Reports of Bricked NVIDIA GeForce RTX 5090 and RTX 5090D Surge

[LittleBro]

This is such a stupid non-argument. Of course NVIDIA tested it with PCIe 5.0 motherboards, otherwise every man and their dog would be complaining. The fact the latter is not happening should, but apparently hasn't, clue you into the fact that maybe, just maybe, these isolated reports are either user error or incompatibility issues with specific motherboards and their firmware.

Oh, really? And your post is such a valid argument, right? Because of the lack of information on this problem, we can only provide opinions and speculations, just like you did in quoted text above.

You even forgot about the possibility of driver issue to be a culprit here.

No doubt there are hundreds, maybe thousands of different motherboard SKUs with a PCIe Gen 5 slot and each board could have many BIOS updates, but how many motherboards does NVidia test before they release a GPU driver update? Do the problems occur mostly on motherboards with an out-of-date BIOS, several years old or is it a hardware problem? I bought my mobo in November 2022 and I'm wondering if I should pass on a PCIe Gen 5 GPU upgrade, even if I update the BIOS.

Keep in mind that PCI-Express is a standard. BIOS is not meant to be updated on regular basis every time a new PCIE expansion card is released. Expansion card is not a processor.

BIOS - if it ain't broken, don't touch/fix it. Same applies to GPU BIOS.

As for your wondering, I had been using PCIE 2.0 board with PCIE 3.0 GPU for years, even my previous PCIE 3.0 board had BIOS version from around 2019 and I ran PCIE 4.0 GPU from 2022 on that board just fine. I only update BIOS when there is security patch involved, when there is added support for newer CPU I'm going to switch to, or when they fix performance-related issue that possibly negatively affected my system (memory speed support - though I do a thorough research here, because sometimes it turns out to be opposite - performance does not get actually improved).

No Gen 5 GPU products even existed up until now, so on field deployment it's bound to run into issues. Gen 5 PCIe testing equipment in labs is one thing, real world usage likely another beast altogether.

Are you implying that it is normal to cause failures when a first product that supports a new technology arrives? This should have been ironed months before date of release. It's not like they were supposed to do start from a scratch here. PCIE Gen 5.0 standard is known for years. It has exact specifications which are required to be fulfilled in order for a device to be compatible with the standard.

We still don't know whether this problem is related to PCIE 5.0 only. In the article, it is mentioned as one of thought possible causes.

 

[AVATARAT]

I'm not sure why, but I think in the last few years large corporations have been using end users as beta testers for everything software and hardware. Am I wrong?

 

[bug]

I'm not sure why, but I think in the last few years large corporations have been using end users as beta testers for everything software and hardware. Am I wrong?

Wrt hardware, if it was that bad, you'd see people stop buying, class-action law suits. When you make hundreds of thousand or millions of any product, there will be defects. The defect rates are probably in the low single digits. We just have more channels to make noise about that, that's all.
And keep in mind defect rate goes up naturally as the products get more and more complex.

 

[RootinTootinPootin]

I'm not sure why, but I think in the last few years large corporations have been using end users as beta testers for everything software and hardware. Am I wrong?

spot on!!

 

[Prima.Vera]

Nice.

 

[Dr. Dro]

Are you implying that it is normal to cause failures when a first product that supports a new technology arrives? This should have been ironed months before date of release. It's not like they were supposed to do start from a scratch here. PCIE Gen 5.0 standard is known for years. It has exact specifications which are required to be fulfilled in order for a device to be compatible with the standard.

We still don't know whether this problem is related to PCIE 5.0 only. In the article, it is mentioned as one of thought possible causes.

I don't disagree, but it remains that it's the only Gen 5 graphics card design (AMD won't support Gen 5 on 9070 series, though that doesn't really matter for it), so we've had this technology motherboard-side for years (since 2021/Alder Lake) that basically shipped with just lab testing and no real world field deployment. I definitely expected some bumps on the road.

We'll have to wait and see, I suppose.

I'm not sure why, but I think in the last few years large corporations have been using end users as beta testers for everything software and hardware. Am I wrong?

You're not wrong.

 

[Prima.Vera]

Wrt hardware, if it was that bad, you'd see people stop buying, class-action law suits. When you make hundreds of thousand or millions of any product, there will be defects. The defect rates are probably in the low single digits. We just have more channels to make noise about that, that's all.
And keep in mind defect rate goes up naturally as the products get more and more complex.

What hundreds of thousands, what millions are you hallucinating??
It seems that a very big % of the 100 or less RTX5090 cards that were scalped, it's affected.

 

[mkppo]

It's not only the 5090. Roman had issues on his 5080, and trust him of all the people to know a thing or two and not be using a cheap motherboard. Dude even tried to troubleshoot it. Quickly pointing the fingers at motherboard manufacturers isn't the best thing to do tbh. He's not the only reviewer having issues either..

A very disappointing launch really, I expected more from them. It's probably the most disappointing launch i've witnessed in a good while.

