Vulkan API might be better than DX12, says Magicka 2 developer

[Bouzoo]

Quote:"During an interview with WorldsFactory, Pisces Interactive, the developer behind Magicka 2 talked about DirectX 12. The studio said that DX12 will have large improvements over DX11, but teased that the Vulkan API could be even better.

The studio said: "DX12 is a big improvement over DX11, it enables games to move to the next generation of graphics. But we're also eagerly awaiting the official release of OpenGL Vulkan, which might even be better than DX12. Currently we are only experimenting with the API's but we will of course support them eventually"."

We suspected this much, so nothing new on that point, but now we got "pros" teasing it. More competition for DX12 is good. What do you think?

Source:http://www.tweaktown.com/news/45258/magicka-2-developer-vulkan-api-even-better-dx12/index.html

[Trik'Stari]

When will we actually see any of this in the wild? Tired of hearing about it but nothing happening.

[AxelRantila]

I'll wait and see before I decide on whenever DX12 or Vulkan is better.

 

Slight off-topic, but AFAIK Microsoft is rather strict about GPUs needing to support certain features in order to say that they support DX, but OpenGL was not as strict. Does anyone know if this will change with Vulkan.

[werto165]

Haven't really seen much of vulkan so far. I'm hoping it's good because I would like options when it comes to gaming, DX-12 you're just limited to windows. But I really don't think it's going to be the magic fairy dust that everyone is hoping for, I mean mantle really had little affect on performance and that API was designed specifcally for GCN.

[Hans Christian | Teri]

That's like saying Schrödinger's cat might still be alive. The information is completely useless until you've actually opened the box.

[Samfisher]

Haven't really seen much of vulkan so far. I'm hoping it's good because I would like options when it comes to gaming, DX-12 you're just limited to windows. But I really don't think it's going to be the magic fairy dust that everyone is hoping for, I mean mantle really had little affect on performance and that API was designed specifcally for GCN.

 

? Mantle had a huge performance upgrade in draw-call limited scenarios.

[MrDynamicMan]

But I can slfire them for 2x performance!

But Vulcan is the next mantle, great tech, but failed because of adoption.

[werto165]

? Mantle had a huge performance upgrade in draw-call limited scenarios.

Yeah I guess, perhaps me just thinking of battlefield 4, which didn't really change performance a whole lot on higher end systems. I guess it just gives devs more flexibility with what they can do. Personally I'm more excited for DX-12 because all I really want is forza on PC.

[patrickjp93]

? Mantle had a huge performance upgrade in draw-call limited scenarios.

Which Nvidia proved never existed. When they improved their DX 11 drivers, Nvidia beat out AMD's Mantle performance, with less bandwidth on the same CPUs. I'm sorry but AMD's Mantle performance is only better against their own DX11 performance.

[RuecanOnRails]

Vulkan will not rely on, or be restricted to systems running windows 10+. That alone makes it better and far more appealing than DX 12.

[Dzzope]

Take what game devs say with a grain of salt..

 

End of day it doesn't matter, whichever is used your machine will run.. as long as they pick the best tool for the job and then use it right.. Which lets be honest, most games aren't great at.

[Humbug]

I expect them to perform the same, don't expect meaningful performance differences between the two. It would probably be splitting hairs. Vulkan will obviously be widely used on mobile platforms and on Linux, whether it becomes used on Windows may be down to how much developers like it.

 

When will we actually see any of this in the wild? Tired of hearing about it but nothing happening.

Nothing happening?

It was only announced in August last year. And they have got the job done so fast that it's going to be done this year only a few months later than directX 12. Has a graphics API ever been created that fast?

 

http://www.phoronix.com/scan.php?page=article&item=valve-lunarg-vulkan&num=2

 

http://www.gdcvault.com/play/1022018/

 

http://youtu.be/miZmas6sGqM

 

 

[Trik'Stari]

I expect them to perform the same, don't expect meaningful performance differences between the two. It would probably be splitting hairs. Vulkan will obviously be widely used on mobile platforms and on Linux, whether it becomes used on Windows may be down to how much developers like it.

 

Nothing happening?

It was only announced in August last year. And they have got the job done so fast that it's going to be done this year only a few months later than directX 12. Has a graphics API ever been created that fast?

 

http://www.phoronix.com/scan.php?page=article&item=valve-lunarg-vulkan&num=2

 

http://www.gdcvault.com/play/1022018/

 

-snip

I was mainly talking about DX12. I'd like to see some titles coming out with it.

 

I guess you could say I'm just bored with the current crop of games, and want to see some awesomesauce games come out.

[deviant88]

Vulkan will be supported by many used by few, and even fewer released games.

OpenGL will never succeed because developers dont want to use it.

[sof006]

'Cept no one will use Vulkan... ever.

[Humbug]

Which Nvidia proved never existed. When they improved their DX 11 drivers, Nvidia beat out AMD's Mantle performance, with less bandwidth on the same CPUs. I'm sorry but AMD's Mantle performance is only better against their own DX11 performance.

I would not have expected such a post from you.

 

SamFisher is talking specifically about draw call performance. There is no way that Nvidia's DX11 draw call performance (however optimized) can match Mantle or DirectX 12.

It may be more optimized than AMD's DirectX 11 but that's it...

