Is DirectX 12 really going to take over?

[Hi-Fi Man]

It seems every game release with DirectX 12 has been problematic or divisive in some way. Since the onus is now all on developers to do things that were traditionally done by the driver there seems to be a certain level of skill is required to even get a benefit out of the API and not every developer has the skills or resources.

Is it really better to have a low level API and shovel more work on developers and hope they get everything right and come out with a performance boost? Or have the vendor do most of the work with their driver and give the developer more time to do other things?

Another concern is about architecture. GPUs do eventually change architecture and when they do will all the optimizations the developer did for the then current GPUs apply to the new GPUs with new architectures? There is evidence already to suggest no. GCN 1.2 didn't do so hot in mantle with BF4 defaulting to D3D instead of mantle on GCN 1.2 cards. Not sure if that was ever fixed but if it was I don't think that will be the case for every game that uses a low level API.

Will DirectX12 be relegated to DirectX10 like status? Used but not the main API?

 

[Samwell]

I'm not so negative about DX12 as you as DX12 is still in an early status and the programmers need to learn a lot how to use it properly. The bigger commercial studios will all find their way to DX12 at some point, but DX11 will definitely coexist as many smaller studios won't bother with DX12, because it's too complicated and costs too much to develop good DX12 stuff. I think most bigger developers are waiting for Shader Model 6 to be released, as this is also one of the major advancements of DX12. Devs are pretty enthusiastic about SM6 as far as is read, i'm just not sure when it will be released. Next year we should see better running DX12 titles which really use the api.

My biggest concern is the same as yours, i hope that new architectures won't be used fully but at least run the games properly. It would be pretty bad for PC gaming if new Architectures would lead to crashes and run like crap in older games.

 

[TheELF]

Samwell,I believe the opposite will happen,devs are forced to use low level for the consoles anyway so smaller devs will only use dx12,less cost for porting games over and [strike]smaller[/strike] devs don't care about pc performance anyway,there are sooooo many games that run like crap even with dx11.
Only big devs will keep on providing an additional dx11 path, for the same reason they provide unlocked FPS or ultra settings on PC.

 

[Piroko]

It took roundabout six years for DX9 to be completely replaced by DX11. Along its way many were sceptical because of how it forced you to update from XP to 7. Here's an article from a couple years ago:
http://semiaccurate.com/2012/03/14/the-state-of-directx-11-gaming/
Yes, you're reading that right. It took well over two years until we had 33 games with DX11 support. Developers of RTS games spearheaded DX11 adoption along with the Frostbite engine. Sounds familiar? It's way too early to see if DX12 fails.

As for hardware support, I counter with trends. Modern GPUs are increasingly capable of general compute tasks and reprogrammability. I don't see a sign that a GPU of generation +2 does something so immensely different to the current generation that it won't run current gen games at superior fps.
Tweaks will always be neccessary to get the best performance out of a specific card, but I fear there's this general misconception over DX12 in that it locks Nvidia and AMD out of optimizations from their end. That just isn't true. DX12 is very far from Assembler code, it's just lower level compared to DX11 with a direct access to memory. Replacing shader programs and in-flight command restructuring is still a thing.

 

[whm1974]

Myself, I am hoping that Vulkan takes over. However I am a Linux Gamer.

 

[Azix]

yes. Lots of games coming out with it just around a year after its release looks good for its adoption.

 

[maddie]

If not DX12 or Vulkan, then we are looking at a gaming plateau. The large disparity between CPU and GPU advancements make it inevitable that both Vulkan and DX12 will have to be used if gaming advances in physical effects, graphics realism and AI continue.

We still have the sandy bridge processors being adequate for gaming. Can anyone say the same for a GPU of that era?

Present buyers of top of line Skylake CPUs realistically expect to use these processors for several years going forward. Will anyone say the same for even a 1080 in 3-4 years time?

 

[Innokentij]

Myself, I am hoping that Vulkan takes over. However I am a Linux Gamer.

I agree with you and im a windows gamer.

 

[ConsoleLover]

Of course it will, its inevitable. No one is going to be developing DX11 games 3 years from now for example, so that is just a silly question.

DX12 is literally 10x times better than DX11 if done right. I think many game engines need to be reworked significantly or designed from the ground up to take full advantage of DX12, but once it does then we'll see the incredible benefits of DX12.

Especially with newer and newer hardware, as cpu's start reaching 6 and 8 cores as mainstream, etc...

 

[Phynaz]

Not for years.

No one is going to be developing DX11 games 3 years from now for example

DX9 games are still being developed.