Edit: Here's your chance on a silver platter AMD. You ramped production at the same time at 50 series but didn't have a joke of a launch with hundreds of products. Iron out bugs, do a proper launch, price it well and actually have stock. Judging by their track record, something should be amiss but let's not hope it's all of those points.

 

[dyonoctis]

At least Patty Cakes was fired for that and probably other things. The big leather jacket will probably get the medal of freedom.

Well Intel is a "faceless" company, so old that the founders are already dead. Kicking Jensen out, the founder, and the man who's strategy made the company so sucessful in the first place would probably made investors more nervous than waiting to see how they fix the problem.

Apple kicked Jobs after two commercial fail. The dude went back in shorts and tongs after the supposed better leadership nearly ran the company to the grave. In the current datacenter focused market, Jensen probably got as much aura as Jobs had for the mainstream market (The man is now being interviewed by mainstream vulgarisation tech channels). Nvidia without Jensen sounds scary If you have money invested.
NVIDIA CEO Jensen Huang's Vision for the Future

 

[bug]

What hundreds of thousands, what millions are you hallucinating??
It seems that a very big % of the 100 or less RTX5090 cards that were scalped, it's affected.

He was talking about hardware, in general, being released in beta form. I was answering to that.

 

[mkppo]

I'm wondering how reports have surged when production itself hasn't surged. But erm..having reports with this little of quantity does not bode well. Let's hope they get it sorted, especially the FE issues.

 

[ThomasK]

This irrefutably proves that rtx 5000 was rushed, just so nvidia had something to show at CES.

I was hoping the fanboys would begin to shut up by now, but no, they'd rather defend the trillion dollar overvalued company, which HAS the resources, but cares so little about gamers, that managed to "launch" an overpriced, unavailable and failing product.

 

[Crackong]

You forgot:
2000 Series - Teaches people the "importance" of Ray-Tracing. Lessons continue to this day.

At least Ray-tracing does not brick your card...

 

[bug]

This irrefutably proves that blackwell was rushed, just so nvidia had something to show at CES.

I was hoping the fanboys would begin to shut up by now, but no, they'd rather defend the trillion dollar overvalued company, which HAS the resources, but cares so little about gamers, that managed to "launch" an overpriced, unavailable and failing product.

"This" meaning what? Do you have a number of the failing cards? A number of cards being sold?
Because if you look at this objectively, one of the reports posted literally says "all 5090 and 5090D have exploded". Are we not allowed to take that with a grain of salt?

 

[DigitalMaikal]

Just registered here to post. I was using 572.16 on my RTX 4080, but hadn't gamed at all until today. This morning while playing a game I had a black screen. I could not get a video signal when trying to reboot. After doing the basic trouble shooting steps and reseating the card/power multiple times I eventually got a video signal. When the bios splash screen came up the screen went black loss of video signal. Eventually got signal back (reseating the card etc doesn't seem to really help. Just sometimes it will "work" for a bit and others not at all). Anyway I got a signal back long enough to get into windows. Do a search see all the crashing etc going on with 572.16 and I rollback to the previous driver. Reboot have a signal and everything works great for about 10 minutes then black screen loss of signal.

I bought my Galax 4080 SG in August of 2023 and haven't had any problems with it until today. Anyway at this point I hook a hdmi cable up to my Asus x670e Hero and reset the bios with the button because I had the IGPU disabled and couldn't see to get in and enable it. So I boot into Windows 11 with the IGPU and device manager is seeing my RTX 4080 and reports it's working correctly. So I hook a DP cable up to it and try to change the source.. black screen loss of signal. Eventually my Galax vanishes from device manager. I've tried reseating etc many times and currently can't get it back at all. There is no obvious reason as to why it suddenly stopped working. The plug which has caused problems at various points (just meaning the issues people have had since they changed to the new power plug.. I haven't personally had any issues) looks brand new still. So no scorches, damage etc Pulled the card out and all the contacts on the PCI slot look good. There was no spark, explosion or smell.. many years ago I had a card fail and blow the pci slot right off the motherboard and looked like a small sun in my pc case when it happened. So nothing like that. It's just like the card and pc are no longer talking to each other.

I'll have to get into my parts closet to grab another motherboard/power supply so I can set up a open bench machine to see if that can see/post with the 4080. Right now it seems pretty dead for no apparent reason or bricked etc (something).

Either I was just really unlucky today or this driver has some real issues.

 

[ThomasK]

"This" meaning what? Do you have a number of the failing cards? A number of cards being sold?
Because if you look at this objectively, one of the reports posted literally says "all 5090 and 5090D have exploded". Are we not allowed to take that with a grain of salt?

This, meaning all the reports. You should work on your interpretation skills.

Also, there are barely any cards launched and available to begin with, so any amount counts.

Lastly, blackwell was overheating on datacenters before that.

"The first shipments of racks with Blackwell chips are getting too hot and have connection issues between chips"

NVIDIA's GB200 "Blackwell" Racks Face Overheating Issues | TechPowerUp

 

[bug]

This, meaning all the reports. You should work on your interpretation skills.

