DX12 step back
Images are sometimes not shown due to bandwidth/network limitations. Refreshing the page usually helps.
You are currently reading a thread in /g/ - Technology
You are currently reading a thread in /g/ - Technology
>new patch to render over dx12
>it is slower
To be fair, the developer said the advantage they were expecting was better multi-core CPU usage. But if it is just that, it is not really a good start compared to the alternative.
By the way, expecting GPU manufacturers to optimize drivers goes contrary to the purpose of a low level API, if anyone still has reason to truly rely on this then DX12 is as good as useless.
Source: http://www.hardocp.com/article/2016/03/28/rise_tomb_raider_dx11_vs_dx12_review
>>53735214
>By the way, expecting GPU manufacturers to optimize drivers goes contrary to the purpose of a low level API, if anyone still has reason to truly rely on this then DX12 is as good as useless.
Explain? If the low-level API is being called with the same high-overhead functions then of course it's going to be slow as shit.
Just implementing a low-level API doesn't require everybody else to use it, and it being low-level doesn't somehow magically optimize the code above it.
You're talking out of your ass, parallelization is a difficult problem to solve and it's going to take awhile before both game makers and driver programmers understand them. Any kind of major shift like Async compute is a big deal.
DX12 was just a meme news at 11.
With Vulkan directx is absolutely useless.
Microsoft better uses Vulkan on their next console or no one will bother with it.
>>53735320
never...No one steps foot with in the kingdom. NO ONE!
>>53735280
>before
- Developers call high-level functions
- They have no control over the implementation of such functions
- Driver manufacturers must optimize their drivers to provide good performance
>now
- Developers use low-level functions
- They craft their own high-level functions
- Driver manufacturers have no control over their implementation
- Developers must optimize their own functions to provide good performance.
>>53735214
>>53735320
>>53735429
basically devs will have to re-write or create new engines with Vulkan or DX12 for any benefit to exist
>>53735724
Ashes of Singularity/Unreal/CryEngine/DICE are available
I'm pretty sure all of them will be used in multi games. Some are license-able too.
>>53735214
I'm currently using an old CPU (Phenom II X4 965) with a GTX 780 and minimums are way better in DX12 mode. In the final area of the benchmark it drops down to ~4fps at its lowest point with DX11, whilst it stays over 30fps at all times with DX12. It's not going to be useful for anybody with a competent CPU (and indeed I got slightly lower performance when I had my 4790K), but it'll help people using shitty ones massively, just like Mantle did.
>>53735320
More than enough people purchase Microsoft consoles for them to hold the cards. Major publishers will continue to make games supporting DX if it's the only option. Plus, see below.
>>53735724
Most major middleware engines already make abstractions to support multiple graphics APIs anyway.
>>53735960
Nobody buys microsoft consoles, AND they have to sell them at a loss to begin with.
VULKAN
U
L
K
A
N
>>53736127
not the fucking point asshole. you know what the fuck he's saying stop being a kunt.
>>53736196
>he
Good joke samefag. How about you go be retarded somewhere else, like >>>/reddit/
>>53736127
I know you'd like to think so, and trust me, so would I. But they sell far, far more of them than you think. You wouldn't believe how many people own an Xbox One and don't even have a computer, often even an internet connection. And I mean just in the US. Then consider that they market and sell their consoles worldwide.
>>53736352
Found the paid shill.
>>53735960
The old abstractions may not be enough, the new APIs especially Vulkan are too different, to get maximum performance the engine should be rewritten - if not then at least the time consuming parts should be remade.
>>53736380
You shitting me? I'm just telling you the facts. I'm not happy that's how it is. The world isn't made up of people who post on /g/, it's made up of people who play Angry Birds and watch Netflix on their tablets and at most have a console so they can play Call of Duty and Destiny. Are you really that oblivious?
>>53736606
0.334 rupees have been deposited in good account sir Pajeet Rajesh. Keep please kindly do the needful thank you sir.
>>53736406
To a degree, yeah. With the right abstractions that can still be somewhat accounted for. It just means the abstractions for Vulkan and DX12 will be more involved (though those two could likely be built on the same set of lower abstractions since they're more similar than Vulkan and OpenGL, or OpenGL and DX11). Middleware engines already use custom allocators anyway, so it's really just command buffers. You're never gonna have ideal performance with an engine that supports a bunch of APIs anyway, though, because they're all different enough from each other that there's no way to totally avoid bloat, redundancy, wasted time, etc. Ideally, a new engine or a rewrite could really just support Vulkan and DX12, since that actually covers pretty much everything now, because of Vulkan's compatibility back to OpenGL 3.x. And Vulkan and DX12 are similar enough to avoid overabstraction.
>>53736732
I haven't said a single thing in support of Microsoft you dumbshit. You're trying way too hard.
>>53736815
Another major advantage of the new APIs is that there is no leeway for the driver to guess what the developer wants, the developer is the one who has to properly set things up. It will surely lead to less bugs, but not without rewrites, we have already seen how automatic porting tools work regarding DX->GL wrappers.
If anything, a DX-VK wrapper supposedly should be better than a DX-GL wrapper, but still far from ideal.
>>53736877
Vulkan and DX12 are far more similar than any version of GL and DX have been. You could make a pretty ideal abstraction provided you're only supporting those two APIs and you're only targeting what you'd expect from a game engine. Supporting more broad generic computation capabilities would make things messier, but a pretty standard game pipeline and renderer could potentially be pretty damn clean, especially compared to anything done before with older DX and GL versions.
>It will surely lead to less bugs
On the driver side, yes. On the developer side, hell no. But as long as people know what they're doing and use the support level and validation tools, it should be alright.
>>53736815
Is very good pajeet sir but no have more money in bank for needful of send! thank you sir
>>53735429
so if i get a game running like shit it's 100% the devs fault? also If a dev is using gameworks would it still yield poor results on a non nvidia card?(yes i know nothing about this stuff)
>>53737174
With dx12 or vulkan, yes.
yes because that's the whole point of gameworks. Gameworks over dx12/vulkan is "the high-level functions" the devs are supposed to make themselves, except in blackbox format.
>>53735214
>idk what I'm memeing anymore
>>53737034
While everything remains on the developers side they can do something about it themselves. Against driver bugs their hands are tied.
