NVIDIA GeForce RTX 4090 Power Adapter Failures Can Be Blamed on User Error and Foreign Debris, according to New Investigation

Tsing

The FPS Review
Staff member
Joined
May 6, 2019
Messages
11,075
Points
83
User error and foreign object debris are two of the main reasons why some GeForce RTX 4090 owners have ended up with melted connectors, according to a new in-depth investigation by GamersNexus' Steve Burke that shows footage of a 12VHPWR power connector bubbling and smoking alongside one of NVIDIA's Founders Editions cards.

See full article...
 
As much as Steve likes his high horse, this is hardly conclusive. Rambling half an hour about foreign debris inside of the plastic casing, that doesn't even do anything. Then basing his entire conclusion on one example they managed to burn by almost completely unseating the connector. And on the guesswork of some anonymous test lab. Guesswork that doesn't seem to make any sense. What is a high resistance alternative route? I thought electricity will take the path of lowest resistance. As long as there is a lower resistance path for electricity to flow it will go that way. So the only way this works, is if all other pins have no contact while one has a high resistance contact. Hence why they had to almost completely unseat the connector.

Now I'm not saying that this is not what happened to the examples in the wild. But his almighty test jesus is trying to make it sound they performed rocket science.
 
New investigation, if the debris is pieces of tiny copper wire, and they happen to entangle themselves between several pins, it might burn.
 
Complete bullshit.

It's clearly a combination of poor design (dumb port location) and manufacturing issues (adapter cable quality control).

Nvidia has a long history of blaming their problems on others and not taking responsibility.

Anyone remember the Geforce 7800 - 9800 "wrong solder" issues? They completely dumped that **** sandwich on AIB's to handle under warranty, and abandoned customers who were out of warranty.

Nvidia is truly a ****ty company, but they get away with it because of inadequate competition.
 
Complete bullshit.

It's clearly a combination of poor design (dumb port location) and manufacturing issues (adapter cable quality control).

Nvidia has a long history of blaming their problems on others and not taking responsibility.

Anyone remember the Geforce 7800 - 9800 "wrong solder" issues? They completely dumped that **** sandwich on AIB's to handle under warranty, and abandoned customers who were out of warranty.

Nvidia is truly a ****ty company, but they get away with it because of inadequate competition.
Exactly, inadequate competition and add in some level of Apple style cultism. This design is unacceptable, and needs to be phased out ASAP. Unless power supplies manufacturers also want to be caught in the fire, they should announce they are phasing it out, and not supporting it, requiring the user to use adapters if they want to keep things hot. In theory its a 600w plug used still at a fraction (large percentage, but still less) of that 600w , imagine failures at the rated 600w, you would probably have actual fires then.
 
In theory its a 600w plug used still at a fraction (large percentage, but still less) of that 600w , imagine failures at the rated 600w, you would probably have actual fires then.
It's an up to 1200W plug.
 
Nvidia has a long history of blaming their problems on others and not taking responsibility.
Though I agree with you on their practices at times, I don't believe it's Nvidia directly blaming the end user for this as it's been circulating amongst the PC community for some time since the 4090 came out. I even stated it may be user error when I got my card a while ago. Al in all it is a poor design that should have been made better for the less than careful do it yourself PC builder IMO.
 
I have to say that yesterday when I was installing the new PSU I was a bit nervous walking out the adapter from the card. I don't have the full experience as some others here in our forums but definitely have far more than an average user and can say that in my 30-40 years of working with PC equipment the amount of tension needed to get that thing out would've been enough to break pin solder points on a PCB. It all worked out but I was as concerned with that as with everything else we already know.
 
Note the 'sustained' column header. Not expected to sustain more than 600W, but transients are a different issue.
 
Note the 'sustained' column header. Not expected to sustain more than 600W, but transients are a different issue.
Unless you are talking car amplifiers, I think most people are thinking sustained power draw. That's what TDP / TBP refers to now at any rate.
 
Unless you are talking car amplifiers, I think most people are thinking sustained power draw. That's what TDP / TBP refers to now at any rate.
They very well may be only thinking of sustained power draw, but sustained power draw hasn't really been an issue.
 
They very well may be only thinking of sustained power draw, but sustained power draw hasn't really been an issue.
How would anyone know that? AFAIK they still haven't really hit on a consistent mode of failure - just some prevailing theories. And of the failures out in the wild - who's looking at power delivery at the exact moment the cable starts to melt?

And sure, there's always a transient spec - is there a reference for that though? I haven't seen anything on transient rating, or even transients being an issue on this. All the reviews I've seen to date that have a power measurement don't indicate the cards are even getting close to the 600W steady state limit, let alone anything about transients going crazy. Now sure, overclocks all bets are off, but I'm not seeing this failure is correlated with crazy high overclocks either.

 
All the reviews I've seen to date that have a power measurement don't indicate the cards are even getting close to the 600W steady state limit, let alone anything about transients going crazy. Now sure, overclocks all bets are off, but I'm not seeing this failure is correlated with crazy high overclocks either.

This looks like they're still waiting for better equipment; one second measurement times are worse than they'd get from the card itself.
 
This looks like they're still waiting for better equipment; one second measurement times are worse than they'd get from the card itself.
They couldn't measure all the rails and pins individually, but that's millisecond resolution - the 4090 is less spikey on transients than a 3090Ti.

None of the spikes on either card get all that high though - average rating is low 400W's on both cards, the max spike is maybe 15%-20% of the average on the 3090, around 10% on the 4090. Nothing no where near 600W, steady state, transient or otherwise
 
If you've been building PC's a long time... like I have and many here have. I think it's fair to say we've forgotten some of the hurdles and challenges of earlier in our PC build time.

Remember installing SIP's or Cache modules on a motherboard? Especially the cache modules where you had to carefully populate the motherboard directly... and it was pretty easy to screw that up even with the huge *** pins because there wasn't a master pin. Or better all of the stupid *** jumper style pins on motherboards back in the day when you had to individually hook up everything and you could easily get it wrong? Sure it didn't melt... normally. But it was still a PITA.

They have over the years made it a lot easier no lie. No more upside down floppy cables, no more futsing with 40 pin IDE cables where you could toast a drive. Many other advancements... pin blocks for USB 1-3 being one. But seating completely was a thing then two especially with those huge *** molex connectors.

What needs to happen for power is the companies need to take a read of SATA standard connectors and power connectors. Look at how they do it and develop a way to push power through a locking connector of that style. Then you would have a good seating, know it's being seated correctly. And take a big step to eliminating user error on builds.

this is an issue that should only be impacting DIY users. I can say for a fact I would build out my power adaptor and seat it into the card BEFORE I put the card in the case. So I can make sure there is as close to zero gap as humanly possible at this stage. Tempted to get out a C clamp to do it right. lol. But yea... the force needed, the limitations of connections, and everything else makes this a poor design. If I were an SI I would be running these systems 48 hours torture test before sending anything out the door that uses one of these connectors. Just to make sure I don't send a customer something that's going to melt/catch fire.
 
Become a Patron!
Back
Top