-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Is DirectX 12 Worth the Trouble?
- Thread starter Kaapstad
- Start date
I think the most important thing about DX12 is the ability to move the workload much better accross the cores/threads. Some games you tend to see CPU bottlenecks at 1080p which is what DX12 was meant to be good at overcoming. But the thing is how many engines or games have been built from the ground up specifically for DX12? Most are made for DX11 with DX12 just patched in later for the sakes of compatibility seen as majority of gamers are stick rocking older gpus. It's like MGPU. Devs CBA with it because of the time it takes for a small majority of gamers. I think to see some actual gains from DX12 things need to move completely over.
Many of the big studios these days are simply using off the shelf products like unreal engine and doing very little engine coding or even necessarily working that close to the nuts and bolts of the programming. Small studios are very mixed with some using off the shelf platforms to design and publish their games and others pioneering their own engines.
Its a common fallacy that developers have to program for MGPU in the way many think - its not like an API where there are functions you specifically target - for the most part it is just building your game, seeing if it breaks driver level multi GPU, then going back in and changing the way you do things to try and not break compatibility or going to the GPU vendor and seeing if they can produce driver workarounds for your application.
Not the same but there was some vague similarities with DX10 in that:
1) It required a new version of Windows (Vista)
2) It was supposed to be light on new graphical features (unlike DX8, 9 etc) but instead just the opportunity for better performance. The idea was you'd chose DX10 over DX9 because it was faster, but in practice that usually wasn't the case, at least initially.
Nail on the head right there... most of the games are merley DX11 back ends patched to DX12 so performance gains are not going to be the numbers we expect.
This comes simply down to time and money... if devs built a game using DX12 from the ground up it would take time but we would get the performance gains expected.. instead its about money and they just dont have or want to spend the resources to do this.
Doom and the ID engine (think thats what it is) is an example of how to do it right.
Oh look another DX 12 is doomed thread... Is this by any chance because of nvidia's supposed dx 12 driver falling flat on its face? Is Chips in damage control mode? 
We already have a thread for all the dx 12 naysayers here:
https://forums.overclockers.co.uk/threads/is-directx-12-worth-the-trouble.18770873/
So to avoid a repeat of the same arguments, here is how that that thread can be summed up.
The only people who don't care for dx 12 are:
- people who stupidly buy 4 way setups and blame dx 12 for nvidia dropping support etc.
- people who have nvidia GPUs
- people who have beastly setups/cpus already
Is dx 12 perfect? No.
Is it worth the time/effort in the end? Depends on who you ask.
Will dx 12 die? No, look back at all the previous dx versions when first launched, exact same thing happened yet look at all the games that came out using said dx version in the end and Microsoft are pushing dx 12 more than any previous dx version. No doubt we will see dx 12.1 or something in the next year or 2 but it is only going to be improving upon the core of dx 12.
The fact that some developers are going back to add dx 12 to their games says that they do value what DX 12 is capable of and that it is worth the effort. If they didn't care then why bother doing this? Seems odd to waste time and resources when they could either be fixing the game or working on DLC/the next game...
I really can't wait till the day when nvidia have proper software/hardware support for dx 12 as it will be amusing to see how some people change their tunes on dx 12, I've already got posts bookmarked for that day!
Meanwhile I will go and play hitman on my 290 in dx 12 mode, which matches a 980ti In fact, I'm probably getting better FPS than what I would be getting with a 980ti in my system since my CPU is weak/older...
We already have a thread for all the dx 12 naysayers here:
https://forums.overclockers.co.uk/threads/is-directx-12-worth-the-trouble.18770873/
So to avoid a repeat of the same arguments, here is how that that thread can be summed up.
The only people who don't care for dx 12 are:
- people who stupidly buy 4 way setups and blame dx 12 for nvidia dropping support etc.
- people who have nvidia GPUs
- people who have beastly setups/cpus already
Is dx 12 perfect? No.
Is it worth the time/effort in the end? Depends on who you ask.
Will dx 12 die? No, look back at all the previous dx versions when first launched, exact same thing happened yet look at all the games that came out using said dx version in the end and Microsoft are pushing dx 12 more than any previous dx version. No doubt we will see dx 12.1 or something in the next year or 2 but it is only going to be improving upon the core of dx 12.
The fact that some developers are going back to add dx 12 to their games says that they do value what DX 12 is capable of and that it is worth the effort. If they didn't care then why bother doing this? Seems odd to waste time and resources when they could either be fixing the game or working on DLC/the next game...
I really can't wait till the day when nvidia have proper software/hardware support for dx 12 as it will be amusing to see how some people change their tunes on dx 12, I've already got posts bookmarked for that day!
Meanwhile I will go and play hitman on my 290 in dx 12 mode, which matches a 980ti
In fact, I'm probably getting better FPS than what I would be getting with a 980ti in my system since my CPU is weak/older...AMD more or less sorted their DX 11 performance out over a year ago. I'm pretty sure that there was a thread about it on here comparing old and new drivers and showing the 2xx matching/overtaking the 7xx series cards.
