Question Why is NVidia allowed to get away with flatly lying about gen-on-gen performance uplifts?

[YouFilthyHippo]

This is quite a rant. If this is against forum rules, I apologize, go ahead and delete. When the RTX 4090 came out, NVidia claimed it was 2-4x better performance than the 3090. This is a lie. When you compare them side by side, on two exact identical machines and run an equal objective neutral benchmark, both with DLSS disabled, the 4090 is about 70% better performance. In NVidia's extremely biased and cherry-picked tests, the 4090 is getting help from DLSS 3.0, and I'm betting the 3090 has DLSS disabled. HOLY CRAP!!!! 4x UPLIFT!!!!.

It's not even exaggerating or stretching the truth. It is full-on straight up lying. They did this exact same thing with the 3090. They said it was 2x faster than the 2080Ti. In a side-by-side, same machine, level playing field, objective, neutral test, both with DLSS disabled, the 3090 has about 50% better performance.

Why is NVidia allowed to get away with flatly lying like this? How is their no class action lawsuit for false advertising?

 

[zx128k]

Not a bad as the RX 7900XTX performance claims but yes the nVidia RTX 4090 benchmarks appear to use frame generation from DLSS 3. The RTX 3090 TI is DLSS 2 feature level in DLSS 3 with no frame generation. Thus it makes the performance of the RTX 4090 look far better.

 

[TerryLaze]

In NVidia's extremely biased and cherry-picked tests, the 4090 is getting help from DLSS 3.0, and I'm betting the 3090 has DLSS disabled. HOLY CRAP!!!! 4x UPLIFT!!!!.

So it did happen, so it's not a flat out lie.
Yes, everybody puts their best foot forward, from a personal level up to as high as it goes...nobody leads with their flaws...you cracked the code.

Why is NVidia allowed to get away with flatly lying like this? How is their no class action lawsuit for false advertising?

Because they lie within the accepted leeway of the law, can't sue them if they are on the side of the law.

 

[KyaraM]

From what I found, the claim was made about raytraced games, where the claim isn't too far off the mark to be considered an outright lie, and I think the gains without RT are very high, too. Also, if you want to sue Nvidia over this, be my guest, but please also include AMD since they are just as bad in that department. Look at their idiotic comment about 8k gaming before the launch of their 7000 series GPUs, for example. And it's not as if it's an outright lie in every case, either.
https://www.techspot.com/review/2588-amd-radeon-7900-xtx/#RT_1

Here, they really do show 2x performance between those cards. That's also why they always say "up to x times the performance of y". If they claimed "we offer 2x the performance of y or more!", you could actually peg them on it. But "up to"? Well, that's actually a true statement, if quite a bit misleading maybe.

Edit:
I actually wanted to link a specific image of that article, not the entire thing. Somehow, it did link the article, though. Oh well... the image I wanted to link was the "Dying Light Stay Human 4k Raytracing Performance" one. On the upside, the article highlights my comment on AMD also liking to exaggerate, so there is that at least xD
One could also argue that DLSS and FSR are features of card (more so with DLSS), so they do, in fact, flow into the comparison.

 

[DSzymborski]

Marketing departments tend to try and push the envelope. It's not easy to just file a class action lawsuit; you have to show specific harm (basically, demonstrate that you had a good faith basis to rely on the specific claims and suffered damages as a result) and state AGs don't generally get involved unless there's something egregious.

And Nvidia's exaggeration here seems to largely involve new features which the competing products don't have. They can make a very reasonable argument that they're highlighting a new feature that's part of the card and showing the benefits of a new feature.

 

[mjbn1977]

I agree that the marketing of Nvidia and also AMD sucks a lot when it comes to graphics cards. I can understand it a little bit in AMDs case, as their products lagging quite a bit behind Nvidia when it comes to a lot of features, such as Raytracing performance, Upscaling technology (DLSS 2 vs FSR 2), not having frame generation. They can only compete in raw rasterization performance, but at higher power draws, less efficiency, and worse drivers. So yeah, I can see how AMD might exaggerate their marketing a little bit as they have done in the past (new standard of PC gaming?? what???). But I don't understand why Nvidia has to do that. Just explain your tech, explain the features, show the raw power and don't come up with weird claims about performance increases that the average gamer doesn't understand. That should be enough for Nvidia. Also, their market share is really big enough and doesn't require such stupid markeing...what is it know 80%-85% right now for gaming GPUs???

