NVIDIA GeForce RTX 5090 Spotted with Missing ROPs, NVIDIA Confirms the Issue, Multiple Vendors Affected, RTX 5070 Ti, Too

[JustBenching]

The more you spend the harder you will fall.

- Poor performance
- High prices
- A lot of issues
- Worst nvidia gpu generation ever

What was the last time when nvidia had such bad times in general ?

I thought RTX 40 Series was the worst but this seems to be is even worse thanks to nvidia buyers.

Poor performance made me chuckle, not gonna lie. Poor performance... 6 fastest cards on raster performance are all nvidia.

 

[Krit]

Poor performance made me chuckle, not gonna lie. Poor performance... 6 fastest cards on raster are all nvidia.

Just stop it or wake up! When looking at history in general (performance gains) RTX 50 Series on average are the weakest generation ever released.

 

[JustBenching]

Just stop it or wake up! When looking at history in general (performance gains) RTX 50 Series are the weakest generation ever released.

Imagine how freaking terrible everyone else is doing on performance gains if they can't even get top 5 against these crappy nvidia cards. Insane, right?

2 crap nvidia generations in a row and instead of everyone else passing them by in performance, the gap gets larger and larger. Unfathomable...

 

[Krit]

Imagine how freaking terrible everyone else is doing on performance gains if they can't even get top 5 against these crappy nvidia cards. Insane, right?

AMD is not going to high end (big mistake) only mid range similar to RX 5700 XT.

 

[MacZ]

AMD is not going to high end (big mistake)

They probably prefer to produce AI GPUs, where the money is.

And until you are willing to pay $11,000 for a high end gaming GPU, it won't change.

 

[JustBenching]

AMD is not going to high end (big mistake) only mid range similar to RX 5700 XT.

They are going to high end with prices, it's just the performance that's going to be low end. The 9070xt is rumored to be a 649$ gpu at minimum. It should be easily smacking the 5080, especially considering your claims about 2 consecutive crap generations by nvidia. Lets wait and see

 

[Godrilla]

A full $1.3 millions ?

I'm sure Jensen at the head of its $3 trillions company will have a hard time sleeping.

Especially when you can RMA the faulty cards.

Doesn't Nvidia have 2 investigations with the justice department as well. Few weeks ago deepseak just sneezed and the stock responded by dropping by 17% although which recovered and then some. I guess if Nvidia doesn't mind the negative publicity from every tech forum and tech media 24/7 news cycle by doing an ai simulation that proves the risk is worth it. All we want is the beta stage to end and Nvidia fix all the problems with fire risk, mislabeling out of spec, not removing legendary features, stable drivers providing enough stock to match msrp.
I guess if we live in a bubble and the mis labeling out of spec was the only problem I would we cool as a cucumber. Unfortunately Blackwell has a laundry list of issues and new card burning on a daily basis at this point.
Although imo I believe it's all intentional*. Someone on the board wants to paint gaming business as not profitable and is intentionally sabotaging the gaming side. This might lead to a solidification of a only ai business from the looks of it.

 

[Krit]

And until you are willing to pay $11,000 for a high end gaming GPU, it won't change.

Sorry but i'm not an idiot. For gaming anything over 1000$ is a nonsense "get a life" there are way better ways how to spend money for me gpu is just a nice looking brick.

 

[Dingizo]

A full Blockwell chip is between 30 to 40 K and Nvidia cant produce enough of them. I have a feeling Zotac was aware of what they are doing. The 5090 with 168 ROPs will comeback in the form of RMA and will be fused off to become 5080 Ti super Duper. Not sure if that is possible after the chips have been packaged, but from a business stand point it makes a lot of sense. Im not completely convinced Nvidia gives a shitt about the 5090 2k cards that any AIB is selling

 

[Hecate91]

Doesn't Nvidia have 2 investigations with the justice department as well. Few weeks ago deepseak just sneezed and the stock responded by dropping by 17% although which recovered and then some. I guess if Nvidia doesn't mind the negative publicity from every tech forum and tech media 24/7 news cycle by doing an ai simulation that proves the risk is worth it. All we want is the beta stage to end and Nvidia fix all the problems with fire risk, mislabeling out of spec, not removing legendary features, stable drivers providing enough stock to match msrp.
I guess if we live in a bubble and the mis labeling out of spec was the only problem I would we cool as a cucumber. Unfortunately Blackwell has a laundry list of issues and new card burning on a daily basis at this point.
Although imo I believe it's all international. Someone on the board wants to paint gaming business as not profitable and is intentionally sabotaging the gaming side. This might lead to a solidification of a only ai business from the looks of it.

