News Nvidia Calls For Melted 4090 Cards To be Returned for Analysis

TechieTwo

Notable
Oct 12, 2022
234
209
960
What a cluster for consumers. You'd think extra attention would have been paid to the design and production of the adapters knowing the power load they would be handling.
 
  • Like
Reactions: JoBalz

InvalidError

Titan
Moderator
If you are going to send 50A down a bunch of #16 wires and tiny pins, you'd better have some form of active per-pin current balancing (ex.: 3-4 Vcore phases hard-wired to each individual socket pin) to make sure all pins and wires are contributing roughly equal current under all circumstances and that any excess droop on any pin which would indicate a bad wire or connection gets reported to the user so it can be fixed.

Just wait until NVidia has to buy customers a new house because of this
As long as Nvidia can demonstrate that its own GPUs and adapters are made of materials that meet the 94V-0 standard (self-extinguishing materials) then Nvidia's own liability may not extend much beyond components directly damaged by the GPU and cable before they could put themselves out.
 

Exploding PSU

Honorable
Jul 17, 2018
461
147
10,870
And a signed leather jacket from jensen himself!

Where do I sign up!?

ON a more serious note, after watching and reading some stuff that claims the reason why this happen is in part due to the connector being bent, I just realized how bent the cables that connect to my GPU are. Man, those cables are rock solid, it's a miracle I haven't lit my PC on fire all these years.
 

YouFilthyHippo

Prominent
Oct 15, 2022
168
84
660
If you are going to send 50A down a bunch of #16 wires and tiny pins, you'd better have some form of active per-pin current balancing (ex.: 3-4 Vcore phases hard-wired to each individual socket pin) to make sure all pins and wires are contributing roughly equal current under all circumstances and that any excess droop on any pin which would indicate a bad wire or connection gets reported to the user so it can be fixed.


As long as Nvidia can demonstrate that its own GPUs and adapters are made of materials that meet the 94V-0 standard (self-extinguishing materials) then Nvidia's own liability may not extend much beyond components directly damaged by the GPU and cable before they could put themselves out.

At the very least there will be a class action lawsuit, unless the recall happens soon. The recall is coming. Its only a matter of when. I give it a couple of months, max. AMD is going to have a field day with this dumpster fire at their keynote. "Don't worry gamers, you will never have to worry about your GPU catching fire". Shots fired!!!!
 
  • Like
Reactions: daworstplaya
At the very least there will be a class action lawsuit, unless the recall happens soon. The recall is coming. Its only a matter of when. I give it a couple of months, max. AMD is going to have a field day with this dumpster fire at their keynote. "Don't worry gamers, you will never have to worry about your GPU catching fire". Shots fired!!!!
I wouldn't go as far as you did with the language used, however, no matter the merit people of my ilk (Citizens of the USA) are famously litigious. You see people sue companies for significantly less. I would not doubt a lawsuit in the future if enough people are impacted by this issue.
 

InvalidError

Titan
Moderator
At the very least there will be a class action lawsuit, unless the recall happens soon. The recall is coming. Its only a matter of when. I give it a couple of months, max.
The only thing that I can foresee getting recalled is straight HPWR cables and adapters since they are practically unusable in just about any reasonable PC case without bending. Nobody wants their GPU power cables sticking out straight into the side panel in a normal slot, straight up in a bracket mount with riser where it gets in the way of accessing everything else or straight into the front panel for GPUs with power connectors on the end.
 

plateLunch

Honorable
Mar 31, 2017
89
29
10,560
Someone needs to go back to the pin manufacturer and verify their specs and production processes. When the pins mate, there should be a specified max contact resistance. There should also be some max variation in the resistance. With all the power wired in parallel, current should distribute itself automatically across all the pins. The problem is if the variation in contact resistance is too high, the low resistance pins will be asked to carry a lot more current than the higher resistance pins. This is where the overheating will occur.

It is important to understand it is the relative resistance of the pins that is important, not the absolute value. Two pins may have a spec contact resistance of <20 milliohms. But if one has a contact resistance of 20 mΩ and the other has a resistance of 10mΩ, the pin that has the lower resistance will be currying twice as much current as the other pin. And when we're talking about little pins and big currents, that can be a problem.
 
  • Like
Reactions: JoBalz

InvalidError

Titan
Moderator
Someone needs to go back to the pin manufacturer and verify their specs and production processes. When the pins mate, there should be a specified max contact resistance. There should also be some max variation in the resistance.
I predict any such 'audit' will merely find out that the connectors do indeed meet or outperform all of their specs through the whole 30 insert-remove specified service life when assembled and installed in accordance with the manufacturer's procedures. I strongly suspect the failure is with the application of the connector, not the connector itself.
 
  • Like
Reactions: Devoteicon

ManDaddio

Reputable
Oct 23, 2019
99
59
4,610
Melted cards? The title is a little misleading don't you think?
My guess is either people were overclocking the cards or they didn't plug them in properly.
Or it could be the kind of power supply they're using.
I don't know for sure but that's my guess.
 

JoBalz

Distinguished
Sep 1, 2014
101
42
18,640
What a cluster for consumers. You'd think extra attention would have been paid to the design and production of the adapters knowing the power load they would be handling.

I have to wonder if they simply tested the adapters under ideal conditions, not under conditions that actual owners would potentially put them through...overclocking, cable management involving significant bending, etc. If I understand what I've read correctly, the melting occurs after the cards have been in use for hours to days. Since they have been only recently released, it's possible that the cases of melting currently reported are just the tip of the iceberg.
 

InvalidError

Titan
Moderator
Since they have been only recently released, it's possible that the cases of melting currently reported are just the tip of the iceberg.
Since the cable dissections show that it is very likely the problems are due to the copper strips wires are soldered to ripping apart, this ripping can occur gradually: continuous forces from gravity, hitting an obstruction or other wiring-induced forces, vibrations, etc. can initiate a tear and slowly rip a piece apart with thermal cycles.
 

warezme

Distinguished
Dec 18, 2006
2,450
56
19,890
My MSI 4090 came with a 3 prong adapter and I was a little taken back it wasn't a 4 prong. Now I'm glad plus it's 450W card and I don't plan to overclock. Still runs just as fast as it needs to run.