 

[Endymio]

Because they lie within the accepted leeway of the law, can't sue them if they are on the side of the law.

While the rest of your post was spot-on, allow me to add a small caveat here. Anyone can sue over anything, at any time, legal basis or no. And -- with enough proper venue-shopping -- they may even win. Consider the case of a woman who sued Geico because that company insured a vehicle in which she contracted an STD while having sex in the back seat. She received a $5.2M settlement from arbitration; it was upheld by the appeals court, and, had the MO Supreme Court not chosen to intervene, Geico would now be paying that out -- and charging the rest of us higher premiums in result.

 

[hotaru.hino]

The thing to note here is what defines "performance"? Is it simply just the frames per second number? Okay, but in what scenarios? Which scenarios are invalid? And if they're invalid, why?

If the RTX 4090 can get 240 FPS with DLSS3 FG on while the RTX 3090 struggles to get 60 FPS, all other settings being equal, then yes, it's a 4x uplift in performance because the number we're measuring went up by 4 times. And if NVIDIA has data that can prove their claims with the settings and whatnot used so that anyone else can replicate it, and those people also get similar results, then the claim is more or less verified, no matter how much you disagree with how to get to that result.

And marketing's job is to make a product look good. They'll take any value they can get that can, with a bit of stretching, won't get them outright sued. And why the term "up to" is abused.

 

[sizzling]

This is why we don’t listen to marketing and use independent reviews. This type of cherry picked marketing happens in so much advertising we see.

I was on a course about 5 years ago led by a statistician and he showed us how the numbers often used on TV that say X amount of people agree or use a product are statistically insignificant and cannot be used to draw any useful conclusions. Ultimately the numbers quoted are true but are presented in a way to make the majority of people think it means something when in reality it does not, but because the number is true they are allowed to use it even if people don’t understand it.

 

[TerryLaze]

While the rest of your post was spot-on, allow me to add a small caveat here. Anyone can sue over anything, at any time, legal basis or no. And -- with enough proper venue-shopping -- they may even win. Consider the case of a woman who sued Geico because that company insured a vehicle in which she contracted an STD while having sex in the back seat. She received a $5.2M settlement from arbitration; it was upheld by the appeals court, and, had the MO Supreme Court not chosen to intervene, Geico would now be paying that out -- and charging the rest of us higher premiums in result.

LOL, right 'murica the land of the free and the home of the brave...lawsuits.
Do you know what the basis for the claim was?! Was it enough that they didn't state that they don't cover STDs contracted in the vehicle?

 

[YouFilthyHippo]

The thing to note here is what defines "performance"? Is it simply just the frames per second number? Okay, but in what scenarios? Which scenarios are invalid? And if they're invalid, why?

If the RTX 4090 can get 240 FPS with DLSS3 FG on while the RTX 3090 struggles to get 60 FPS, all other settings being equal, then yes, it's a 4x uplift in performance because the number we're measuring went up by 4 times. And if NVIDIA has data that can prove their claims with the settings and whatnot used so that anyone else can replicate it, and those people also get similar results, then the claim is more or less verified, no matter how much you disagree with how to get to that result.

And marketing's job is to make a product look good. They'll take any value they can get that can, with a bit of stretching, won't get them outright sued. And why the term "up to" is abused.

If those are the stipulations of the test, then it's not an apples to apples test. Remember, we are comparing the performance of 2 physical components. DLSS is not physical. What is the performance of 2 physical components? An RTX 4090 and an RTX 3090 are identifiers of 2 physical components. When you say "the performance of a 4090 is 4x a 3090", you have stated that physical component 1 can perform better than physical component 2 by a 4x scale. That is objectively incorrect. That is a false statement. The 4090 is getting an unfair advantage from a SOFTWARE component that the 3090 does not have. Now, run the 3090 with DLSS 3 and then we will talk. If it's a fair and equal apples to apples test, they either both run DLSS 3 during the test, or they both don't.

Or, how about this? You run your Cyberpunk bench on your 4090 and get 120FPS or whatever it is. I'll run my cyberpunk bench on my GTX 960 and get 180 FPS. So what if my 960 test was done on minimum settings at 240p. My GTX 960 got a higher frame rate IN THE SAME GAME than your 4090. I guess my 960 beats your 4090. No? No....

