NVIDIA gimping benchmarks: GTX 780 tested {322.21 vs 353.62 vs 376.33 vs. 398.36 vs. 411.70}

Wait for it, this is only the beginning. Soon we'll get RTX games which will probably run on Pascals via DXR fallback producing 2-3 fps instead of 30-40 on Turing - and all the "NV iz gimping mah 1080Ti carrrd" threads will start popping on a daily basis.

On a more serious note though, I do expect Paxwells to start falling back a bit in newer games alongside devs switching their optimization efforts to newer h/w. This can already be seen in some benchmarks where 2080 is considerably faster than 1080Ti - this will likely be the norm in a year or two.

 

There are possibilities of DXR fallback support in official way?
DXR fallback it's an non-official injector as I've understood

 

Why injector? It's just a compute based emulator provided within the DXR API. So if a device doesn't support DXR natively then the code is run through this fallback layer. The only catch here is that the IHV driver must allow this - or to be more precise it must report DXR as being supported.

 

Hmm.. that reminds me, I think my Maxwell 2 is getting gimped by nvidia on purpose to make me buy RTX.

Need another thread STAT!

 

People are already bashing Nvidia for the RTX performance cost, I don't see why Nvidia would spend any resources whatsoever optimizing DXR for Pascal when performance probably l would be in single digits anyway

 

(Tinfoil hat engaged.)

If there's any gimping, then it's going to be in a way that's difficult to test and impossible to prove. Meaning, games that didn't exist before could perform very poorly on previous gen GPUs. You can't test for example how AC Odyssey performed two years ago. It didn't exist. But it does run 70% slower on Maxwell compared to Pascal for example than what is expected. You'd need to investigate the reason why that is, and that would be a lot of work, and way outside the capabilities of most tech youtubers. Is it a driver bug? Driver gimping? "Game Ready" driver fixes/improvements intentionally left out for older GPUs? Game bug? Who knows.

People mostly buy new GPUs for the new games. If you were to gimp old GPUs, you'd do it for new games only, where nobody can test it. You can see the missing performance, but you just can't prove it's due to gimping.

 

I was talking about this wrapper/Injector: https://github.com/Microsoft/DirectX-Graphics-Samples/tree/master/Libraries/D3D12RaytracingFallback

 

Good point. To an extent though, I think probably developers and the GPU driver teams don't do as much work to optimize for the older GPUs as much as they possibly can, like they do for newer ones.
Purely from a business/amount of work to do it type of deal.
You can't officially purchase new stock of those, since manufacturing has been discontinued, so why bother beyond the minimum?
It's not like consoles, where they are stuck with the same hardware for forever, so they have to put maximum effort into getting as much optimization as possible, which is easier when it's one set of fixed hardware I guess.

My main gripe is, that if you buy a GPU that far exceeds that of current consoles. It should be able to run every game that ran on those consoles, at the same settings and same resolution bare minimum. Ever, with better framerates.
That doesn't seem to always happen.

 

the video in the OP is not true at all he show that the GTX 780 can reach 72 FPS on deus ex mankind divided at max settings at 1080p with taa. when the official benchmark of this game here at guru3d the gtx 780 Ti hardly reach 42 fps !!!!

 

wow superhappy to see that. i also wasn't sure if nvidia is doing this crap to old generation gpus or not and i'Ve seen those speculations circle around guru3d for the past few years.. i even kept some drivers for half a year because i wanted to be on the secure side..

now after watching this video i feel supergreat, knowing that nvidia is NOT gimping my 980ti with new drivers. in contrary, its either the same performance or better on new drivers thanks soooo much for this thread and video after 2 years this topic can finally be closed (at least for me personally)

 

Why? There were dozens of benchmarks done on this which have proved without any doubt that they don't.

 

AMD fanboy troll bias topic.

 

Good catch, I noticed the numbers in that vid looked fishy earlier today. OSD but FPS not shown.... non Ti 780 consistently beating 390X?!

 

Nvidia isn't gimping anything. The only thing weird and suspicious I would say they are guilty of is not publicly releasing those rtx and dlss demos. All the reviewers (including Hilbert) can use them but why can't we try them for ourselves? If the results are so great and the demos already exist, why can't we use them already?

 

That's more the developers at fault than the driver team. Nvidia and AMD shouldn't be responsible for fixing the devs crappy code. It's become a normal thing for them though but you can't expect them to continue to optimize drivers for 5 year old hardware.

The last time one of these threads came up somebody was complaining that their Fermi card wasn't getting any performance improvements with each driver.

 

This is an unfounded and meaningless myth.
It is true that some drivers have come out better than others, but to say that Nvidia layer the previous series scratches in the ridiculous.

I have here a pair of 780Ti in SLI and I did a Superposition test 6 months ago, with the drivers 388.84, and this is what they got:



And here is another test I did today with the most recent drivers to date (416.16):



As you can see I haven't lost performance and I even won some minimum and maximum fps.
So rest assured that Nvidia is not slowing down yours old non RTX cards.

Greetings.

pd: don't pay attention to the temperatures, so it happens with the SLI configurations of triple ventilation (no blower, no wc).

 

You really have to apologize to the crap you just thrown at me. Because of your accusations, i have recorded a video about GTX 780 reaching 70 FPS in the specific scene with maximum settings and T-AA, which i showed in my benchmark:



Your linked benchmark, where GTX 780 Ti scored a pathetic 42 FPS average, was probably done in another specific, more demanding scene.

But more important, that test was somehow very unfair to GTX 780 Ti, because i also have GTX 970 stock, and it is equal to GTX 780 OC in Deus Ex Mankind Divided at the settings i showed, in the place i showed with GTX 780 boosting up to 1189 MHz core clock.

@toasty, @Vidik your apologies should be in order too.

I've been 10 years in this forum and presented a lot of benchmarks over the years. If you don't know me, it does not mean you can just accuse me of BS like faking results, get it?

 

It's very common to receive new drivers only to find older ones performed better, AMD experiences the same.
It has nothing to do with gimping, but with added programming in the drivers to "improve" certain games, which in turn can handycap older cards.
It usually takes several driver versions to adjust that also.
But nowadays drivers are pretty consistent, you either gain fps and benchmark scores, or stay around the same.

And after all, those buying expensive 1080TI/2080/2080Ti will definitely want to know if their purchase is running as expected. They know their way around those benchmarks and will report if those are out of place.
Those people not concerned with all of this will most likely not notice degrading performance either. They wont run any bench, or oc for that matter.

It will be the enthousiasts that buy and test Nvidia/Amd cards, and those are the critics both companies respect and fear.
On a side note, back in the days, improved game engines could introduce functions older cards weren't fully capable of, thus newer drivers trying to optimize those cards could fail, and produce the opposite.
I suspect we will see the same when RTX is fully used in all or most games.
Drivers will try to improve performance and thereby handycap those cards that don't support it. We will see many posts claiming to stay with older drivers since the new ones are bad etc etc.

 

Since I don't own the game I look to reputable sites for their numbers to compare, Guru3D's results were already posted:

Spoiler