Last edited:
Permabanned
- Joined
- 12 Sep 2013
- Posts
- 9,221
- Location
- Knowhere
DE:MD is a good example of this. It shares the load excellently across the cores, to the extent that I had 100% cpu utilisation on all four cores in places (extremely rare in most games because there is always a bottleneck thread somewhere) and in general pretty high and well distributed throughout. However by definition that means that it hasn't removed the CPU bottleneck, it has just been alleviated somewhat. To be honest, I am not sure how efficient it is, i.e. it maxes out my CPU but doesn't perform noticeably better than DX11, for all I know it could just be doing 1+1-1+1-1+1.... etc etc in the background on a couple of threads for no benefit, increasing cpu utilisation but not actually increasing framerate.
If you are going to get familiarity with a new API, especially one that has some significant new concepts to get your head around, one way of doing that is to update an existing engine to support it. This doesn't mean they care - just that they are looking at the potential.
I don't think most people understand the mindset of the average game developer - there is a reason (aside from some dubious moves by MS) why Direct3D became more and more popular over Open GL with each iteration - Open GL has always been a little more leaning towards working under the hood while D3D has been a more unified approach at an almost purely abstracted level. Largely this is exactly what developers want not the romanticised vision some like AMD have been purveying of working closer to the metal, however developers do want to be able to selectively deep dive where they feel there is a benefit while traditionally high level APIs have largely locked ability to do that out. DX12 forces them to get down and dirty under the hood for everything whether it provides a benefit for what they are doing or not.
For instance some functions in Direct3D will say spawn a copy of a texture, even when the developer knows that with what they are doing they specifically won't be using functionality that copy is there for and doesn't present the developer with any handle to that data so they can manage it themselves i.e. free it up to save memory (there are some hacky ways of doing this via for instance the RTL API but that is another story) that data will stay resident in memory until the API itself decides it no longer needs it - which often is when the API is unloaded when the program quits. What developers want is to be able to go in and manage that memory on a case by case basis but what DX12 makes them do is essentially reinvent that whole function for themselves including all the memory management, from the ground up, every time, which 9 times out of 10 has NO BENEFITS and it just isn't worth it for that 1 time in 10 where it is a benefit.
DX12 does have more advanced support for stuff like taking advantage of increased parallelisation, but largely that could be exposed via something more akin to DX11 without requiring the full on implementation of DX12.
Unfortunately I can't even come close to explaining concisely why DX12 is a misstep, and a big one, along the way and this is nothing to do with the state of nVidia's support for DX12.
EDIT: DX12 done right can massively improve on the results either in visual richness or performance or a mix of both over older APIs when done right - that isn't something I'm disputing - the problem is that what is required to get there is largely a massive roadblock and isn't going to become any less of one with time. Likely people will start wrapping it, losing some of the benefits along the way, which might eventually see more extended uptake but by that time I suspect another API will likely supersede it.
Last edited:
^^
Yup very true, they will need to get to know the API before starting a completely new project using just dx 12 but still, it seems odd for them to waste time and resources doing this if they are happy just to stick with dx 11 and not move forward at all, especially on the PC as we are already an after thought and often have to shout from the roof tops just to get developers to fix their games.
And like I said, dx 12 isn't perfect and no doubt we will see numerous "revisions" of it.
Yup very true, they will need to get to know the API before starting a completely new project using just dx 12 but still, it seems odd for them to waste time and resources doing this if they are happy just to stick with dx 11 and not move forward at all, especially on the PC as we are already an after thought and often have to shout from the roof tops just to get developers to fix their games.
And like I said, dx 12 isn't perfect and no doubt we will see numerous "revisions" of it.
Are you serious? Are you just going to totally ignore the fact that AMD cards get good performance gains out of low level APIs such as DX12?
It's not about running DX12 faster which actually is retarded as DX12 is just a API and it's actually used to render a game.
It's about what benefits DX12 brings over DX11 and there are multiple benefits over DX11 but you want to ignore that just because nvidia have faster cards in games? Not DX11 or DX12 just games. If nvidia got a similar boost to AMD cards would you still hold the same opinion? when your "faster cards" get similar performance gains? Because to me that seems really biased.
If dx12 is going nowhere, why did Nvidia and Microsoft waste four years developing it? At least that was their claim at that presentation a couple of years ago. You would imagine that two such powerful companies, who put so much time into Dx12, would be pushing it to the max to be successful. Just seems strange that they would go to so much effort for that Presentation then do nothing more.
I think they were just bored and did it for the S**** and giggles. lol
No but seriously, good point. MS and nVidia as well as all the other countless people helping must have seen a benefit to it.