I sort of doubt the average person buying Nvidia cards is even reading tech news, or perhaps they do and it's just stockholm syndrome people thinking there are no other options and must hand over their credit card for the latest from team green. Nvidia is damaging the gaming market with stagnation, proprietary features, hazardous power connectors, and now defective cards that people will still buy. But I'd rather buy from AMD or Intel, as Nvidia has priced me out of the xx80 market since the 30 series.
Also didn't Jensen say the future is in AI, not with GPU's? If Nvidia wants only an AI & datacenter business then they can sell gamers an ARM powered SoC with an Nvidia GPU and make even more money by selling geforce now subscriptions on every SoC they sell.

 

[lexluthermiester]

I sort of doubt the average person buying Nvidia cards is even reading tech news

There are a large percentage of them. @W1zzard has the official numbers for views on each review page. I'm betting it's not an insignificant number. And that is just for this site. There are others and then there are the tech reviewers on video sites. Lots of people look for info on PC hardware.

 

[MacZ]

There are a large percentage of them. @W1zzard has the official numbers for views on each review page. I'm betting it's not an insignificant number. And that is just for this site. There are others and then there are the tech reviewers on video sites. Lots of people look for info on PC hardware.

I think this is another case of wishful thinking.

Most of those people just want to play games and couldn't care less about the nitty gritty details of PC hardware politics. They just want to be able to play games and that's it. For most of them their knowledge stops after they have they PC setup and able to play.

Also, the vast majority isn't buying high end GPUs in the first place. We'll see if any of the current problems translate to more mundane gaming GPUs.

Edit : for example, there are 140 million steam accounts and 16 millions LTT subscribers. That is 11%. Not completely irrelevant, but not a sizeable chunk either. I think the revolution against Nvidia will have to wait a bit more.

 

[OkieDan]

Zotac should just ship out some 5080 Ti stickers people can slap on their cards and call it a day...

That would be like Ford's very old "park to reverse" recall that the government let them "fix" by sending owners are sticker to place on sun visor.

Ford Transmissions Failure to Hold in Park

%

www.autosafety.org

 

[FreedomEclipse]

Funny how they are downplaying it and labeling it as an "rare issue" but this issue probably affects hundreds of not thousands of people. That's not rare at all.

 

[NesteaZen]

So no, you're not done then. Citing Reddit does not help your silliness. Cite something with

except Macz is correct

Second, To maximize AI GPUs production, they need to lessen gaming GPUs production. It

the truth is probably somewhere in the middle since die production goes from top to bottom. in the sense that if they could make
no 5090, they probably would. 5090 dies is the beginning of the shit tierlist.
what Nvidia is interested in, is h100 and gb100. to nvidia, 5090 is a necessary evil.

how will this look like in 10 years when or even if AI potentially bursts, who knows. for all we know AI could make bitcoin encryption pointless.
I'd like to remind you all: read the comment especially lol

Intel predicts 10GHz chips by 2011 | Chips | Geek.com

Intel is predicting that its microprocessors will hit 10GHz by the year 2011. In addition, it is currently working on a system bus that is 10 times faster than its upcoming 400 MHz [...]

web.archive.org

seems like zdnet comments got nuked https://web.archive.org/web/2020111...nts-show-70136aa3-2f69-11e4-9e6a-00505685119a

RTX 5090 supply to greatly increase in the coming weeks

An X leaker says the RTX 5090 shortage could be over soon due to low demand for Nvidia's B200 GPUs. The GB100 GPUs powering it will be repurposed for the RTX 5090's GB202.

www.notebookcheck.net

Published 02/17/2025
Getting your hands on a GeForce RTX 5090 has been nigh impossible due to low supply and scalpers. That could change in the coming weeks if a leak from X leaker @Zed__Wang is accurate. Apparently, the market will get a "stupidly high" number of RTX 5090 cards in about a month.

The leaker says Nvidia has an oversupply of B200 data centre GPU due to plummeting demand. It uses a GB100 GPU, from which the RTX 5090's GB202 is derived. Hence, the leftover GB100 dies will be repurposed into RTX 5090s, effectively ending the shortage.

That said, this goes against an earlier report which stated Nvidia's RTX 50 series laptop chips have been pushed back due to supply issues. Now, the exact reason for that delay reason isn't clear, but some speculate it to be due to a shortage of GPUs, while others opine it is due to performance and functionality issues.

nvidia advanced packaging capacity at DuckDuckGo

DuckDuckGo. Privacy, Simplified.

duckduckgo.com

Oct 11th, 2024:
NVIDIA's "Blackwell" series of GPUs, including B100, B200, and GB200, are reportedly sold out for 12 months or an entire year. This directly means that if a new customer is willing to order a new Blackwell GPU now, there is a 12-month waitlist to get that GPU. Analyst from Morgan Stanley Joe Moore confirmed that in a meeting with NVIDIA and its investors, NVIDIA executives confirmed that the demand for "Blackwell" is so great that there is a 12-month backlog to fulfill first before shipping to anyone else. We expect that this includes customers like Amazon, META, Microsoft, Google, Oracle, and others, who are ordering GPUs in insane quantities to keep up with the demand from their customers.

