Is Nvidia deliberately downgrading Kepler performance in favour of Maxwell?

If people report performance drops after driver "updates", then there is something. Those tables from the links are not compelling evidence? WOW!!!
Yep, there's driver bugs if those reports are correct. I only wonder why Nvidia would have timed their alleged Kepler downgrading to roughly coincide with the launch of AMDs latest graphics chips. Do they want to driver green customers to the red team?

If with "tables from the links" you refer to the imgur pictures, i think i have reasonably well explained why the data there is no evidence for anything, beginning with the missing clock speeds.

Those are synthetic benchmarks.
Which are perfectly suited for education about fillrate.

If your theory of Witcher 3 being solely bottlenecked by the fillrate presented in synthetic benchmarks was true, then the GTX 680/770 cards would be faster than the 780, which is not.

How so? Please, do read up on bottlenecks on Fermi/Kepler/Maxwell. HINT: The former two have a pixel fillrate that's not bottlenecked by ROP count, but by shader export rate, thus scaling with the number of active SMs.

Besides, these aren't huge resolutions. This is 1080p, so fillrate requirements shouldn't be exactly stellar.
Plus, it doesn't explain the Kepler's performance deficit on Project Cars and Arkham Knight.
Another view on Arkham Cars:

Project Cars - retested and with new drivers [Benchmarks configurable via dropdown]
http://www.pcgameshardware.de/Project-CARS-PC-238576/Specials/Benchmark-Test-1158026/#a3

Batman with recent drivers [Benchmarks configurable via dropdown]
http://www.pcgameshardware.de/Batman-Arkham-Knight-PC-258399/Specials/Technik-Test-1162745/#a5




Delta colour compression makes a surprising difference in some games.
Which has been in operation apparently since Fermi, only been upgraded in Maxwell. :)
 
Last edited:
GK110 wasn't really built as a pixel pushing monster and at the same time it carries a lot of GPGPU "dead weight" logic. The fact is that Nvidia released the small GK104 first, as a high-end product, while the GK110 was much more carefully pushed into the same market, while already flooding the HPC clients.
 
Also 9/11 was an inside job and moon landing was obviously fake.

I went from a GTX670 to a 970 and have to say I was super surprised at how much better it is. The difference is way more than I expected since they are both 28nm chips.
 
Yep, there's driver bugs if those reports are correct. I only wonder why Nvidia would have timed their alleged Kepler downgrading to roughly coincide with the launch of AMDs latest graphics chips. Do they want to driver green customers to the red team?

I guess there are a lot of people who buy only nvidia, no matter what

Also 9/11 was an inside job and moon landing was obviously fake.

I went from a GTX670 to a 970 and have to say I was super surprised at how much better it is. The difference is way more than I expected since they are both 28nm chips.

I love how people jump fast on conclusions to liquidate something they dont like.

Thanks for letting us know the difference from 670 to 970 was way more than you expected.
 
On the face of it I see the ever present 'what is best driver version for game X' that exists for NV since forever.
Radeons the answer is nearly without exception 'the latest Catalysts'.

May be simply a question of architectural difference.
 
Also 9/11 was an inside job and moon landing was obviously fake

Offtopic: It is not fake because if it were, the people of Russia would have noticed and let the world know. This is the first.
The second somehow is different - yes, it was kind of inside job because they have AirForce which on that day did nothing. Also, there is not sufficient evidence in my eyes that the plane which "hit" the Pentagon, was actually a plane........

I went from a GTX670 to a 970 and have to say I was super surprised at how much better it is. The difference is way more than I expected since they are both 28nm chips.

I do not know what you had expected but these cards are very different - the chip GM204 is bigger and stronger. It is natural to be quite a bit faster overall.
 
Also 9/11 was an inside job and moon landing was obviously fake.

I went from a GTX670 to a 970 and have to say I was super surprised at how much better it is. The difference is way more than I expected since they are both 28nm chips.

For be honest, the 670 is an old gpu's, first line of Kepler, ( excellent gpu's on price/perf ratio at this time ) you have got 760TI - 770-780 then, ofc it is a big jump.
 
I went from a GTX670 to a 970 and have to say I was super surprised at how much better it is. The difference is way more than I expected since they are both 28nm chips.
I wonder if your jump was bigger than mine -- from GTX580 to 780Ti? :p

BTW, I could soon jump yet once more, to 980Ti... yay!
 
On the face of it I see the ever present 'what is best driver version for game X' that exists for NV since forever.
Radeons the answer is nearly without exception 'the latest Catalysts'.

May be simply a question of architectural difference.

In my experience it is not that complicated for Nvidia. The answer is almost always 'the latest WHQL'.
 
He's not the one jumping to conclusions. The "something" doesn't actually exist.

But the something might've existed. And if it had existed, it could've been because of nefarious intent.

So Nvidia is surely guilty of leaving open the possibility of impropriety. I demand an Internet petition on the double.
 