 

[Shivansps]

Samwell,I believe the opposite will happen,devs are forced to use low level for the consoles anyway so smaller devs will only use dx12,less cost for porting games over and [strike]smaller[/strike] devs don't care about pc performance anyway,there are sooooo many games that run like crap even with dx11.
Only big devs will keep on providing an additional dx11 path, for the same reason they provide unlocked FPS or ultra settings on PC.

Devs has been using low level for consoles since forever, but we now seeing what little is that when talking about PC hardware that is not static and detail level is not static either.

The devs where already aiming to get things right for GCN 1.1, now they need to do it for GCN 1.4, while trying to include GCN 1.0, 1.1, 1.2, Kepler, Maxwell, Pascal and Intel intro the mix, in two years they gona need to do it again, then again, then again.

Marketing aside, we needed a better DX11, with proper MT and Async Shader Compute, not a low level api.

When the marketing for DX12 dies, DX12 may die with it if not show an improvement in the next 2 years(thats when the hardware arquitectures will change again on PC).

 

[DidelisDiskas]

I hope it's Vulkan that's going to take over, anything with DX means that gaming will always be mostly windows orientated, while Vulkan has 4 platforms listed already.

 

[ThatBuzzkiller]

Yes, at least for AAA games ...

It's going to be very hard for them to ignore the benefits that DX12 will offer. The performance issues will start to disappear once they focus on DX12 as their main video backend and once shader model 6 comes into play ...

In fact some big publishers like Microsoft, EA, Square Enix, Sega, Take-Two Interactive and Ubisoft have already transitioned or are transitioning to DX12 ...

 

[happy medium]

It will take over when Nvidia says so.
I'm not trying to troll, but dead serious.
Reason? the market leader usually dictates new technology or software.
Pc games that is.

For the near future we will get half ass direct X 12 games that sometimes use some elements of direct x 12 but are built with direct x 11 in mind.

We have seen this before guys.

 

[greatnoob]

This isn't a yes or no question you're asking. As a developer you want:
1) compatible - something that works with a large array of devices, could also be backwards compat.
2) portable - can be ported to many platforms without too much hassle
3) extensible - no need to use hacks to extend functionality
4) adopted standards - a fixed SDK with major versions being released when needed, usually mean a lot of 3rd party library support
5) support - if the hardware doesn't support it (or vice versa) chances are it isn't extensible nor portable

DX12 is not portable and not fully compatible with a large array of devices because of its OS restrictions. Vulkan + a third-party DX9-esque abstraction layer/wrapper would be my (or any indie developers) first choice followed by OpenGL/DX9, DX11 and then finally DX12.

The minor performance regression is very well worth it to any level-headed developer because it is guaranteed performance will improve in the future (combination of developers, H/W and kernel/driver releases) and the code portability you gain by using Vulkan would be tremendous (easily worth > millions of dollars to AAA developers) because now porting a game to/from Android, Windows and Linux would mean one less layer of code to worry about + added perk of easier multi-OS releases.

In terms of DX12 adoption: the tradeoff between a fixed (<DX8) and a programmable pipeline (DX9+) was the performance hit on older hardware and changes in how developers handled the graphics pipeline. The gain was significantly better looking games and much more developer and artist flexibility. DX12 and Vulkan gives even more flexibility and introduces proper non-hacky use of the API (which, again, makes code more portable). Anybody who says DX12 and Vulkan won't have much adoption is deluded: current hardware already supports DX12, future hardware will support it better and medieval APIs like DX11 and OpenGL will gradually be dropped by AAA developers for DX12/Vulkan.

 

[Shivansps]

Yes, at least for AAA games ...

It's going to be very hard for them to ignore the benefits that DX12 will offer. The performance issues will start to disappear once they focus on DX12 as their main video backend and once shader model 6 comes into play ...

In fact some big publishers like Microsoft, EA, Square Enix, Sega, Take-Two Interactive and Ubisoft have already transitioned or are transitioning to DX12 ...

The shift to DX12 is marketing, all those where already working with consoles low level api for years, and also thats considering MS let DX11 die, and seems unlikely considered they just released DX11.4 feature set, they are keeping it alive as a fallback by implementing DX12 features, you know that DX11 can be improved? improving MT support, adding SM6, adding Async Shader compute.... all that can be done, whiout having to add load to developers by forcing them to work on a lower level api and the nightmare that will be DX12 games when we start seeing long term issues, (old games, new hardware) and (new games, old hardware).


Im fully agree that we needed an improvement over DX11, i just dont agree that going to a lower level api is the way, at least no on PC where hardware is not static.

 

[ThatBuzzkiller]

If people believe that one sole IHV like Nvidia are the ones that's going to determine the adoption of a gfx API then they are only kidding themselves when we take a look at past history ...

