in development Vulkan API Discussion

[beppe_goodoldrebel]

Base programming and testing should be finished next month. We'll then work on transferring our shaders to the new API. This is a very new technology for us, and it is quite possible that we will run into unforeseen issues during that phase. DCS World uses many complicated shaders that may not necessarily play nice with the Vulkan API.

 

Thanks!

 

How much do you guys think this API would benefit the FPS?

Or at least where are you aiming for?

[NineLine]

How much do you guys think this API would benefit the FPS?

Or at least where are you aiming for?

 

Probably not something we want to guess at right now, better to wait till its closer to release.

[wilbur81]

Base programming and testing should be finished next month. We'll then work on transferring our shaders to the new API.

 

Thanks!

 

Thank YOU, Matt...for the response... and all your hard work! :thumbup:

[Precog]

Base programming and testing should be finished next month. We'll then work on transferring our shaders to the new API. This is a very new technology for us, and it is quite possible that we will run into unforeseen issues during that phase. DCS World uses many complicated shaders that may not necessarily play nice with the Vulkan API.

 

Thanks!

Thanks for the update Matt. Looking forward to hearing more about Vulcan implementation. I hope it goes smooth for you guys.

[falcon_120]

How much do you guys think this API would benefit the FPS?

Or at least where are you aiming for?

 

Its not that simple, you can't say i will improve Fps a 10% or 13,5%. You will never get an answer to that because is impossible to say until is done and you benchmark on your system.

 

Lets wait patiently and support ED on their quest.

[Mr_sukebe]

I’d also expect system specs to influence the improvements possible.

For example, if Vulkan does allow offloading of video tasks from the GPU to unused CPU cores, then a Threaripper system with 16 cores will have far more free resources than say my wheezing i5 2500.

[sobek]

For example, if Vulkan does allow offloading of video tasks from the GPU to unused CPU cores, then a Threaripper system with 16 cores will have far more free resources than say my wheezing i5 2500.

 

 

Doing that would actually be a rather bad idea because of the comparatively low bandwidth of the connection between RAM and GPU. To render a frame you need to work on big amounts of data to generate even more data and ultimately that data needs to end up on the graphics card anyway. Offloading things to the CPU would most likely cost you huge amounts of performance while the GPU waits for stuff from RAM to be loaded. More cores is mostly beneficial if your CPU isn't able to fully saturate the GPU (and the application properly uses a modern graphics API) or if you're doing lots of stuff that is not well suited to be run on the GPU on the side.

Edited August 10, 2018 by sobek

[kazereal]

Base programming and testing should be finished next month. We'll then work on transferring our shaders to the new API. This is a very new technology for us, and it is quite possible that we will run into unforeseen issues during that phase. DCS World uses many complicated shaders that may not necessarily play nice with the Vulkan API.

 

Thanks!

 

Thanks for info! I have been anticipating for this to happen :)

[kazereal]

Doing that would actually be a rather bad idea because of the comparatively low bandwidth of the connection between RAM and GPU. To render a frame you need to work on big amounts of data to generate even more data and ultimately that data needs to end up on the graphics card anyway. Offloading things to the CPU would most likely cost you huge amounts of performance while the GPU waits for stuff from RAM to be loaded. More cores is mostly beneficial if your CPU isn't able to fully saturate the GPU (and the application properly uses a modern graphics API) or if you're doing lots of stuff that is not well suited to be run on the GPU on the side.

 

I agree with you: if CPU isn't doing things fast enough you essentially stall the GPU (make it wait for CPU to give more tasks).

 

PlayStation 4 uses different bus-architecture than PC called HSA-architecture: that essentially allows GPU and CPU share same data without copying between memories. If you had that on PC it could improve efficiency.

 

Often GPU is more efficient in highly parallel tasks and offloading things from CPU to GPU makes more sense. There are other tasks that don't scale well on GPU and are better to be done on CPU threads: network code is one example of such. So offloading things GPU and threading the rest of CPU would be most useful approach. And simulation code does have plenty of things that can benefit from parallelism if it is done so that it can benefit from parallelism: this means data structures need to work well with this design.

[etherbattx]

Vulkan API

 

good ideas and good analysis.

please send this to the developers.

[Worrazen]

Base programming and testing should be finished next month. We'll then work on transferring our shaders to the new API. This is a very new technology for us, and it is quite possible that we will run into unforeseen issues during that phase. DCS World uses many complicated shaders that may not necessarily play nice with the Vulkan API.

 

So they're not being written from scratch? Heh, don't get me wrong I'm aware of the headache that would do if it was required, just to clear out the specifics. Pretty much nobody does that yet, I've heard it's most difficult and the Vulkan API team thankfully delivered the conversion tools.

 

It's good for people to know that performance improvements would come gradually as time goes on, but the fact that this new API implementation is happening at all is more than great news.

 

 

For example, if Vulkan does allow offloading of video tasks from the GPU to unused CPU cores, then a Threaripper system with 16 cores will have far more free resources than say my wheezing i5 2500.

 

Oh that's most probably not how it would work ... no API or anyone is trying to do that at all. The point of Vulkan API is also to let the GPU work more without needing the CPU to babysit it, which is what the whole ERA of PC gaming of last 20 years was, horribly unoptimized!, which is another way of saying to lower the cost of draw calls on the CPU.

