• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

NVIDIA GeForce RTX 5090 Spotted with Missing ROPs, Performance Loss Confirmed

ChadD said:
So many open box deals. lol
Click to expand...
Yeah, in the JayzTwoCents video he says people can't just return them to Microcenter because they'll just turn around and sell them as open box. He suggests that Microcenter would have to test every one for the "defective" ones...yeah, that's never going to happen. They'll just send them back to the manufacturer.
 
MavericK said:
Yeah, in the JayzTwoCents video he says people can't just return them to Microcenter because they'll just turn around and sell them as open box. He suggests that Microcenter would have to test every one for the "defective" ones...yeah, that's never going to happen. They'll just send them back to the manufacturer.
Click to expand...
Probably lots will just be sold as open box product. Ya they aren't going to test every return, and people that spent 2-2.5k for a card they feel ripped off on are likely to just say "changed my mind".

I imagine a lot of people however will be testing Scalper tech support.... lol
 
At the register they print a yellow sticker. If it's not going on the shelf it's RTV (Return To Vendor) and if being sold as open box, typically it's marked down 10%. If it sits there for a while it gets marked down further but with these things that's never gonna happen!
 
zehoo said:
I can't help but assume Nvidia knew about the missing ROPs and shipped them hoping to not get caught. If supply wasn't so constricted they probably wouldn't have been.
Click to expand...
NO. These chips were released by accident, I mean incompetence by QA.
 
Flogger23m said:
Ignoring price and low generational performance improvements, this seems to be Nvidia's worst launch in recent memory. At least their worst in 10+ or so years.
Click to expand...
Blackwell has been their worst nightmare at every level. It just hasn't worked. They have had major issues fabricating all versions of Blackwell. One big advantage of fabricating the data center versions first is that it should smooth out fab issues with the later consumer parts. Doesn't seem to be the case for Blackwell.
 
Flogger23m said:
Ignoring price and low generational performance improvements, this seems to be Nvidia's worst launch in recent memory. At least their worst in 10+ or so years.
Click to expand...
To my memory, the FX cards were possibly worse in different ways... badly delayed, under performing and fan loud enough to wake the dead.
 
Nvidia has just confirmed it shipped some RTX 5090, RTX 5090D, and even some RTX 5070 Ti graphics chips that were missing render units, as TechPowerUp originally reported — and that you’ll be able to get a replacement if your card was affected.

Nvidia GeForce global PR director Ben Berraondo tells The Verge:
We have identified a rare issue affecting less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D and 5070 Ti GPUs which have one fewer ROP than specified. The average graphical performance impact is 4%, with no impact on AI and Compute workloads. Affected consumers can contact the board manufacturer for a replacement. The production anomaly has been corrected.
Click to expand...

While limited, the manufacturing issue affected multiple Nvidia graphics card partners: reports came in of Zotac, MSI, Gigabyte, Manli, and even an Nvidia Founders Edition card with missing ROPs. You can use GPU-Z to check your card and see if it’s showing the proper number of 176 ROPs; if fewer, you should probably get it replaced.


https://www.theverge.com/news/617901/nvidia-confirms-rare-rtx-5090-and-5070-ti-manufacturing-issue
 
Can we just have a mulligan on this series?
Bring back the 40 Supers for another year while Nvidia cleans up, builds supply. Then do it properly.
 
El_Capitan said:
That's not how QA works. QA will easily find stuff like this. The Product Manager and upper management just approve releases despite what QA finds.
Click to expand...
This statement leaves me gobsmacked.

When I worked (for a long time) as a product manager for enterprise software, I would never, ever approve a release without QA support and concurrence. Depending on the severity code of the bug the product might be released with a release note calling out the issue and suggesting a workaround. Can't speak to upper management although there were companies, no names to protect the guilty, which had a "ship it and fix it" attitude no matter what. Gotta book that revenue.

Hardware is much more difficult. If the issue can be fixed with a firmware update, then see above. If the issue is "hardcoded" in silicon, then that chip needs to be binned down to a lower level.