The previous generation of "Hopper" GPUs was ordered in 10s of thousands of GPUs, while this "Blackwell" generation was ordered in 100s of thousands of GPUs simultaneously. For NVIDIA, that is excellent news, as that demand is expected to continue. The only one standing in the way of customers is TSMC, which manufactures these GPUs as fast as possible to meet demand. NVIDIA is one of TSMC's largest customers, so wafer allocation at TSMC's facilities is only expected to grow. We are now officially in the era of the million-GPU data centers, and we can only question at what point this massive growth stops or if it will stop at all in the near future.

NVIDIA "Blackwell" GPUs are Sold Out for 12 Months, Customers Ordering in 100K GPU Quantities

NVIDIA's "Blackwell" series of GPUs, including B100, B200, and GB200, are reportedly sold out for 12 months or an entire year. This directly means that if a new customer is willing to order a new Blackwell GPU now, there is a 12-month waitlist to get that GPU. Analyst from Morgan Stanley Joe...

www.techpowerup.com

 

[PerfectWave]

so techpowerup inventor of gpuz didnt notice the missing rop ROLF cant be true lol

 

[W1zzard]

so techpowerup inventor of gpuz didnt notice the missing rop ROLF cant be true lol

Unlike some stories that you might have heard that claim the opposite, I'm just a human. When I found out, I published this article (after investigating, and doing my research homework)

 

[Bwaze]

It isn't a charity when Nvidia just cannot treat the gaming customers like crap, Nvidia is still invested in the gaming market and not caring about the gaming market is a dumb move as the AI bubble can pop, or the need for Nvidia for GPGPU as AI cards can diminish as many other companies are producing their own compute hardware.

Of course being first in a market that brings in (inflated) $3.3 billion in a quarter sounds a lot, but if AI server dies are supply constrained, sacrificing Gaming hardware to increase the focus on the sector that does bring in the money has all the benefits, and no downsides. It shows the commitment Nvidia has, it shows the investors and the AI hardware users where the focus lies - not in making toys.

And even after all the debacles, and very little cards actually sold we’re in for a surprise revenue. I imagine at this point with the overpriced RTX 40x0 old gen cards and nonexistent stock of RTX 50x0 in many markets, and especially in the shops that show sales numbers (like Mindfactory), even Intel might overtake them on actual gaming card sales, not just AMD. But the new quarter revenue numbers will show that “Gaming and AI PC” sector will grow by some absurd value, and Nvidia will claim it’s also from the strong sales of gaming cards - and we have no way to dispute that.

Also according to US laws, Jensen could be in trouble for selling the AIBs cards with missing ROPs, he's already lied about the 5070 being faster than a 4090 when the 5070Ti doesn't even beat a 4090.

First one they will claim is clearly a malfunction in a product, which you can RMA for that reason. Will you be able to, since there is no stock? Will Nvidia and AIB partners inform their buyers to check their cards with some obscure third party software for a very specific number among dozens, or are the articles in the daily tech news enough? Were these cards really “repaired” by using special bioses, basically making them separate SKUs? That’s for investigative journalism to find out. Oh, we basically don’t have one?

About performance claims - these are marketing claims. Everyone knows they are not to be believed, if you think otherwise, you’re just a laughing stock, not a deceived customer, sadly. And they are true, in a very specific way - not usable by most, but that’s not their problem.

 

[Crazybabe]

Is this real?

http://imgur.com/a/uQqSKcP

96 vs 88 ROPs
Pixel Fillrate: 287.7 Gpixel/s vs 223.7Gpixel/s
Texture Fillrate: 839.2 GTexel/s vs 711.8 GTexel/s
Bandwidth: 1088 GB/s vs 896 GB/s

 

[JustBenching]

Is this real?

http://imgur.com/a/uQqSKcP

96 vs 88 ROPs
Pixel Fillrate: 287.7 Gpixel/s vs 223.7Gpixel/s
Texture Fillrate: 839.2 GTexel/s vs 711.8 GTexel/s
Bandwidth: 1088 GB/s vs 896 GB/s

Yes, from computerbase. As my german friends would say, das ist nicht gut.