When you do a performance comparison of 2 physical components, the playing field must be level. Neither can have an unfair advantage. Both machines must use the exact same components, same game, same settings, same software. This is a true apples to apples performance comparison. In this comparison, the 4090 is about 70% faster than the 3090, not 2-4x

 

[randyh121]

Money. Thats why

 

[USAFRet]

When you do a performance comparison of 2 physical components, the playing field must be level. Neither can have an unfair advantage. Both machines must use the exact same components, same game, same settings, same software. This is a true apples to apples performance comparison. In this comparison, the 4090 is about 70% faster than the 3090, not 2-4x

If Component A can utilize different software than Component B, and if that software can result in XTimes the performance of COmponent B....no lie has been perpetrated.

Tests can be contrived in any way the tester wants.
Nowhere did they state "4x the performance with exactly the same conditions"

Cherry picking is absolutely typical.
We, the potential buyers, need to understand how marketing works.

If DLSS enables the 4XXX series to do better than a 3XXX that cannot use DLSS, no lie given.

 

[DRagor]

I'm waiting to see car manufacturers use same tactics then. "Yay, our new model is up to 4 times faster then old one!*"
LOL

Spoiler: *

bah, we forgot to put tires on old model though during test

 

[USAFRet]

I'm waiting to see car manufacturers use same tactics then. "Yay, our new model is up to 4 times faster then old one!*"
LOL

Spoiler: *

bah, we forgot to put tires on old model though during test

"PCIe 4.0 drives are twice as fast as PCIe 3.0 drives!!"

(only in certain very limited use cases)

 

[Endymio]

I'm waiting to see car manufacturers use same tactics then. "Yay, our new model is up to 4 times faster then old one!*"
LOL

Spoiler: *

bah, we forgot to put tires on old model though during test

NVidia's tests were conducted in raytracing mode. People actually do play games in raytracing mode. How often do you see people driving cars without tires?

 

[YouFilthyHippo]

If Component A can utilize different software than Component B, and if that software can result in XTimes the performance of COmponent B....no lie has been perpetrated.

Tests can be contrived in any way the tester wants.
Nowhere did they state "4x the performance with exactly the same conditions"

Cherry picking is absolutely typical.
We, the potential buyers, need to understand how marketing works.

If DLSS enables the 4XXX series to do better than a 3XXX that cannot use DLSS, no lie given.

By that logic, refer to my GTX 960 analogy above. I can run a test that could prove a GTX 960 outperforms a 4090, then sell it to some idiot on the second hand marketplace who knows nothing about conputers. Here's the test.... see.... 960 wins. It's faster then the 4090 at 1/20th the price.

Nuh-uh, sorry pal. If the tests are not under the same conditions then it's not an accurate comparison, and to display the results accordingly is a lie. I can tie a 300KG stone to Usain Bolt's legs and be like hahahahahaha I'm faster than you!!!!!! Lol. Sorry pal. Exact same conditions or the test is a sham and a lie.

 

[USAFRet]

NVidia's tests were conducted in raytracing mode. People actually do play games in raytracing mode. How often do you see people driving cars without tires?

And how did the 3 series do in raytracing mode?

 

[USAFRet]

By that logic, refer to my GTX 960 analogy above. I can run a test that could prove a GTX 960 outperforms a 4090, then sell it to some idiot on the second hand marketplace who knows nothing about conputers. Here's the test.... see.... 960 wins. It's faster then the 4090 at 1/20th the price.

Nuh-uh, sorry pal. If the tests are not under the same conditions then it's not an accurate comparison, and to display the results accordingly is a lie. I can tie a 300KG stone to Usain Bolt's legs and be like hahahahahaha I'm faster than you!!!!!! Lol. Sorry pal. Exact same conditions or the test is a sham and a lie.

Sorry, but no.

In Condition X, there is a 2% difference between A and B.

In Condition Y, A is 4 times faster than B.

I'm quite sure you can find actual reviews showing Condition X.

Of course, the marketing team only shows you Condition Y.
It is up to you to believe or not.


I have no dog in this fight. I am brand agnostic...AMD, Nvidia, Intel.
I do not have a 4XXX or even a 3XXX GPU.

Truly, I do not care.

But I do know how marketing works.

You can feel free to try to crank up a class action suit, for false advertising.
I would enjoy watching.

 

[YouFilthyHippo]

Sorry, but no.

In Condition X, there is a 2% difference between A and B.