Imagine just how much "stuff" would hit the fan if AMD did this with their GPUs. For NVidia, either their QC needs to be seriously fixed, or someone's ass should be fired pronto. If this was an upper level management decision, then NVidia is way to smug and arrogant for our good.
 
Nvidia's directive for their screwup is to contact the board partner for a replacement. Even if Asus doesn't deny the claim by alleging customer that you the purchaser somehow damaged the chip, Asus and the other AiB's will reluctantly send you a replacement in twelve to sixteen months after they have received the return from you, shipped at your own expense.
 
philb2 said:
This statement leaves me gobsmacked.

When I worked (for a long time) as a product manager for enterprise software, I would never, ever approve a release without QA support and concurrence. Depending on the severity code of the bug the product might be released with a release note calling out the issue and suggesting a workaround. Can't speak to upper management although there were companies, no names to protect the guilty, which had a "ship it and fix it" attitude no matter what. Gotta book that revenue.

Hardware is much more difficult. If the issue can be fixed with a firmware update, then see above. If the issue is "hardcoded" in silicon, then that chip needs to be binned down to a lower level.

Imagine just how much "stuff" would hit the fan if AMD did this with their GPUs. For NVidia, either their QC needs to be seriously fixed, or someone's ass should be fired pronto. If this was an upper level management decision, then NVidia is way to smug and arrogant for our good.
Click to expand...
Hardware testing is typically a lot easier. There's strict tolerances that must be passed, first, before undergoing further tests, otherwise you're just wasting time. You're basically testing the specifications. A simple scenario would be case fans and RPM's at 12V. I think the tolerance is 10%, so if you're stating a fan has 1800 RPM's, as long as it gets up to 1620 RPM's, it could be considered a pass. However, the ROPs scenario is much easier. It's like making sure you count the fan blades on a case fan, and there should be 8 blades with 0 tolerance. If you find one with 7 blades, then it's a fail. With software testing, you're testing the specifications, but also variables the specifications may not have covered.

As a QA Manager and Lead, I've never signed off on builds that didn't pass testing, but mostly because liability could then fall on me (which I never stated, but they also never stated I would be liable if I did sign off on a release). It's a dog-eat-dog world out there in the corporate sector. The Product Manager or upper management would make the decision to release despite testing not passing, and it was done often (which never went well). No matter what, QA still got the blame, but we couldn't be held liable since we didn't sign off on it being released.

At any rate, 8 ROPs out of 176 is 4.55%, and maybe the tolerances are having no more than 5% missing ROPs? I don't know, but we've never seen this before, have we?
 
So this issue has just surfaced yesterday and nVidia already knows exactly what percentage of cards are affected? Yeah, they knew all along this was a problem and shipped them anyway because they knew they had barely any cards to sell.
Not too worry though, Jensen will sign these defective cards and sell them as a 5090 SE with a MSRP of $2999 and people will buy them.
 
They intentionally shipped the GTX 970 despite its flaw, there's a very high probability they found this too far into production and just said fuck it.
 
vegeta535 said:
Don't blame QA always for shit like this. I do QA at my co.panu and I get overruled by management all the time when I bring up issues so they can push it out the door to meet their numbers.
Click to expand...
Happens only all the time. Me, when I was working in enterprise software product management, QA and support were my BFFs. Screw up either of these and the stuff hits the fan in the CEO or VP's office. Way above my pay grade. but I could still get it in the neck. If product is shipped late, maybe some sales guys won't make their personal number, and they won't go to "club." Boo hoo boo hoo.. I could deal with that.
 
vegeta535 said:
Don't blame QA always for shit like this. I do QA at my co.panu and I get overruled by management all the time when I bring up issues so they can push it out the door to meet their numbers.
Click to expand...
This. When I did QA we were overridden all of the time. It's apart of the business: sales loss due to issues vs cost of shipping and missing deadlines.
 
Does anyone know if the rops are fused off on the gpu or if it’s a bios thing?
 
