direct x 12 is very promising...for AMD

[3DVagabond]

Doesnt make sense to use AMD's Direct3D drivers to show how "great" Mantle is when the competition can even beat or come close to Mantle with Direct3D. :|

I think you are heavily generalizing things. Just look at the facts, Mantle does affect performance. No one denies that.

The post above seems to insinuate otherwise.

 

[sontin]

No, it doesnt.

In every Mantle game, nVidia is much faster than AMD under Direct3D. Which implies that AMD doesnt care to optimize the Direct3D path - great service for their own non Mantle costumer btw.

 

[Spjut]

No, it doesnt.

In every Mantle game, nVidia is much faster than AMD under Direct3D. Which implies that AMD doesnt care to optimize the Direct3D path - great service for their own non Mantle costumer btw.

And for games that do use Mantle, what reason does AMD have to do specific optimizations for the Direct3D path in those games?

Optimizing the Direct3D path for Mantle games would just be a waste of resources. And although one could argue that the pre-GCN GPUs don't support Mantle, the optimizations for those GPUs wouldn't automatically carry over to AMD's GCN GPUs either

 

[AtenRa]

No, it doesnt.

In every Mantle game, nVidia is much faster than AMD under Direct3D. Which implies that AMD doesnt care to optimize the Direct3D path - great service for their own non Mantle costumer btw.

You have data to prove that non Mantle GPUs suffer from Mantle optimized drivers ???

 

[witeken]

The post above seems to insinuate otherwise.

The benchmarks prove otherwise.

 

[3DVagabond]

No, it doesnt.

In every Mantle game, nVidia is much faster than AMD under Direct3D. Which implies that AMD doesnt care to optimize the Direct3D path - great service for their own non Mantle costumer btw.

Well, now you are making a different claim to what you did.

Doesnt make sense to use AMD's Direct3D drivers to show how "great" Mantle is when the competition can even beat or come close to Mantle with Direct3D. :|

You quite clearly said that "the competition (nVidia, it can't be anyone else) can even beat or come close to Mantle with D3D. Not that AMD doesn't care about optimizing the the D3D path for their own non Mantle customers. Not that it matters because both of those statements are FUD. AMD is still optimizing their drivers for DX regardless to whether there's a Mantle path or not, and if they can meet/beat them in D3D it only gets worse with Mantle.

Here the reference 290 is ~5% faster than the 780 using D3D. And that's @1080. Not the best case resolution for Hawaii vs. GK110. All of the cards in this graph fall pretty much right into their typical performance brackets.

The benchmarks prove otherwise.

Yes, they do.

 

[SirPauly]

And for games that do use Mantle, what reason does AMD have to do specific optimizations for the Direct3D path in those games?

Optimizing the Direct3D path for Mantle games would just be a waste of resources. And although one could argue that the pre-GCN GPUs don't support Mantle, the optimizations for those GPUs wouldn't automatically carry over to AMD's GCN GPUs either

There could be a potential backlash from the market where a consumer may feel that AMD is focusing too much on Mantle and not enough focus on DirectX. It wouldn't be a waste of resources if the Direct3d optimizations are not just game specific. Another potential backlash would be that the consumer may feel that AMD needs a proprietary API just to compete.

 

[sontin]

And for games that do use Mantle, what reason does AMD have to do specific optimizations for the Direct3D path in those games?

Optimizing the Direct3D path for Mantle games would just be a waste of resources. And although one could argue that the pre-GCN GPUs don't support Mantle, the optimizations for those GPUs wouldn't automatically carry over to AMD's GCN GPUs either

Then AMD should stop talking about the improvements when 80% of it comes from not improving Direct3D.

http://pclab.pl/art58492-4.html

 

[Spjut]

Then AMD should stop talking about the improvements when 80% of it comes from not improving Direct3D.

http://pclab.pl/art58492-4.html

And as I said before, there is no reason for AMD to optimize the Direct3D path for Mantle games.

There can be both general and game specific optimizations for Direct3D. But if you want to prove a case about the Direct3D path being neglected, post benchmarks of games that don't have Mantle.

 

[sontin]

And i followed that there doesnt exist a reason to compare AMD's D3D driver to Mantle then.

We know that AMD doesnt optimize their D3D driver (overhead...) any further.

 

[AtenRa]

Then AMD should stop talking about the improvements when 80% of it comes from not improving Direct3D.

http://pclab.pl/art58492-4.html

And i followed that there doesnt exist a reason to compare AMD's D3D driver to Mantle then.

We know that AMD doesnt optimize their D3D driver (overhead...) any further.

R9 290X with DX11 is faster than GTX780ti,
R9 280X with DX11 is faster than GTX770.

Shocking

http://gamegpu.ru/action-/-fps-/-tps/sniper-elite-3-test-gpu.html

 

[caswow]

there is no actual dx11 overhead driver. nvidia has to optimise their dx11 game drivers for each game. you can almost read in every international forum posts made by nvidia fanboys about the recent driver update beeing a dx11 wonderdriver. this is simply not true.

 

[monstercameron]

and that is why mantle/dx12 will be good, driver optimizations will play a lesser role going forward. Also less driver updates and possibly more stable drivers.

 

[ShintaiDK]

and that is why mantle/dx12 will be good, driver optimizations will play a lesser role going forward. Also less driver updates and possibly more stable drivers.