 

You don't need extra GPU work on the CPU, even if you wanted it wouldn't help much, rather use it for far more important things -> AI, more units on screen (more effects, particles, models, etc) more complicated physics and simulation!!!

Edited August 12, 2018 by Worrazen

[boedha68]

Base programming and testing should be finished next month. We'll then work on transferring our shaders to the new API. This is a very new technology for us, and it is quite possible that we will run into unforeseen issues during that phase. DCS World uses many complicated shaders that may not necessarily play nice with the Vulkan API.

 

Thanks!

 

Thanks for the update Matt. Great news. :D

[Vitormouraa]

I just had a thought, or question pop into my head.

When DCS 2.5 transitions to the Vulkan API, will DCS World be rebranded to DCS World 3.0?

 

Why would it?

 

From a marketing perspective, DCS has seen a few rebrands when it was bringing a new graphics engine, although Vulkan API is a major update, I don't see DCS receiving a new number because of that.

 

We recently had two major graphics engine update, new PBR and deferred shading rendering in 2017 and 2.5 this year. Along with 1.5 in 2015.

[Worrazen]

I just had a thought, or question pop into my head.

When DCS 2.5 transitions to the Vulkan API, will DCS World be rebranded to DCS World 3.0?

 

I hope not, I don't like the google-style fast paced versioning, with stupid mozilla following suit and then everyone else, any means to distance away from the mobile-web habits the better IMO, it's a gimmick, I didn't like the jump from 2.0 to 2.5 either because it felt like a void, something missing, or else they should explain better what were all those big changes that put 2.0 to 2.5 because as I remember there was only one big change of things unified and multiple maps, afaik.

 

2.6 would be okay with me. Then, various companies have public and internal versioning split, so it's even more of a mess. Just look at microsoft, there's the internal versioning, then builds, then public, then 3 different namings, it's ridicolous.

 

Windows NT 7.1.x.xxxx

Windows 7 Starter

Windows 7 Home Basic

Windows 7 Home Premium

Windows 7 Enterprise

Windows 7 Ultimate

Anniversary Update, Redstone, Build 1607, 10.0.14393.0

 

Attempting to make things easier by making public versioning is negated by the fact that internal names/numbers get pressed in the media/discussion so much, then when everyone's accustomed it gets another term, it doesn't help at all.

 

https://arstechnica.com/information-technology/2014/11/why-windows-10-isnt-version-6-any-more-and-why-it-will-probably-work/

 

Windows 10 Ultimate 7: Remastered 20th Anniversary Edition 128-bit Service Pack 5 Build 1530 R3 Version 100.007.1803.15434.0.10

Edited August 12, 2018 by Worrazen

[M1Combat]

So you're mad at ED because Google, Mozilla and Microsoft don't version their software the way you want?

 

 

FYI... there was a very significant graphics engine upgrade between 2.0 and 2.5.

 

 

 

 

"Windows 10 Ultimate 7: Remastered 20th Anniversary Edition 128-bit Service Pack 5 Build 1530 R3 Version 100.007.1803.15434.0.10"

 

 

I've never heard of that version...

 

 

I suspect this problem isn't quite as problematic as befits the amount of consternation it seems to have caused you...

[six.dof]

If i play, than 90% DCS, plan to buy a new PC, will be very interesting what change with Vulcan, better wait .....

[Bogey Jammer]

I really hope for ED to succeed with the new API.

[etherbattx]

are there sone benefits to Vulcan, but it’s highly unlikely to be THE silver bullet some of us expect it to be.

[falcon_120]

I see it more like the foundations for DCS in the years to come. It should not increase very much fps now, but it will allow to deploy complex functionality(DC, better AI, etc) later on without exponential hardware requirements, through better hw balancing and usage.

 

Enviado desde mi SM-G950F mediante Tapatalk

[joca133]

I am also skeptical about Vulcan making miracles, but I am hopeful it may allow more flexibility and better resource usage. The difference (in vr) from a medium to high settings in graphics is dramatic, even small 5fps gains matter in my case. Yes, it seems my hardware is always lagging when it comes to FSX, Arma, and DCS. It has been this way for decades. The main issue is money, I need more to keep up $$$.

[boedha68]

If i play, than 90% DCS, plan to buy a new PC, will be very interesting what change with Vulcan, better wait .....

 

 

 

That was my idea too :D

[BeastyBaiter]

I'm not looking forward this.

 

It will, as usually, decrease performance on same PC.

 

Assuming it's properly implemented, it will remove the single core bottleneck that exists when using a top end graphics card. I do not expect it to increase lr decrease performance other than that. This does have far reaching implications though, such as allowing more objects in missions and making 3 GHz quad cores from 10 years ago a whole heck of a lot more viable. It should also make VR at 90fps more reasonable

[msalama]

It will, as usually, decrease performance on same PC.

 

 

1) wrong

 

2) regardless, buy a new PC if this is what you believe

[SkateZilla]

I'm not looking forward this.

 

It will, as usually, decrease performance on same PC.

 

You are incorrect Please Provide Information to Backup this claim.

[corn322]

I'm not looking forward this.

 

It will, as usually, decrease performance on same PC.

 

 

Where you here for the upgrade from DCS 1.2 to 1.5? IIRC, it brought performance improvements for everyone.