DooKey said:
NO. These chips were released by accident, I mean incompetence by QA.
Click to expand...

Vegas P11 said:
So this issue has just surfaced yesterday and nVidia already knows exactly what percentage of cards are affected? Yeah, they knew all along this was a problem and shipped them anyway because they knew they had barely any cards to sell.
Not too worry though, Jensen will sign these defective cards and sell them as a 5090 SE with a MSRP of $2999 and people will buy them.
Click to expand...

It's a screwup, but not primarily by QA. It was inventory management that had a faceplant. Nvidia has historically collected sub-spec binned parts and done low volume OEM or Asian PC cafe cards made with them - typically at the end of the product cycle. Sites like WCCFTech would write about them as outrage bait in the past; they probably still do but I haven't read it regularly for the last 2 or 3 card cycles to catch them.

Once the screwup was found someone checking the records would have been able to determine that a lot of bad-bin parts was accidentally sent into the general supply chain instead of being left for a weird/crap card later.
 
Brackle said:
This is an nvidia issue. Not the aib’s
Click to expand...
Could be on both, AIB test suite could need to catch something like that (if it is not something that occurred over time post assembly after shipping). Maybe they did hide/ignore error code and decided to ship anyway.

The idea that they were having fake issues and manipulated supply to create a FOMO to make more money (not sure how it works when no one can buy any card....) was maybe false, maybe they had actually issue with blackwell all along.

Vegas P11 said:
So this issue has just surfaced yesterday and nVidia already knows exactly what percentage of cards are affected?
Click to expand...
Possible that installing Nvidia driver-geforce experience phone home that type of specs as well (if it is something cpu-z tell a regular users, could easily be in the NV card info) ? AIBs post assembly test suite could have them as well.
https://gamersnexus.net/industry/2672-geforce-experience-data-transfer-analysis
 
Last edited:
erek said:

RTX 5080 missing ROPs​

https://www.reddit.com/r/nvidia/?f=flair_name:"PSA"
https://www.reddit.com/r/nvidia/com...p&utm_name=ioscss&utm_source=share&utm_term=1
Click to expand...
Reminder: nvidia announcement only said 5090, 5090D, and 5070 Ti. They are clueless.

Nvidia GeForce global PR director Ben Berraondo tells The Verge:

We have identified a rare issue affecting less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D and 5070 Ti GPUs which have one fewer ROP than specified. The average graphical performance impact is 4%, with no impact on AI and Compute workloads. Affected consumers can contact the board manufacturer for a replacement. The production anomaly has been corrected.
 
TaintedSquirrel said:
Reminder: nvidia announcement only said 5090, 5090D, and 5070 Ti. They are clueless.

Nvidia GeForce global PR director Ben Berraondo tells The Verge:

We have identified a rare issue affecting less than 0.5% (half a percent) of GeForce RTX 5090 / 5090D and 5070 Ti GPUs which have one fewer ROP than specified. The average graphical performance impact is 4%, with no impact on AI and Compute workloads. Affected consumers can contact the board manufacturer for a replacement. The production anomaly has been corrected.
Click to expand...
They should offer a partial refund through retailers for affected units in addition to offering a replacement. I'm sure many would be ok with the minor perf hit if they got $100 back or so at least. If the GPUs are fully functional otherwise, just seems like a waste to RMA them or not re-SKU the affected models for a slight discount. But I'm sure somehow recalling/refunding all affected units is a negligible financial hit to Nvidia and the AIBs as well.
 
5687af5f4fe7621dbbe2582a70a9be2688e5275a5c2bd5f6bf0d9b6a087a4a5c.gif
 
vegeta535 said:
Don't blame QA always for shit like this. I do QA at my co.panu and I get overruled by management all the time when I bring up issues so they can push it out the door to meet their numbers.
Click to expand...
Agreed. I worked in QA for over twenty years. If management wanted to meet quota and the defect wasn't too severe it went out the door.
 
You must log in or register to reply here.
Back
Top