case?
 
  • Like
Reactions: cyrusfox

russell_john

Honorable
Mar 25, 2018
115
81
10,660
The 12V and 48V HPWR specs actually come from the PCI-SIG's CEM 5.0 spec and the 12V part got adopted into the ATX 3.0 spec in February this year, well over a year after Nvidia started using it, albeit sans sideband pins.
Only true with the 4 extra sense pins which don't carry power ...... Intel announced their intentions for ATX 3.0 over 2 years ago and their intention was to use this connector to replace the 24 pin connector and the 12V molex connectors on motherboards with this one 12 pin power connector and have all the voltage conversion/regulation on the motherboard itself

https://www.pcworld.com/article/631...upplies-their-first-overhaul-in-20-years.html
 

InvalidError

Titan
Moderator
Only true with the 4 extra sense pins which don't carry power ...... Intel announced their intentions for ATX 3.0 over 2 years ago and their intention was to use this connector to replace the 24 pin connector and the 12V molex connectors on motherboards with this one 12 pin power connector and have all the voltage conversion/regulation on the motherboard itself

https://www.pcworld.com/article/631...upplies-their-first-overhaul-in-20-years.html
That story does not appear to make any mention of when Intel first mentioned its intention of deploying 12+4 12V cables, only that it was revealed in 2022, which is this year.

BTW, while the ATX12VO motherboard connector may look similar, it only has 10 pins which includes 12VSB, pwr_on and pwr_good signal pins, completely different thing. Also, even the ATX12VO v2 spec still specifies the EPS12 connector for high-power CPUs, not the HPWR one.
 

Diceman_2037

Distinguished
Dec 19, 2011
53
3
18,535
On a side note, we can tell who the hard core nVidia reps are because there are multiple failure modes, only one of which is the Steve Jobs "your not holding it right".


They can burn out even when fully inserted. Not that this would matter to some folks, they'd just assume everyone else is lying.

The lawsuit itself for those interested.

https://dockets.justia.com/docket/california/candce/5:2022cv07090/403711

That WCCF article was based on a single source youtuber video who had no evidence of their claims in the video

Anyway, this is a social media stunt, the plaintiff is the founder of High Roller
 
Last edited:
  • Like
Reactions: TJ Hooker

InvalidError

Titan
Moderator
The efficiency benefits are only there under specific power envelopes, so when the power draw gets too high and the PWM hits 50%, that is when these connectors hit a peak power they were never supposed to hit and not designed for.
The HPWR connector is very much capable of delivering 50A. Gamer Nexus chopped off four out of six 12V pins and the connector barely reached 70C under full load, which means that a properly inserted connector can handle nearly 3X its maximum rated continuous power.
 
Though what you say is true, an engineer, who would be required in this case, might also need to examine the test data to prove that there was malpractice.
If Nvidia can prove that no failures ever happened to them in testing and their testing fulfilled the laws statutes, then no matter what the plaintiff says or does, or even the judge, the company would be deemed to have released their product in good faith.
The guilt might actually lie with the middle man, the brand who redesign their boards with custom PCB and silk screen, and add their own connectors.
Note that this connector is an Intel patent, and Nvidia can offset liability to them if it proves the connector is a dud.
Just because in California, a judge can see a precidence, it can still be overturned on appeal or by the Federal Judges, making this even more murky and fudge like.

None of that matters. This isn't a criminal trail so throw all that law drama nonsense out the window, there is no "guilt", no law is required to have been broken, only one side has a claim of damages from another. There will be no "engineer prove X" or some youtube guy making some comment, or even a publicity statement from nVidia being used, none of that matters. Also there is no such thing as "good faith" or "best effort" in a civil case, this is why there are insane warning tags on everything.

The plaintiff claims damages and will have their lawyer make a statement to a jury, of random people, of why they thing they deserve those damages. The defendant will have their lawyers on why they think they plaintiff doesn't deserve those damages, then both sides get to present any evidence supporting their claim while also refuting the evidence presented by the other side. When it's over with, that jury, of random people, go into a room and discuss which side had the better argument and how much damages, if any, should be awarded. The important part about this is these random people don't own nVidia stock, aren't fans and generally don't like large corporations in general. If it becomes a class action, meaning multiple plaintiffs, then it gets super bad for the defendant because it immediately bolsters the arguments of the plaintiffs. Pictures of obvious burning on a consumer electronic with the argument of "they made a dangerous product" do not go over well with jury's. The defense of "hey it wasn't that dangerous and the dude didn't plug it in all the way" really doesn't go over well with jury's. It's why almost every civil lawsuit gets settled long before reaching trial. A settlement gets entered with an NDA attached to not discuss the settlement, the plaintiff drops the lawsuit and everyone walks away.

Also people be thinking that a civil suit is this one sided affair where nVidia says "dude didn't plug it in" and the judge says "oh I guess trial over". Each side gets to attack the other sides arguments, and the jury is already predisposed to disbelieve the big multi-billion USD corporation.
 
Last edited:

InvalidError

Titan
Moderator
And there you go spouting temperature as the only issue here! You neatly forget parasitics due to quick switching which is inherent on pwm systems, and as any engineer who is worth anything knows that phase shifts come as a cost with parasitics when high power is used.
There is no "phase shift" in bulk DC power distribution, only bypassing to compensate for wiring inductance, resistance and source/load transients. There is control loop bandwidth and phase margin on the PSU and PoL control loops side of things to keep them stable regardless of what happens downstream and that is about it for "phase" in DC.
 
  • Like
Reactions: digitalgriffin
Indeed, I agree with most of your points here, I was arguing my point based on how the defense would present their case, not one of certainty.
Criminal liability is not an issue only because corporations gain the ability to not have individuals prosecuted. If Nvidia was not a corporation, this would be a criminal case.

It wouldn't be a criminal case, a person can not bring a criminal case against anyone. The state would have to bring such a case after a lengthy investigation. The standards of both are drastically different, criminal uses a beyond a reasonable doubt, meaning the facts must be taken in a light most favorable towards the defendant, civil causes use preponderance of evidence which favors the plaintiff. If we were to use percentages, criminal cases require 99% certainty that the defendant is guilty, civil cases would be the defendant has 51% of fault, as determined by a group of random people.
 
  • Like
Reactions: helper800
My thoughts exactly. I think he is going to have a hard time with this lawsuit, not like his house burned down or he was immolated.

Free month of GeForce now and a bag of Doritos. I sure hope he found a pro bono lawyer, otherwise the master computer builder is going to have a lot of lawyer fees.

There's material damages and punitive ones. The later is sometimes used to punish a company to remove incentive from being careless again.
 
There is no "phase shift" in bulk DC power distribution, only bypassing to compensate for wiring inductance, resistance and source/load transients. There is control loop bandwidth and phase margin on the PSU and PoL control loops side of things to keep them stable regardless of what happens downstream and that is about it for "phase" in DC.

I wonder if he is referring to inrush current between phase switches and inductive losses on load lines.
 

InvalidError

Titan
Moderator
I wonder if he is referring to inrush current between phase switches and inductive losses on load lines.
Smoothing out switching and load transients so power distribution only needs to deal with near-DC is why boards have a crap-ton of supply bypass caps on them, usually with an inductor between the local bypass and input cables to reduce the amount of EMI ('AC') back-feeding into the wires.