PCPer Tests 3DMark Timespy: Pascal Does have Async Compute

[Briggsy]

I've updated the OP to show how many points each card improved by in the benchmark. I think its far more indicative of real world gains in FPS.

 

FuryX: 652 points 

Nano: 496 points 

GTX 1080: 471 points

RX 480: 335 points

GTX 1070: 305 points

1080 SLI: 164 points

[patrickjp93]

51 minutes ago, HalGameGuru said:

But, that's not a driver handling its business that's bad code breaking the driver requiring more effort. In a perfect world, where the driver was that robust a bad dev couldn't break the driver or ruin performance on their end, the driver would cope out of the box. The question is is that a limitation of the driver or a limitation of the hardware? If the driver is compensating for a hardware issue (shader incompatibility) that strikes as more a hardware issue the driver is having to compensate for than the driver having an issue of its own, but if the driver itself cannot handle something it was meant to, or having not been written to be able to stably handle the hardware at its full utilization, etc.?

It's a limitation of the driver. You can't scan a whole game for buggy code in real time. Nvidia finds out what is broken and then creates a driver workaround. That's how their drivers stay thin while fixing broken code.

[Humbug]

 

http://www.guru3d.com/articles-pages/futuremark-3dmark-timespy-benchmark-review,1.html

[Khvarrioiren]

15 hours ago, patrickjp93 said:

Just counting AAA being released in 2016.

9 AAA games on DX12, one AAA game on DX11 which got a DX12 patch.

[laminutederire]

10 hours ago, patrickjp93 said:

It's a limitation of the driver. You can't scan a whole game for buggy code in real time. Nvidia finds out what is broken and then creates a driver workaround. That's how their drivers stay thin while fixing broken code.

That's funny, all the computer science graduate where I am think their drivers aren't that thin and particularly don't think nvidia are the maestro that you describe, biased much?

 

Anyway that begins to be more and more off topic to me, going into an amd - nvidia war which isn't what the thread is going for.

[Humbug]

28 minutes ago, laminutederire said:

That's funny, all the computer science graduate where I am think their drivers aren't that thin and particularly don't think nvidia are the maestro that you describe, biased much?

 

Nvidia does have plenty of driver issues yes.

but they are good at minimizing CPU overhead on traditional graphics APIs, in that particular aspect of the drivers yes they are good at what they do.

[DXMember]

9 minutes ago, Humbug said:

Nvidia does have plenty of driver issues yes.

but they are good at minimizing CPU overhead on traditional graphics APIs, in that particular aspect of the drivers yes they are good at what they do.

they are good at tesselating Batman's cape

 

[Humbug]

5 minutes ago, DXMember said:

they are good at tesselating Batman's cape

 

I am not a supporter of wasteful overtesselation, but I was talking about the CPU overhead on their drivers so not really relevant.

[laminutederire]

28 minutes ago, Humbug said:

Nvidia does have plenty of driver issues yes.

but they are good at minimizing CPU overhead on traditional graphics APIs, in that particular aspect of the drivers yes they are good at what they do.

Well that's the point, we can see they seem to limit cpu overhead, but nobody really knows what's inside their drivers.

Whereas AMD drivers are more open. When they have a new technology they provide source code and explanations, which explains also why game works run like shit on AMD cards but all AMD technology run fine on Nvidia's card. And to me that is also why nvidia drivers are better than AMD's, since if AMD would do better nvidia would know how they did it and could reajust and the contrary isn't possible.

But anyway I digress.

[DXMember]

10 minutes ago, laminutederire said:

Well that's the point, we can see they seem to limit cpu overhead, but nobody really knows what's inside their drivers.

Whereas AMD drivers are more open. When they have a new technology they provide source code and explanations, which explains also why game works run like shit on AMD cards but all AMD technology run fine on Nvidia's card. And to me that is also why nvidia drivers are better than AMD's, since if AMD would do better nvidia would know how they did it and could reajust and the contrary isn't possible.

But anyway I digress.

PureHair?

[laminutederire]

6 minutes ago, DXMember said:

PureHair?

Probably. They publish so much things that I lost count of. They make the world progress (with GPUOpen etc) and only fanboys can deny that.

[DXMember]

3 minutes ago, laminutederire said:

Probably. They publish so much things that I lost count of. They make the world progress (with GPUOpen etc) and only fanboys can deny that.

Umm... Pure Hair is nVidia tech which was built form an open source project TressFX 2.0 that was published by AMD, and the Pure Hair also runs like crap on AMD, it's still better than HairWorks but still crap

[laminutederire]

2 minutes ago, DXMember said:

Umm... Pure Hair is nVidia tech which was built form an open source project TressFX 2.0 that was published by AMD, and the Pure Hair also runs like crap on AMD, it's still better than HairWorks but still crap

You make my point, nvidia often takes what AMD publishes to take the best out of it. Explaining also their better softwares.

