Jump to content
dna_uk

DirecxtX 12 for ArmA 3?

Recommended Posts

That's a change in tone from the CEO stating he hoped it would be their "golden egg" a few months ago.

 

"He hoped" is the important part, I'll guess.

Share this post


Link to post
Share on other sites

"He hoped" is the important part, I'll guess.

Well, hoping that someone else is going to fix performance problems that have been long-known with this engine, shouldn't be that encouraging in the first place... :)

Share this post


Link to post
Share on other sites

Let's say NOW the graphics task on the CPU core 0 takes about 16ms. Ideally on DX12/Vulkan you can spread that on a 4 core CPU to about 4ms/core which frees 12ms on the main thread for AI, simulation, etc. On a 6 core you get ~2,67ms/core which takes you to ~ 13,32ms extra performance ( dx12 scaled to around 6 cores in the preview 3DMark test and Mantle to around 8 if I'm not mistaken). Now, if they go even further and NOT render what the player CANNOT see (like they're doing now), you should see even higher gains. Basically what we have with their games so far, is a vision of a game/gameplay that is hold back not by the hardware, but rather by the "old" way of doing things.

It's the same with the AI that can be done on the GPU... like those 3 GPUs you have in your own system. :D

 

Will this happen? I don't know. Probably not since their next engine is Enfusion as far as I know (a mix between this and another one). With a 2500k@4,5GHz I've got around 40-50% extra performance in Mantle vs. DX11.2, plus A LOT smoother gameplay (BF4). And that's a waaaay better optimized engine than this is.

 

 

That could be because the CPU is now forced to prepare more data to be sent to  GPUs. More data on the main thread/core, less processing time available for other tasks - AI, simulation, etc.

Or maybe just because that after 3 years we still dont have a Crossfire Profile for ARMA 3 and also because LOD transition does not work as it should, in multi gpu (AMD) configurations. These are due to issues with LOD transition, not CPU related.

Anyway, now that we solved DX12, can we go back to DX9 and get your happiness back?

Share this post


Link to post
Share on other sites

you do realize crossfire is all in hands of AMD profiles and AMD drivers ...

as developer you touch nothing on it ...

  • Like 2

Share this post


Link to post
Share on other sites

''All of these changes are unrelated to our investigation into DirectX 12 by the way. This has not yet yielded useful results, so we don't have any concrete news on that front.''

 

Really? I thought it was clearly stated that Tanoa would come with dx 12. Did BI actually state that or was it always a big maybe?

Share this post


Link to post
Share on other sites

CEO says "question is not IF, but WHEN"... but, I don't recall them ever explicitly stated that it was to come with Tanoa... I think that was just the conclusion folks drew given the relative timing of that statement and the announcement of the expansion.

Share this post


Link to post
Share on other sites

you do realize crossfire is all in hands of AMD profiles and AMD drivers ...

as developer you touch nothing on it ...

 

The engine still has to support multi gpu or else (just like in JS 3 for instance), the IHV can't do much.

''All of these changes are unrelated to our investigation into DirectX 12 by the way. This has not yet yielded useful results, so we don't have any concrete news on that front.''

 

Really? I thought it was clearly stated that Tanoa would come with dx 12. Did BI actually state that or was it always a big maybe?

 

They pretty much confirm it when asked about the new features (4:47 onward):

 

Dx12 is gonna be something that will really kick on the performance which is a really important part of the ArmA 3 experience; we really wanna nail that down for the expansion.

 

 

Anyway, let's just hope that at least their next engine is approached just like Star Ctizen has done with Cry Engine - take something and build it to what you want, by harvesting the most of what there is available. For me the support of this game and their games in general stops here. Bough a lot of products, but at the moment even buying to support the developer is "no go" option. Nothing has changed.

Share this post


Link to post
Share on other sites

''All of these changes are unrelated to our investigation into DirectX 12 by the way. This has not yet yielded useful results, so we don't have any concrete news on that front.''[/size]

 

Really? I thought it was clearly stated that Tanoa would come with dx 12. Did BI actually state that or was it always a big maybe?[/size]

 

Alright so I understand that Dx12 to be fully utilized needs fundamental rework on the lower engine part, I understand that development company

has different priorities and lots of different tasks to do.My question is more in line with pcgamer show when they first announced that they are

looking into dx and that was in june, which was 8 months ago, and now I'm sitting and wondering how can their work I quote "not yet yielded useful results"

in such long time?It just seem strange.I mean I expect this kind of comments for things "that we want to have" but they would took to long to make like

when marksman DLC was around of "bullet in chamber" for example.Did they not really set priorities for this task (performance, Dx12 mentions) then?

Share this post


Link to post
Share on other sites

I just dont care about DX12 beacuse i stick to the Win7, nobody force me to move on this appstore called Windows10. Its real shame they decided to move on DX12 instead of Vulcan. I guess Vulcan will be relased when they start to "optimalise" the game engine.

Share this post


Link to post
Share on other sites

Display: 5280*1050= 5,54 Mpx  (1080p is 1920*1080= 2,07Mpx)

My sis: 2500k@4,5GHz (for test was at 3,3GHz as well), 16GB RAM @1600Mhz, R290@ 1090/1380MHz. Terrain and Objects details low, 2k for view distance and object distance, the rest pretty mach maxed, except MSAA and some blur and co. turned off.