 

[sbacc]

There are a large percentage of them. @W1zzard has the official numbers for views on each review page. I'm betting it's not an insignificant number. And that is just for this site. There are others and then there are the tech reviewers on video sites. Lots of people look for info on PC hardware.

For people buying directly a dGPU through retail, you are probably right, but what about people buying prebuilt ?

 

[LittleBro]

I still think missing performance has something to do with why RTX 5070 and 5060 got delayed. Officially they got delayed due to performance issues.

This might be a Blackwell architecture design flaw and might as well impact AI GPUs, since AI GPUs are Blackwell, too. In AI compute no one gives a shit about ROPs, so it's actually not a problem there. As for gamers, this is bad. That's why I don't like next AMD RDNA to be UDNA, that means same architecture for AI and gaming. Sure, it saves costs, and maximizes profits. The downside is it allows them to fully focus on AI market and serve only (defective) leftover chips to gamers. Hopefully I'm terribly wrong, but I think that once AMD unifies the architecture, it becomes greedy as Nvidia.

Dedication to original commitment is no more, all is about money now: selling potentially dangerous products, selling defective products, selling unpolished products...

Maybe Intel will save us gamers. For a while. For greed is eternal and contagious.

As for prices, TSMC margins are brutal. There is no competition. So even there is very little competition in GPU market, unfortunately, TSMC has impact on all those GPU chips. Which means get used to getting mid tier GPU at $1k. Or worse. Unless TSMC gets a competition, this is the new standard.

 

[jnv11]

Gamers Nexus wants to buy one of these RTX 5090s or RTX 5070 Tis with fewer than specified ROPs, and will pay the cost on your receipt + $500 for your defective GPU according to https://www.youtube.com/channel/UCh...unity?lb=Ugkx0WLxvKTWH-hKeMABJoYuxc0Pz935DnLZ

 

[jnv11]

I've already explained it in a previous post, AIBs know that they are shipping these chips because they need a different VBIOS and they have to make sure it's the right one otherwise the GPU wont function correctly. Both Nvidia and their board partners did this on purpose, it wasn't a mistake, they knew exactly what they were selling.

I sincerely doubt this. A full GB202 has 192 ROPs. An RTX 5090 that meets spec contains 176 active ROPs, so 16 are disabled. The kinds of defective RTX 5090s discussed in the article have 168 active ROPs, so they can be disabled in groups of 8 or factors of 8. I would think that disabling the same 16 ROPs via VBIOS would cause yields to plummet because those left could be damaged or failed. I think that to keep the yields high, one should disable the ROPs clusters that contain bad ROPs first, and if too many are left, disable any to make them meet spec, but probably disable the ROPs that have the slimmest margins to failure for the best results. Generally, chips can be designed so that undesirable parts can be fused off by blowing fuses internal to the chips to disable the undesirable parts of the chips. With this kind of chip harvesting, one could write one BIOS that detects which parts of the chip are active and work with the active parts while working around the parts of the chips that were fused off, removing the need to have multiple BIOSes for different variants of the chip which contain the same number of components but have components located in different parts of the chip fused off based on which ones failed or were the flakiest.

 

[Tek-Check]

I sincerely doubt this. A full GB202 has 192 ROPs. An RTX 5090 that meets spec contains 176 active ROPs, so 16 are disabled. The kinds of defective RTX 5090s discussed in the article have 168 active ROPs, so they can be disabled in groups of 8 or factors of 8. I would think that disabling the same 16 ROPs via VBIOS would cause yields to plummet because those left could be damaged or failed. I think that to keep the yields high, one should disable the ROPs clusters that contain bad ROPs first, and if too many are left, disable any to make them meet spec, but probably disable the ROPs that have the slimmest margins to failure for the best results. Generally, chips can be designed so that undesirable parts can be fused off by blowing fuses internal to the chips to disable the undesirable parts of the chips. With this kind of chip harvesting, one could write one BIOS that detects which parts of the chip are active and work with the active parts while working around the parts of the chips that were fused off, removing the need to have multiple BIOSes for different variants of the chip which contain the same number of components but have components located in different parts of the chip fused off based on which ones failed or were the flakiest.

Whatever the case, the bottom line is that shipped cards with fused off ROPs do not match the official spec. This was supposed to be picked up by QA/QC process of optical scanning of each chip. This is on Nvidia.

Failing that, board partners had another opportunity to check the specs during their internal testing, which was very tight due to Nvidia distributing chips very late. Remember, cards on dispplay at CES were dummy cards, without Blackwell chips inside. Board partners could have simply used GPU-Z tool to check each parameter, which is how tech community discovered the issue.

The fact that tech community discovered this issue by using GPU-Z tool brings serious questions about the quality of QA/QC process at Nvidia and board partners.