We still have to see that. Right now Mantle is a per game driver basicly. Mantle and DX12 is mainly to reduce the CPU overhead. The optimizations we have seen in DX previously have been speedup of the GPU part. So as it stands, there is no reason to think it wotn be the same with Mantle and DX12.

 

[monstercameron]

We still have to see that. Right now Mantle is a per game driver basicly. Mantle and DX12 is mainly to reduce the CPU overhead. The optimizations we have seen in DX previously have been speedup of the GPU part. So as it stands, there is no reason to think it wotn be the same with Mantle and DX12.

Mantle is so much more than cpu overhead reduction. Its a whole new api.

 

[Carfax83]

there is no actual dx11 overhead driver. nvidia has to optimise their dx11 game drivers for each game. you can almost read in every international forum posts made by nvidia fanboys about the recent driver update beeing a dx11 wonderdriver. this is simply not true.

I don't know about that. NVidia has clearly stated that the driver enhancements were NOT game specific, but general in nature:

NVIDIA claims that these fixes are not game specific and will improve performance and efficiency for a lot of GeForce users. Even if that is the case, we will only really see these improvements surface in titles that have addressable CPU limits or very low end hardware, similar to how Mantle works today.

Source

From my own experiences with these drivers, I'd say they are definitely not game specific. The reason why I say so is because I noticed performance increases in older games as well, such as BL2... Now BL2 is DX9 and not DX11, but it goes to show that NVidia has spent time optimizing not only DX11 functions, but DX9 functions as well..

 

[BrightCandle]

One thing that bothers me about DX12 is how close to Mantle it actually gets. One of the "features" of Mantle is that the use of 2 graphics cards is no longer transparent to the game, so you need to actively write support for Crossfire into your game. You have the ability to use both cards for one frame, but if you just want to use AFR you need to code it in. That would mean the loss of frame pacing with Mantle unless each company developed it themselves, and as we know that took AMD over 1.5 years and they never supported all the possibilities. Equally support for eyefinity would presumably require special support from the developers.

I do wonder if DX12 intends to copy this (anti) feature or not. It sounds like a very bad thing in general, something that would increase development costs and decrease the utility of multiple cards dramatically but presumably its still too early to know.

 

[Final8ty]

We still have to see that. Right now Mantle is a per game driver basicly. Mantle and DX12 is mainly to reduce the CPU overhead. The optimizations we have seen in DX previously have been speedup of the GPU part. So as it stands, there is no reason to think it wotn be the same with Mantle and DX12.

Mantle is not a Per game driver at all, all improvements to any particular game with mantle thus far has come from the game developer through game patches.

 

[KingFatty]

No, it doesnt.

In every Mantle game, nVidia is much faster than AMD under Direct3D. Which implies that AMD doesnt care to optimize the Direct3D path - great service for their own non Mantle costumer btw.

Can you clarify the logical path here? I don't follow how nVidia's performance can imply something about AMD?

Is there any logically sound way to show that AMD is intentionally causing their Direct3D path to be worse than it could be?

I'm just thinking it's a really big jump to think that AMD would intentionally sabotage itself, just to try to make Mantle look better. That seems like a huge leap without any support.

 

[3DVagabond]

Can you clarify the logical path here? I don't follow how nVidia's performance can imply something about AMD?

Is there any logically sound way to show that AMD is intentionally causing their Direct3D path to be worse than it could be?

I'm just thinking it's a really big jump to think that AMD would intentionally sabotage itself, just to try to make Mantle look better. That seems like a huge leap without any support.

There are no facts at all backing this. Graphs have been shown that demonstrate Mantle games run right where you would expect them to on AMD (*and nVidia as well) hardware with the DX pathway. This whole position makes no sense.

*So far Mantle games have shown the ability to be highly optimized on the DX pathway for nVidia. Think about it. If a company decided to knowingly leave optimizing out for their own customers would they allow for the game to support their competitor's optimizations?

 

[tweakboy]

Very promising that is. thx

 

[krumme]

One thing that bothers me about DX12 is how close to Mantle it actually gets. One of the "features" of Mantle is that the use of 2 graphics cards is no longer transparent to the game, so you need to actively write support for Crossfire into your game. You have the ability to use both cards for one frame, but if you just want to use AFR you need to code it in. That would mean the loss of frame pacing with Mantle unless each company developed it themselves, and as we know that took AMD over 1.5 years and they never supported all the possibilities. Equally support for eyefinity would presumably require special support from the developers.

I do wonder if DX12 intends to copy this (anti) feature or not. It sounds like a very bad thing in general, something that would increase development costs and decrease the utility of multiple cards dramatically but presumably its still too early to know.

Its about making the api transparent to the game engine devs.

Its needed for the future high-perf games, and i actually think its more about keeping dev. cost down, because optimizing for prior DX to get best perf is a pain. (dx as a black box argument)

And its driven by a major consolidation of engines used ->fewer engines & cross platform compatability. It makes perfect sense and thats why everyone and his brother is running the same way now.

And its not like NV, Intel and MS would have preferred it otherwise. The old dx standard is protecting their existing position, so there must have been a great pressure for devs for that to change. AMD obviously therefore have different interest so thats why they were the one developing the clean sheet API with the leverage of the consoles. The development of DX12 already makes Mantle a major strategic win.