 

http://www.anandtech.com/show/9112/exploring-dx12-3dmark-api-overhead-feature-test/3

 

Now there are practical gaming scenarios where an Nvidia card running DX11 can beat an AMD card running Mantle. But that's not a proper scientific assessment of the API because the hardware is different and there may be other more significant bottlenecks apart from CPU limited draw calls.

[Humbug]

Vulkan will be supported by many used by few, and even fewer released games.

OpenGL will never succeed because developers dont want to use it.

Vulkan is not openGL. Ground up redesign.

It's closer to Mantle and DX12, nothing like openGL. Expect for the fact that it's made by Kronos.

[FaustoGomez]

I hope Vulkan turns out to be something widely used. With it, devs will have easier times porting their games to Linux which probably wil save you a lot of money, say, when you're building on a tight budget. Or a cheaper Windows version would do if Microsoft does something like "barebones" Windows for $30

Either way, I would love to see Vulkan succeed. It'd allow far easier porting of games compared to DX12.

[Humbug]

I hope Vulkan turns out to be something widely used. With it, devs will have easier times porting their games to Linux which probably wil save you a lot of money, say, when you're building on a tight budget. Or a cheaper Windows version would do if Microsoft does something like "barebones" Windows for $30

Either way, I would love to see Vulkan succeed. It'd allow far easier porting of games compared to DX12.

Well Vulkan is going to be popular on certain platforms like mobile and Linux-desktop.

 

The big question is will it be popular on windows?

We know that all the big name engines are going to support it but will devs actually use it to release windows games is an unknown, they may just default to DX12.

To achieve that utopia where everybody uses Vulkan across OSX, Windows and Linux there needs to be good developer support, documentation, drivers, conformance and good toolsets. There are very good signs this time that Kronos is not making the mistakes they made in the past with openGL. But only time will tell for sure...

 

I don't know whether the porting process from DX12 to Vulkan will be made easier by their similar design philosophies? If it is then it's a boon for LInux gamers.

[Thony]

When will we actually see any of this in the wild? Tired of hearing about it but nothing happening.

Im with u bro ! Sick of the talk, bring on the action ffs !

[Trik'Stari]

Im with u bro ! Sick of the talk, bring on the action ffs !

Damn straight.

 

Lets hope for a GIANT sandbox sequel to Metro Last Light.

[Mister Snow]

Which Nvidia proved never existed. When they improved their DX 11 drivers, Nvidia beat out AMD's Mantle performance, with less bandwidth on the same CPUs. I'm sorry but AMD's Mantle performance is only better against their own DX11 performance.

 

Sooo, this is not true or relevant?

 

[patrickjp93]

I would not have expected such a post from you.

SamFisher is talking specifically about draw call performance. There is no way that Nvidia's DX11 draw call performance (however optimized) can match Mantle or DirectX 12.

It may be more optimized than AMD's DirectX 11 but that's it.

-snip-

http://www.anandtech.com/show/9112/exploring-dx12-3dmark-api-overhead-feature-test/3

Now there are practical gaming scenarios where an Nvidia card running DX11 can beat an AMD card running Mantle. But that's not a proper scientific assessment of the API because the hardware is different and there may be other more significant bottlenecks apart from CPU limited draw calls.

Neither is this benchmark scientific either. You'd have to be one crappy programmer to use as many draw calls as this benchmark suggests as maximums, partly because you should be including many objects in your calls, and two, this benchmark leaves no room for AI or anything else. Also, yes Nvidia's upgraded DX 11 drivers provide for as many or more draw calls than Mantle for a single core. DX 11's flaw is putting all draw calls to a single core, but that's a separate issue.

Would it kill people to look at all dimensions of a problem before trying to speak definitively about it?

[patrickjp93]

Sooo, this is not true or relevant?

-snips-

True? Yes. Relevant? No, because you're misinterpreting the results. Mantle and DX 12 are both fully multithreaded and can even provide for Hyperthreading. You're comparing 5 cores to 1 if you run those on a 4790K for example. If you limit yourself to a single core, Nvidia proved its DX 11 driver provided for as many or more draw calls with respect to Mantle. Now, that is not to discount the gains of multiple cores, but no one uses that many draw calls anyway, and if we already struggle with CPU PhysX and AI, the new APIs will do nothing to change that.

[Humbug]

Neither is this benchmark scientific either. You'd have to be one crappy programmer to use as many draw calls as this benchmark suggests as maximums, partly because you should be including many objects in your calls, and two, this benchmark leaves no room for AI or anything else.

That's besides the point. It's a synthetic benchmark designed to isolate one aspect of performance. Nobody is claiming that real applications will see massive gains like this.

It just shows that draw calls are much cheaper in DX12/Mantle than in any DX11 implementation.

That doesn't mean that real games have to use 20 million draw calls per second just because they can. It means that that they can go higher before hitting a ceiling. And it means that once they use up whatever amount of draw calls they need to then there will be more CPU time left over to be put to other good uses such as AI, physics etc...

 

Also, yes Nvidia's upgraded DX 11 drivers provide for as many or more draw calls than Mantle for a single core

Please share the source of these results

And even if it's true; nobody uses single core CPUs.

 

here are the results for two cores.

it's hard to imagine DX11 catching upto DX12/Mantle if we go down to one core. But please share your results...

 

DX 11's flaw is putting all draw calls to a single core, but that's a separate issue.

It's not a separate issue, we are discussing the limitations of DX11 draw call performance.