Nvidia is the one at the mercy of Microsoft and not the other way around. Nvidia itself is easily replaceable when there are other alternatives (AMD/Intel). To Microsoft, all IHVs except Intel are just their puppets since they can decide to drop support for their products at a whim ...

Microsoft has a monopoly when it comes to gfx APIs because of their developer support and the added benefit of being more portable from an older codebase due to AAA developers writing their shaders in HLSL, Nvidia to my knowledge hardly has such benefits ...

If Nvidia ever decides that their not content with their partnership with Microsoft then people shouldn't go around encouraging the misinformation that Nvidia has absolute say when that is totally not the case since their not rolling out their own gfx API ...

Nvidia MUST play by Microsoft's rules and that goes for everyone else ...

 

[ConsoleLover]

If MS allowed DX12 to win7, DX11 will die immediately. No one would use it. DX11 would stay relevant because of MS bullshit restrictions, so for the next 3 years we are still going to be having some DX11 titles, especially from smaller indie developers, but most of those are worthless, they literally sell 1000 copies of their games, if you combined most indie games you wouldn't even reach 1 million in sales.

So its completely irrelevant to the big picture.

 

[Red Hawk]

If MS allowed DX12 to win7, DX11 will die immediately. No one would use it. DX11 would stay relevant because of MS bullshit restrictions, so for the next 3 years we are still going to be having some DX11 titles, especially from smaller indie developers, but most of those are worthless, they literally sell 1000 copies of their games, if you combined most indie games you wouldn't even reach 1 million in sales.

So its completely irrelevant to the big picture.

By that logic, is DX11 going to die immediately simply because Vulkan is available to program for on Windows 7?

 

[Pottuvoi]

As long as DX11 and OpenGL are supported by drivers they will be used as they are high level APIs.
Even though DX12 and Mantle will get all fun features first and more control to your software, there are still developers who do not have resources to use them.

 

[Dribble]

Obviously it will eventually but this is the same with all DX releases...
1. There is much hype in the forums.
2. A load of games say they use the new DX, there is much fanboy excitement, but it hardly makes any difference over the old version.
3. Several years later that DX is old hat, the hype has died down but if you look games are finally start to use the new DX better (i.e. it's quite a bit better then the last DX).
4. Several years later cards not supporting that DX have gone and it becomes fully used.

 

[Pottuvoi]

Obviously it will eventually but this is the same with all DX releases...
1. There is much hype in the forums.
2. A load of games say they use the new DX, there is much fanboy excitement, but it hardly makes any difference over the old version.
3. Several years later that DX is old hat, the hype has died down but if you look games are finally start to use the new DX better (i.e. it's quite a bit better then the last DX).
4. Several years later cards not supporting that DX have gone and it becomes fully used.

DX12 wasn't meant to replace DX11, but offer a choice for highly skilled engine developers who want to control everything happening in graphics.
I'm quite sure that high level APIs are still needed in future, what is the name they will be called is not known.

Agreed that results from Vulkan and DX12 will get a lot better as they mature and developers can write software properly for them.

 

[Headfoot]

Im fully agree that we needed an improvement over DX11, i just dont agree that going to a lower level api is the way, at least no on PC where hardware is not static.

Except we really don't. Have you ever programmed? If so then you'd realize what a sad state of affairs it is that one of the very few consumer applications that requires maximum performance (Gaming) was crippled by a high level API. High level APIs sacrifice performance, in every case. It's like saying you need to host Google's sub-1-second searches but you can only use Ruby to do it.

 

[crisium]

Myself, I am hoping that Vulkan takes over. However I am a Linux Gamer.

As a Windows gamer I agree, so long as Doom is the standard. That game has the best performance-per-visuals (a subjective rating) I have ever experienced. If all Vulkan games can properly take advantage of Compute like Doom then I want this to be the standard. But I'm sure it won't, sadly.

 

[Dribble]

Except we really don't. Have you ever programmed? If so then you'd realize what a sad state of affairs it is that one of the very few consumer applications that requires maximum performance (Gaming) was crippled by a high level API. High level APIs sacrifice performance, in every case. It's like saying you need to host Google's sub-1-second searches but you can only use Ruby to do it.

Except that what actually happens is Nvidia/AMD would just replace sections of the games code that called the high level api with their own low level optimisations. With DX12 they are now meant to leave more of it to the game dev who may or may not do a better job. Imo problem here is game dev will:
a) probably do a good job for some cards (depending on who sponsored the game) but not bother with others.
b) are unlikely to update the game for new cards (which need different low level optimisations due to slightly different architecture) which Nvidia/AMD would do.

So DX12 isn't exactly the wonderful answer to all our problems.