I'm not sure that developers are that happy to stick with DX11 - some might be that just use off the shelf products to build their games but engine developers, etc. will be finding it a limitation.
Problem with DX12 is you have to build something complex, then focus on the specifics while traditionally with older APIs you'd build something relatively simple as the core of your game and then add complexity - very few people like to work with complexity off the bat and I think this is where these next gen APIs are falling down.
I think part of the issue there is MS have long since stopped really listening to their users - a lot of the time it seems they only hear what they want to hear. Certain parties seem to have run with this idea that game developers want to work closer to the metal and gone the whole hog with it despite that not being what developers are actually asking for if you really listen.
Has it actually been confirmed that nvidia were working with Microsoft on dx 12? I remember seeing lots of people saying that they were even working with Microsoft long before AMD... Seems odd that considering the state of dx 12 on all of their gpus...
Because dx 12 is what forced nvidia to drop 3/4 way SLI support 
If/when DX 12 enables great 3/4 crossfire/sli support, watch what happens
He is just butt hurt as he can no longer use his 4x GPU setups to top the benchmark leaderboards on here now
Because dx 12 is what forced nvidia to drop 3/4 way SLI support If/when DX 12 enables great 3/4 crossfire/sli support, watch what happens
What about vulkan? Would it be any easier than dx 12? If so, then why aren't we seeing more up take of it?
Has it actually been confirmed that nvidia were working with Microsoft on dx 12? I remember seeing lots of people saying that they were even working with Microsoft long before AMD... Seems odd that considering the state of dx 12 on all of their gpus...
He is just butt hurt as he can no longer use his 4x GPU setups to top the benchmark leaderboards on here now
If/when DX 12 enables great 3/4 crossfire/sli support, watch what happens
What about vulkan? Would it be any easier than dx 12? If so, then why aren't we seeing more up take of it?
Yea sometimes i see bias in some people's opinions.
But yea i thought Vulkan was supposed to be easy for Devs to use? That was the whole point wasn't it? With great scale-ability as in can be used in Linux, windows. Even android etc.
Vulkan largely takes a very similar route to DX12 - though it seems to have been a bit better thought out - uptake is mostly due to it being a bit different (new API, etc.) so developers aren't so familiar with the overall package as DX.
Not really sure what happened with Mantle/Vulkan as the roots look like it started in the right direction then took a 90 degree turn somewhere.
nVidia seem to be trying to shape Vulkan to their own advantage somewhat by getting to hearts and minds first https://developer.nvidia.com/Vulkan it wouldn't surprise me to see them pushing it a lot more in the near future.
He is just butt hurt as he can no longer use his 4x GPU setups to top the benchmark leaderboards on here now
If/when DX 12 enables great 3/4 crossfire/sli support, watch what happens
Are you serious?
I have got 4 of AMDs fastest card which do support quadfire and now run pretty good now they have finally sorted out the drivers.
The funny part about it is that NVidia only support 2 way SLI yet 4 way SLI does run very well - AMD support 4 way Quadfire and up until recently it has been awful since the introduction of the Fiji cards, fortunately they have finally got it running properly.
My problem with DX12 is it is a new API that should give better results with hardware going forward, something that is not happening at the moment as there is too much extra work for the game devs. A new API should not be judged by what it can with old hardware as this is a journey to nowhere.
Yup but hey that's what you can expect in this sub forum
Except once again, you are ignoring that it does give better performance on cards that HAVE the support for it (even with newer/better CPUs partnered with AMD cards).... Don't blame the API just because nvidia can't make proper use of it.
And also, I'm pretty sure the main point of dx 12 (at least for performance wise) has always been to help the older/weaker CPUs partnered with newer/higher end GPUs.
Regarding dx 12, multi-gpu etc. you more or less said it yourself here:
Well I don't care what API games use, as long as I see the same benefits that I do with dx 12 on my setup then I'll be happy!
EDIT:
Also 4 way SLI/crossfire is working great for you? Well that is surprising considering everyone who only has 2 GPUs (which "should" be far better supported than the incredibly niche 4 way...) always complain about flickering, stuttering or/and poor scaling or just down right doesn't work for months, if at all...
Last edited:
I really can't wait till the day when nvidia have proper software/hardware support for dx 12 as it will be amusing to see how some people change their tunes on dx 12, I've already got posts bookmarked for that day!
Meanwhile I will go and play hitman on my 290 in dx 12 mode, which matches a 980ti
Nv are getting the ball rolling, wasn't expecting Hitman result as Rottr/DX-MK DX12 only got to parity with DX11, think Nv is going to nail DX12 software/hardware support with Volta.
You might have to wait a while, but despite the negativity, when Volta arrives, get your bookmarks ready when DX12 really arrives.
@Kaapstad although you don't get the gains on your halo system, doesn't equate to it's crap in general, as way, way more people are getting DX12 gains on lesser hardware and that's one of the main reasons for DX12.
