that is the point of driver optimization, just look at the witcher 3, before and after kepler patch.
One game and one engine are not enough to know anything for sure.
As for GCN getting faster than Kepler, how can we be sure it's not the game developers who are not coding and optimizing their newer games for Kepler anymore?
There are far more variables than just drivers.
At least you've made it easy. Thanks.
that is the point of driver optimization, just look at the witcher 3, before and after kepler patch.
... and in some other cases it's the game itself that is running badly, like the case of Shogun 2 having performance issues on Fermi cards, back in the day, which was fixed by a game patch.
My point was, it's not always the driver.
Besides, it wasn't as if witcher 3 was running good already and needed 1-4% better performance. No, it was running badly and clearly had issues that needed fixing.
Such major performance issues get fixed as long as the card in question is supported in the driver.
Specific driver optimization in order to extract every single bit of performance are another matter though.
but still the point is not every developer have budged(or in case of gamework like gimping) to fix or to maximize performance and thats the point of driver optimization.
Huh?! You cannot fix a badly coded game with a driver, or vice versa. These are separate things.
you can actually, did you ever read driver log, and usually you can see what the new driver fixed some glitch or add some performance in certain games.
sometimes the dev isn't allowed to even optimize for team red.
Welcome to nV's GameBarelyWorks program where AMD is cut off from source code for optimizing driver code.Was this ever proven? NVidia and AMD work with developers to optimize the games to run their best on their respective cards, but to actually prohibit the developer from optimizing for the other company's cards? That's a stretch. It might even be illegal.
Welcome to nV's GameBarelyWorks program where AMD is cut off from source code for optimizing driver code.
That's because those issue are in the driver, not in the game. If there is a performance bug in the game itself, it must be fixed by the game's developer. I even gave you an example.
Was this ever proven? NVidia and AMD work with developers to optimize the games to run their best on their respective cards, but to actually prohibit the developer from optimizing for the other company's cards? That's a stretch. It might even be illegal.
how even you sure it always driver ? and not because that game calling some specific feature that card lacked and causing bug like async in nvdia ?
that is why driver specific support was critical because not every dev know every card capabilities.
i read it somewhere but I'm forget where i read it, but still gamework is blackbox and only nvdia know the source code.
Is that current? I know it's from AT but a Link to source would help. Interesting results there if that's the latest for that game.
I can surely understand Vulkan patched games being faster on AMD hardware, but this may indicate that's not always the case. I guess there can be exceptions under any circumstances.
Tell that the Kepler usersNVidia and AMD work with developers to optimize the games to run their best on their respective cards, but to actually prohibit the developer from optimizing for the other company's cards? That's a stretch. It might even be illegal.
Maxwell likely won't be as "handicapped" as Kepler in some of the newer games versus Pascal, but that fragmented memory on the 970, oh boy. Time will tell what's going to happen to the best-selling gaming card... or was it 960 ...zlatan said:Kepler is generally a strong architecture when the shaders are optimised for ILP. In this case all the cores can be used in any Kepler product. The only problem when you don't optimize your shaders for independent workloads, so Kepler will loose 33 percent of the theoretical performance. This is the case with GameWorks. These shaders just use 67 percent of the Kepler GPUs, so 33 percent of your shader cores will always be in idle.
Optimising for Kepler is really easy. There might be some workloads when it is not practical to find some independent operations for the idle cores, because of the register pressure, but many times this is not a problem.
The problems are the licences. With some specific codes NV won't allow the devs to optimize the shaders. They can see the source, but they don't able change it, and the original code will hurt Kepler. This is why some games don't run well on these products. Basically the licences won't allow the performance optimization.
It's not current. Here's a much more recent test under Vulkan, although it was done in Linux. Vulkan in Dota2 is still the slower render path so I don't think it's of much benefit as it's still an experimental release, but if interested:
http://www.phoronix.com/scan.php?page=article&item=amdgpu-rx480-linux&num=11
Tell that the Kepler users
My point wasn't exactly NV versus AMD, but NV versus NV. Nvidia finds a smart way to cut the older gpus support to boost the sales of its newer hardware. GameWorks/VRAM gimping/etc are merely just tools. But yeah, carry on with your specific discussionThis is about gameworks only, right?
The discussion was about the claims that nvidia can prohibit a developer from optimizing the ENTIRE source code of a game for AMD cards, not just locking the gameworks code, which is highly doubtful.
Throughout the years there have been nvidia sponsored games that run faster on AMD cards and vice versa.