At this year’s GDC, Jurjen Katsman, CEO of NIXXES Software, talked about DirectX 12 and expressed his opinions about it. As we’ve already said, DX12 is not exactly what it was promised and it appears that Katsman agrees to this. And while he admitted that DX12 can be worth the extra effort, he also claimed that it may not be (at least for some developers).
Katsman claimed that the CPU performance gains that come with DX12 are negligible for enthusiast gamers, for those owning a high-end CPU, for those that game at 1440p or 4K resolutions, and for those using High Quality settings.
On the other hand, Katsman stated that the GPU performance gains from Async Compute are currently inconsistent. According to Katsman, there are currently some API restrictions that actually cause problems. Moreover, hardware support is limited for Async Compute and it requires tuning to very specific hardware.
What’s also interesting here is that in DX12 the GPU driver is still very relevant. DX12 is a low-level API and a lot of people believe that this will make things easier for developers. However, that’s not the case. On some drivers the complexities have gone up in the areas of memory management, manual multi-GPU, support for Async Compute, and the various performance-enhancement tricks the various vendors have implemented in DX11.
Katsman admitted that Async Compute is the way to go, however the DX12 API may need some improvements. Moreover, Katsman claimed that DX12 can help CPU-heavy scenes that are bottlenecked, and that an overall 10% GPU performance increase can actually happen.
Katsman concluded that DX12 is hard, it can be worth the extra effort, but it may also not be. That’s up to the developers, and PC gamers should lower their expectations about this API.
Of course our regular readers are well aware of these things. Back in December, we talked about DX12 and whether it was worth the hype. And while some criticized us for that particular article, it’s pretty good witnessing developers sharing the same concerns and beliefs with us.
Our take on this? Katsman is right, though he forgot mentioning that DX12 will benefit scenes that are limited by drawcalls, and games that rely heavily on a single CPU core/thread (theoretically, DX12 should improve CPU performance thanks to its better CPU multi-thread capabilities, provided the game does not remain single-threaded).
Thanks TECHPOWERUP
John is the founder and Editor in Chief at DSOGaming. He is a PC gaming fan and highly supports the modding and indie communities. Before creating DSOGaming, John worked on numerous gaming websites. While he is a die-hard PC gamer, his gaming roots can be found on consoles. John loved – and still does – the 16-bit consoles, and considers SNES to be one of the best consoles. Still, the PC platform won him over consoles. That was mainly due to 3DFX and its iconic dedicated 3D accelerator graphics card, Voodoo 2. John has also written a higher degree thesis on the “The Evolution of PC graphics cards.”
Contact: Email
or u just suck at it, check doom what did they did with vulkan, maybe u should consider vulkan over dx12
just because you see gain with vulkan meaning vulkan is easy to use compared to DX12. just as some people point out we see nice gain with doom but it also coming from developer that has been targeting OpenGL for decades. doom developer already explained that the gain they have with vulkan is coming from use of extensions which the API inherit from OpenGL.
now funny thing is many people said that Vulkan should ditch extension like feature from the API because it is what end up making OpenGL less successful VS DirectX in the past due to fragmentation in vendor implementation of their OpenGL drivers. but without extensions we most likely going to see similar problem with vulkan that DX12 are having right now.
AMD’s implementation of OpenGL is bad, that’s why you see those gains with Vulkan in Doom. That’s also why you don’t see much improvement on most nVidia cards. Vulkan isn’t magic, neither is DX12.
who said anything about amd or nvidia im commenting on nixxes’s
dx12 use cpu more efficiently and allows multi-gpu way in better way than SLI & Crossfire, of course it’s worth the effort
Nope, you were talking about performance improvements with Vulkan
Would rather have Vulkan.
Any day, any year!
For AMD users there’s some games which benefit from DX12 very, very well then there’s some which don’t and Nixxes ported games always seem to fall in the latter category. (For Nvidia DX12 and Vulkan just aren’t any use)
To those who just keep saying use Vulkan: You can get Vulkan like performance gains with DX12, We’ve seen it before. Nixxes is clearly doing something wrong.
inb4 you f*cking shill
Not going to call you a shill by any means, but saying ” You can get Vulkan like performance gains with DX12″ just equates to “why bother with Vulkan?”, why don’t we all move to a single OS, single gover,ent, single everything?.
We don’t because we’re all different and we like having differences. Yes we can get some Vulkan like gains with DX 12, but we haven’t seen an overwhelming stack of games that show that a lot of the time. Also Vulkan supports all of us, DX 12 only supports Windows 10, so by that logic it’s apparently better to support the one company than everyone else.
I’d rather everyone get the same level of support than being shoved into a box and told what to like.
I agree. I never once meant “why bother with vulkan”.
Nixxes is not doing optimization very well, but please lets ditch ms already.
“You can get Vulkan like performance gains with DX12”
why you’re so sure?
we saw the gain from OpenGL to Vulkan not from DX11 to Vulkan
the problem with OpenGL is it’s not fully standardize
each company has their implementation of the same technology and that make it hard for the developers
the performance gain on AMD cards just show how bad AMD drivers deal with the scattered code of OpenGL
Not a new theory by any means, but I think they wanted to push DX12 to the widest demographic of computer users, just to get us onto Windows 10. And of course, Valve releasing Steam OS, which I haven’t heard anything about from them in a little while now.