He's not the one jumping to conclusions. The "something" doesn't actually exist.

I guess if it had happened a couple of years before, he probabily would have added also the Nsa to his list.

The guy on overclock.net points out as in FC3 680 and 7970GHz were on pair (70fps), as they were more or less at the begin

Same website, same game engine (updated)

http://www.techspot.com/review/615-far-cry-3-performance/page3.html

http://www.techspot.com/review/917-far-cry-4-benchmarks/page3.html

and yet 7970GHz is 40% faster in FC4 (45 vs 63fps, +18)

The delta is quite huge IMHO.

I am not saying something does exist, but I wouldn't put my hand on fire saying it doesn't

The same guy also put in comparison games before and later Maxwell


Games tested from post Maxwell
Alien Isolation, Call of Duty Advanced Warfare, Civilization BE, Dragon Age Inquisition, Ryse Son of Rome, Shadow of Mordor.

R9 290: 100%
GTX 780: 82.88%
R9 280x: 80.06%
GTX 770: 68.72 %
GTX 960: 65.34 %


Pre Maxwell games
Battlefield 3 and 4, Crysis 3, GRID2, Tomb Raider, Batman Arkham Origins, Bioshock Infinite, Metro Last Light, Dead Rising 3

R9 290: 100%
GTX 780: 93.73%
R9 280x: 79.93%
GTX 770: 78.39 %
GTX 960: 65.54%


Most interesting to note how the performance of the GTX 960 is almost unchanged on average pre and post. The omission of UBI games doesnt apparently alter the results.

It is also very interesting looking that also the relative performances of the 280x remained the same.

So, if we would try to find a reason for that, one could say that it is a combination of the three possible theories that pjbliverpool explained very well. Anyway it is certainly a curious coincidence how all of them affect at the same relative way the 290, the 280x and the 960.
 
GTX580 is pretty old now and only has 1.5GB. GTX980Ti is much faster.

Of course if he waits I'm pretty sure something faster will come out at some point :smile: but the 980Ti with 6GB and full DX12 compliance seems like a great choice going forward.
 
@elect it could simply be that the AMD hardware is more forward looking; I believe that has been noted elsewhere on this site. From personal experience my X1900XT really did well compared to it's competition in games released in the next few generations of hardware. Of couse one might attribute that to nVidia gimping their own hardware on purpose, but it seems unlikely.

Its easily possible to cause different kinds of bottlenecks on different games using the same engine. Add some more complex effects here, optimize something else there. The loads could be quite different.
 
The 780Ti was simply a good deal I grabbed in an impulse, but I can still get good money for it and jump in the DX12 features bandwagon just in time with the Win10 release. Plus, it will come handy for the new 1440p monitor.
 
Yes, too bad I didn't bookmark the smoking gun of showing the decrease in performance of kepler with newer drivers. But maxwell also is rather different arch with many feature level enhancements so that games meant for the latter won't run as fast on kepler.

I guess if it had happened a couple of years before, he probabily would have added also the Nsa to his list.

The guy on overclock.net points out as in FC3 680 and 7970GHz were on pair (70fps), as they were more or less at the begin

Same website, same game engine (updated)

http://www.techspot.com/review/615-far-cry-3-performance/page3.html

http://www.techspot.com/review/917-far-cry-4-benchmarks/page3.html

and yet 7970GHz is 40% faster in FC4 (45 vs 63fps, +18)

The delta is quite huge IMHO.

I am not saying something does exist, but I wouldn't put my hand on fire saying it doesn't

The same guy also put in comparison games before and later Maxwell


Games tested from post Maxwell
Alien Isolation, Call of Duty Advanced Warfare, Civilization BE, Dragon Age Inquisition, Ryse Son of Rome, Shadow of Mordor.

R9 290: 100%
GTX 780: 82.88%
R9 280x: 80.06%
GTX 770: 68.72 %
GTX 960: 65.34 %


Pre Maxwell games
Battlefield 3 and 4, Crysis 3, GRID2, Tomb Raider, Batman Arkham Origins, Bioshock Infinite, Metro Last Light, Dead Rising 3

R9 290: 100%
GTX 780: 93.73%
R9 280x: 79.93%
GTX 770: 78.39 %
GTX 960: 65.54%


Most interesting to note how the performance of the GTX 960 is almost unchanged on average pre and post. The omission of UBI games doesnt apparently alter the results.

It is also very interesting looking that also the relative performances of the 280x remained the same.

So, if we would try to find a reason for that, one could say that it is a combination of the three possible theories that pjbliverpool explained very well. Anyway it is certainly a curious coincidence how all of them affect at the same relative way the 290, the 280x and the 960.

BF3 and BF4 run better on nvidia hardware, same with bioshock, crysis 3 at least; the post-maxwell games otoh include ryse and shadow of mordor which favor AMD, so not really a fair comparison.
 
Back
Top