In Condition Y, A is 4 times faster than B.

I'm quite sure you can find actual reviews showing Condition X.

Of course, the marketing team only shows you Condition Y.
It is up to you to believe or not.

Oh, I never said they couldn't lie. I just don't think they should get away with it

 

[USAFRet]

Oh, I never said they couldn't lie. I just don't think they should get away with it

Until there is a judgement in a court of law, they are "getting away with it".

I treat ALL claims from marketing with a very large grain of salt.

 

[MrN1ce9uy]

I don't care how much the performance increase is or isn't (which it's not near what I'd expect), the pricing is so ridiculous even though I have the money and have spent thousands in the past on numerous custom builds, I simply cannot agree with the pricing of these cards. I love custom computers just as much as the next guy, but no gaming part is worth more than $1,000 to me.

 

[hotaru.hino]

If those are the stipulations of the test, then it's not an apples to apples test. Remember, we are comparing the performance of 2 physical components. DLSS is not physical. What is the performance of 2 physical components? An RTX 4090 and an RTX 3090 are identifiers of 2 physical components. When you say "the performance of a 4090 is 4x a 3090", you have stated that physical component 1 can perform better than physical component 2 by a 4x scale. That is objectively incorrect. That is a false statement. The 4090 is getting an unfair advantage from a SOFTWARE component that the 3090 does not have. Now, run the 3090 with DLSS 3 and then we will talk. If it's a fair and equal apples to apples test, they either both run DLSS 3 during the test, or they both don't.

Except there were hardware improves to Ada from Ampere. What makes DLSS3 an Ada only thing is the inclusion of an Optical Flow Accelerator, and this is needed to help the frame generation more accurate to what would actually happen. Also as far as I can tell, DLSS3's only feature was frame generation.

If you still want to say "no, they can't do that", then we should null and void all RT tests because technically the GeForce 10 series can run DXR RT, and it's not fair because it has to do it in pure software while the others get to do it with a hardware accelerator.

When you do a performance comparison of 2 physical components, the playing field must be level. Neither can have an unfair advantage. Both machines must use the exact same components, same game, same settings, same software. This is a true apples to apples performance comparison. In this comparison, the 4090 is about 70% faster than the 3090, not 2-4x

Then this null and voids any comparison between AMD and NVIDIA cards because... guess what? They're using different software: the drivers. Also some games may have performance optimizations for specific GPU architectures, which doesn't make that exactly a fair test either.

Overall sure, NVIDIA saying their GPUs are "up to 4x more powerful" can be construed as misleading, but the wording and their cherry picked data that can be independently verified shields them from any legal action. And should it be on the onus of the seller of the thing to do the "research" that potential customers should be doing? Because they'll just cherry pick those results too.

The customer should always be leery of the claims that any seller makes. This is no different. Let marketing spew whatever BS they want about the supposed performance of the thing because there's always going to be a million factors in the way that prevents me from getting to their ideal result. The only thing they can't BS on are the specifications.

 

[Endymio]

Oh, I never said they couldn't lie. I just don't think they should get away with it

Because they aren't lying. It's right there on NVidia's website. The 4X faster claim is specifically for "CyberPunk 2077, with new RT Overdrive. They then state specifically how that was measured:

"3840x2160 Resolution, Highest Game Settings, DLSS Super Resolution Performance Mode, DLSS Frame Generation on RTX 40 Series, i9-12900K, 32GB RAM, Win 11 x64. All DLSS Frame Generation data and Cyberpunk 2077 with new Ray Tracing: Overdrive Mode based on pre-release builds."

Find a new tree. You're barking up the wrong one here.

 

[DSzymborski]

While the rest of your post was spot-on, allow me to add a small caveat here. Anyone can sue over anything, at any time, legal basis or no. And -- with enough proper venue-shopping -- they may even win. Consider the case of a woman who sued Geico because that company insured a vehicle in which she contracted an STD while having sex in the back seat. She received a $5.2M settlement from arbitration; it was upheld by the appeals court, and, had the MO Supreme Court not chosen to intervene, Geico would now be paying that out -- and charging the rest of us higher premiums in result.

I should note in this case there was an additional complexity. The two people in the car went to arbitration , the arbitrator rules for the woman, and her counsel basically told GEICO to pay up, without GEICO ever being able to intervene. A really odd case, though one that was possible because of the combination of arbitration and the amount of money.