Also, there are barely any cards launched and available to begin with, so any amount counts.

Lastly, blackwell was overheating on datacenters before that.

"The first shipments of racks with Blackwell chips are getting too hot and have connection issues between chips"

NVIDIA's GB200 "Blackwell" Racks Face Overheating Issues | TechPowerUp

To sum it up: you don't have numbers, only some reports. Based on that, you want everyone else to shut up. Mmmkay.

 

[Degreco]

If it's true... It's a typical Nvidia Cluster**ck.

Consider this: Being an early adopter, you go out of your way to buy a 5090 now, then RMA it after a day and then wait weeks for the replacement, while others buy cheaper and better, as they will certainly fix the issue in the next weeks.

But this kind of issues are testable before launching the product!!
It's not like an Intel 14th gen CPU that gets degraded after being fed power spikes over months. It's a 0 day bug, which is unforgivable (if true).

Another signal how they shifted all their focus to their Data Center clients.

Nvidia is still surprisingly unprofessional in testing things. See for example the Nvidia App, which was Beta for almost all of 2024 until they finally were able to release it. It matured with the beta testers. No adequate inhouse testing was active on this (it seems). They even forgot to add Gsync controls into their App at the start.

Don't mistake a high stock value with a professional corporation. They came to fame (and value) unexpectedly after 11/2022. They are toying with their B2C userbase right now because almost 9/10 of a $ in revenue is coming from B2B now. And they are toying with their large data center clients as well as they constantly increase pricing like a monopolist does. Most of us are forced to use their products for now. But customer retention is not in their mind. In the next years Nvidia needs to stay at the top of R&D with this kind of strategy because the minute somebody will catch up with raw GPU power or smarter software solutions they are dead in the water. Most won't shed a tear when this happens.

 

[ThomasK]

To sum it up: you don't have numbers, only some reports. Based on that, you want everyone else to shut up. Mmmkay.

Once again: you're lacking in interpretation skills.

I was hoping the fanboys would begin to shut up by now, but no, they'd rather defend the trillion dollar overvalued company

Are you a fanboy? Then shut up as well, 'cause I never said EVERYONE.

 

[evernessince]

This is a completely expected teething-pain issue with a new PCIe generation and will undoubtedly be resolved in a new BIOS.

Expected? I sure don't have any issues with any of my other PCIe 5.0 devices. Apparently storage manufacturers know how to make products that conform to PCIe standards far better than the 3 trillion dollar Nvidia.

If this turns out to be true, it's a colossal screw-up by Nvidia. Of course I think we should wait for more evidence but people rushing to say this is expected because PCIe 5.0 are absolutely jumping on the sword for Nvidia.

The same people saying this was expected and should be excusable are the same people who are first in every thread bashing AMD, even when the thread isn't related to an issue or when spouting some dated input (like saying AMD drivers are unstable).

The double standard is obvious. Bricking cards from a game (new world) and burning adapters are completely fine. I don't have any problem with people pointing out AMD's screw-ups (and there are a LOT of them) but the rush to defend Nvidia is nuts. A part of the reason Nvidia gets away with this stuff is because of this mentality.

At least Ray-tracing does not brick your card...

Hmm, I would not bet against that on my Nvidia future failure reason bingo card.

 

[mkppo]

Hey this isn't the first time. I honestly don't understand why nvidia usually gets a free pass or things get swept under the rug but I had one of the 3090's die on me a couple of years ago with the not-so-infamous driver bug. Thankfully it wasn't the main one on water which definitely didn't have warranty.

Now wait till nvidia says it's just the motherboard makers that failed hard and you need a new nvidia arm CPU and purpose built mobo for their GPU's. This isn't a fail, it's a weapon to counter attack if they wish.

When is the last time an AMD driver killed their GPU? I remember there was one, but it must've been so long ago that i forgot.

 

[Bomby569]

No free pass here. Nvidia is abusing their dominance to push for all sorts of crap, and worst untested crap. There is no excuse for this, they had or should have had months/years to test this before implementing it, just like the power connectors for example.
If you test a card door thousands of times to just see if it resist open and shutting, I'm sure you can test a gpu, that can start a fire in some cases, before getting out the door.

 

[Godrilla]

One time use utensils.

 

[Vayra86]

Nvidia shares be going rrrrrrrrrb this time

I guess even trillions of dollars of evaluation can't guarantee functioning products. Thank God there weren't that many of these available for purchase anyway.

Trillions of dollars cannot buy time. I think Nvidia scrambled to Blackwell and this is one of the results. That and the just overall meagre uplift even considering the shader count. Its a big box of nothing. But at least you've got 3 extra fake frames now.

 

[Godrilla]

Nvidia shares be going rrrrrrrrrb this time


Trillions of dollars cannot buy time. I think Nvidia scrambled to Blackwell and this is one of the results. That and the just overall meagre uplift even considering the shader count. Its a big box of nothing. But at least you've got 3 extra fake frames now.

Back in the day even sli had better scaling than this monolithic core density scaling. lol.