Yeah, DX12 was way better at converting people to Windows 10 than at giving performance gains in games.
MS was alwayw going to end support in 2020 though. When I look at win10 I see desperation for a new business model. Ain’t nobody paying for an OS anymore.
Who gives a f***. DirectX 11 works just fine.
Yes, it might be fine for you and your console. But on the PC platform, we like to see a bit more innovation.
Inovation are better graphics
And? You can’t make sh*t look prettier forever.
Besides, “innovation” doesn’t just mean “add more pixels to it!” It also means “figure out more efficient ways to run the existing amount of pixels, so that you can add even more pixels!”
Assclown.
psst cpy cracked saints row 3 and the last sherlock holmes game. So there goes your argument, if cpy bothers he can do it, there is nothing uncrackable.
Best part of DX11? With every second we spend on this decade-old API that was outdated almost from the moment it first launched, 90% of everyone’s Multi-Core CPU’s are going to waste because of Microsoft’s sh*tty multi-core support, so, yeah, DX11 ftw, brah.
Not.
This is an odd article considering how much Rise of the Tomb Raider was optimized for Nvidia despite them not supporting Async Compute.
async compute is not the only way of increasing performance with DX12. in TR case they even specifically showcase CPU limited situation and how DX12 improves that. AFAIK they never really talk about async with TR.
Oh I know its not, but it’s one of the most effective. The spent a considerable amount of time optimizing for Nvidia.
only most effective on AMD hardware. but even that because of their utilization issue they have with their architecture. nvidia also have utilization issues with their past architecture but they able to increase their utilization by making direct changes to their base architecture. and because of that they don’t need “new features” being added to the API first to properly benefit from them.
Plz moar Vulkan. But we all know it won’t happen.
One can dream.
meanwhile, in the kingdom of Vulkan…
Do only reason why people jerk on it it’s because of Doom and that’s pathetic.
Well its not like idsoftware did well with rage.
Or because performance enhancements/benefits are huge and real unlike DX12… and it’s platform agnostic… unlike DX12?
there are reasons why vulkan did better in this low level business better than DX12. but that reason that makes vulkan better in low level API probably also the reason why many game developer avoiding OpenGL in the first place.
So are you, peasant 🙁
Have a game with more than a few characters on screen, with more than just a few clothes, with more than just a few repeating objects and then not embrace a low level API.
This is nothing more than remaining close to the same old way of making games, never really adding anything new.
Nixxes some time can optimize and make it worth the extra effort, but they may not bother with optimizing either.
See what i did there?
Its funny because its true and with dx12 it seems to be the same case, they wont bother because Nixxes doesnt want to bother.
SOOO can we ditch dx already?
lol Nixxes.
Presentation: “DX12 is really hard”
No kidding? Nixxes may be at the top of game devs, but they had issues with a mid-high level API like DX11 even with relatively generous deadlines, and then they and AMD flog MS over and over again to make a low level API in DX. So MS does with DX12, which has on a common setup of i7-3770/GTX 970 about 700% of the pure rendering speed of DX11 – if properly optimized and if GPU rendering is the bottleneck.
Their response? It’s “really hard” and “may not be” worth it. *face palm*
Yes, & no. To be fair, Microsoft didn’t have any intentions of making DX12 before Mantle came out. If it wasn’t for AMD, Microsoft’s Windows 10 Direct3D API update would have been the DX11.3 API that also exists on Windows 10, alongside DX12, just without the low-level coding features.
Once AMD did Mantle, Microsoft realised they were falling behind, so they got off their lazy asscracks & made DirectX 12. Other than that, agreed.
The reason why you see barely any gains with good CPUs is simple, these games are all made for consoles. What’s going to stress these CPU beasts if all they have to worry is running dumb down AI, laughable physics and lighting that’s still at the level of the last generation.
Make a game exclusively for PC and push it to the limits, make your engine scale from 4c/4t to 8c/16c. That would be the only scenario where using a low level API could prove worth the effort. As long as the potatoes keep holding the PC back, there’s not much to gain.
“hardware support is limited for Async Compute”, boy I wonder what’s the company responsible for this.
Consoles and NVIDIA have been holding this generation back. People need to stop buying hardware with terrible planned obsolescence.
I’m wondering what they think of Vulkan.
This is rather stupid article. There may be some gamers out there who simply fell in love with the promised gains of DX 12 and forgot to learn how they were supposed to come about, but for those of us with some technical knowledge we always knew this would be a down the line sort of improvement. I can’ remember a single API DirectX or OpenGL outside of Glide that produced near instantaneous improvements with pre-existing technology. And even that is not entirely comparable since the Glide API was a first party API provided by the GPU manufacturer compared to the DX 12 API which is a third party product supported by Nvidia and AMD. To be honest I see DX 12 and Vulkan yielding real benefits in 2-3 three years because it really is a big shift in the established schism both for hardware manufacturer and software developers. And equally important because as long as DX 11 still holds a large portion of the customer base you have to build and optimised around traditional synchronous API calls either way. Simply implementing this via DX 12 will give no real benefit so why bother doubling your workload.