[Humbug]

45 minutes ago, laminutederire said:

Well that's the point, we can see they seem to limit cpu overhead, but nobody really knows what's inside their drivers.

Whereas AMD drivers are more open. When they have a new technology they provide source code and explanations, which explains also why game works run like shit on AMD cards but all AMD technology run fine on Nvidia's card. And to me that is also why nvidia drivers are better than AMD's, since if AMD would do better nvidia would know how they did it and could reajust and the contrary isn't possible.

I think you are mixing up the concept of device drivers with that of middleware solutions like gpu-open and gameworks (which devs can inject into their games) for special effects.

[laminutederire]

3 minutes ago, Humbug said:

I think you are mixing up the concept of device drivers with that of middleware solutions like gpu-open and gameworks (which devs can inject into their games) for special effects.

Well drivers adapt themselves to Middleware solutions, so it's easier to adapt to open source Middleware than it is for proprietary Middleware 

[patrickjp93]

5 hours ago, Khvarrioiren said:

9 AAA games on DX12, one AAA game on DX11 which got a DX12 patch.

Funny, I count 7 on DX 11 with DX 12 patches, 3 in DX 12, and 1 in OpenGL with a Vulkan patch...

 

5 hours ago, laminutederire said:

That's funny, all the computer science graduate where I am think their drivers aren't that thin and particularly don't think nvidia are the maestro that you describe, biased much?

 

Anyway that begins to be more and more off topic to me, going into an amd - nvidia war which isn't what the thread is going for.

The driver itself is more lines of code than the Windows 10 operating system, but, only a small portion of it is running at any one time. In that aspect, though it is a tree with many branches, the central trunk isn't that tall, and the branches aren't that long. Game performance speaks for itself, and for the number of recent bugs in Nvidia's drivers, there have been many more in AMD's in 2014/2015. It's not bias. It's just objectively comparing against the competition.

[laminutederire]

32 minutes ago, patrickjp93 said:

Funny, I count 7 on DX 11 with DX 12 patches, 3 in DX 12, and 1 in OpenGL with a Vulkan patch...

 

The driver itself is more lines of code than the Windows 10 operating system, but, only a small portion of it is running at any one time. In that aspect, though it is a tree with many branches, the central trunk isn't that tall, and the branches aren't that long. Game performance speaks for itself, and for the number of recent bugs in Nvidia's drivers, there have been many more in AMD's in 2014/2015. It's not bias. It's just objectively comparing against the competition.

Yeah but it shouldn't be that way. Games should be optimized by their developers with a real sense of what's going on, like Vulkan games like Doom for instance.

[Khvarrioiren]

1 hour ago, patrickjp93 said:

Funny, I count 7 on DX 11 with DX 12 patches, 3 in DX 12, and 1 in OpenGL with a Vulkan patch...

 

Mind listing some of the DX11-only titles, since it seems our definition of AAA seems to vary greatly. I actually forgot about FC:Primal and Cawadoody.

 

[patrickjp93]

32 minutes ago, laminutederire said:

Yeah but it shouldn't be that way. Games should be optimized by their developers with a real sense of what's going on, like Vulkan games like Doom for instance.

Games programmers often are nowhere near as skilled as programmers for other high performance programs, and they're certainly not as thorough. IDTech is an exception to this rule because it's not beholden to shareholders and management to rush a game out. ID makes most of its money off of seminars and programming camps.

 

4 minutes ago, Khvarrioiren said:

 

You never specified DX11-only. That said, the games will launch as DX 11 and get patched later. It's the nature of the beast. DX 12 is still premature.

[Khvarrioiren]

1 minute ago, patrickjp93 said:

You never specified DX11-only. That said, the games will launch as DX 11 and get patched later. It's the nature of the beast. DX 12 is still premature.

By that logic, we'd still call most of those releases DX9 games, patched or not.

[laminutederire]

8 minutes ago, patrickjp93 said:

Games programmers often are nowhere near as skilled as programmers for other high performance programs, and they're certainly not as thorough. IDTech is an exception to this rule because it's not beholden to shareholders and management to rush a game out. ID makes most of its money off of seminars and programming camps.

 

You never specified DX11-only. That said, the games will launch as DX 11 and get patched later. It's the nature of the beast. DX 12 is still premature.

Maybe, but doesn't that frighten you that game developer aren't good at programming excepting a few of them?

[themaniac]

damn, ran time spy and my result is 

[patrickjp93]

1 hour ago, Khvarrioiren said:

By that logic, we'd still call most of those releases DX9 games, patched or not.

No, because DX 11 and DX 9 are not compatible.

[Khvarrioiren]

26 minutes ago, patrickjp93 said:

No, because DX 11 and DX 9 are not compatible.

True, but nearly every DX11 game has a DX9 mode.

[patrickjp93]

1 minute ago, Khvarrioiren said:

True, but nearly every DX11 game has a DX9 mode.

Source?