4,5GHz ArmA 3: http://imgur.com/NbTtGWw  48fps
3,3GHz ArmA 3: http://imgur.com/V7xIOKd 38fps

 

The CPU was around ~ 50% usage in both cases, so it's not the problem - as in: it has A LOT of room for improvement in terms of raw performance IF the engine will actually be optimized properly.

 

PS: 48fps to render that simple scene is frankly a joke.

Share this post


Link to post
Share on other sites

I just dont care about DX12 beacuse i stick to the Win7, nobody force me to move on this appstore called Windows10. Its real shame they decided to move on DX12 instead of Vulcan. I guess Vulcan will be relased when they start to "optimalise" the game engine.

Did someone from bis actually say they have no interest in looking at vulkan? It hasn't been released no? Or am I missing something.

Share this post


Link to post
Share on other sites

Did someone from bis actually say they have no interest in looking at vulkan? It hasn't been released no? Or am I missing something.

I hasnt been released, but my guess it will be released WHEN THEY START TO IMPLEMENT DX12(then they will be able to aply Vulcan too). So far they didnt sayed anything about vulcan.

Share this post


Link to post
Share on other sites

I hasnt been released, but my guess it will be released WHEN THEY START TO IMPLEMENT DX12(then they will be able to aply Vulcan too). So far they didnt sayed anything about vulcan.

Ahhhh I see what you mean. Cool cool.

  • Like 2

Share this post


Link to post
Share on other sites

PS: 48fps to render that simple scene is frankly a joke.

It's not the scene rendering alone that makes the fps that low. AI (pathfinding etc.) has a lot more to do with it. As I can see you're playing a mission (official campaign?) and missions with AI lower the fps. Though if DX12/Vulkan would be utilized then the scene rendering could use more GPU and give the main CPU core more room to calculate AI and scripts etc. Hard to say how much that could increase the fps.

 

Btw your GPU usage is at full on that 4,5GHz 48fps. The fps is limited by your GPU at that point. Can't say if it's 1fps or 10fps but it's limited by the GPU power.

 

//Edited: Not to utilize more CPU through draw calls on all the core but to move draw calls in GPU to utilize GPU more and give naturally CPU more room to make other calculations.

Share this post


Link to post
Share on other sites

actually, lets just hope that BI totally misses the DX 12 train and gradually picks up onteh Vulcan/AMD GPU Open stuff. Will we see any great benefits from that during arma 3's lifetime? I highly doubt that. But, if they assimilate, bit by bit, what these new, open and os independent technologies have to offer, and develop their new Enfusion engine (once Vulcan is finally released) around it, we should see some real performance improvements in the arma series in the future. Scrap dx 12 for Apex, and focus on the longterm goal of building a highly performant engine for future titles in the arma series.

Share this post


Link to post
Share on other sites

actually, lets just hope that BI totally misses the DX 12 train and gradually picks up onteh Vulcan/AMD GPU Open stuff. Will we see any great benefits from that during arma 3's lifetime? I highly doubt that. But, if they assimilate, bit by bit, what these new, open and os independent technologies have to offer, and develop their new Enfusion engine (once Vulcan is finally released) around it, we should see some real performance improvements in the arma series in the future. Scrap dx 12 for Apex, and focus on the longterm goal of building a highly performant engine for future titles in the arma series.

how can they even think about implementing something that is not released yet? (vulcan)

for whatever reason, some of you consider that the main gaming platform will change from windows to linux which is simply not gonna happen. I am pretty sure the next BI project (not even sure that is gonna be an iteration of the existing franchise - arma3) will still be platform dependent (windows) and will still use DX instead some AMD lala land. (case of point is the really wide spread mantle has seen - as in it didn't)

Share this post


Link to post
Share on other sites

how can they even think about implementing something that is not released yet? (vulcan)

for whatever reason, some of you consider that the main gaming platform will change from windows to linux which is simply not gonna happen. I am pretty sure the next BI project (not even sure that is gonna be an iteration of the existing franchise - arma3) will still be platform dependent (windows) and will still use DX instead some AMD lala land. (case of point is the really wide spread mantle has seen - as in it didn't)

And he shows he doesnt understand the argument. Vulkan (SPELLED VULKAN!!!), is OS Agnostic. AKA ALL operating systems. Sure, it is useful when porting to Linux. But, we are not assuming that Linux will be the new platform of choice. ,

Also, I don't think you understand the makeup of DX12. Especially since AMD gave Mantle code to  Khronos Group's Vulkan. And DirectX 12 is an imitation of Mantle. Vulkan is more heavily based on Mantle than DX12.

 

BI is making the important steps towards other operating systems. especially as we see Microsoft trying to merge XBOne and PC into  one platform. Vulkan, being the more industry standardized API than even DX12 considering OpenGL is an industry standard and it too is by Khronos Group. Vulkan and OpenGL (two separate APIs) are backed/supported by more companies and markets than DirectX. That is why you wont see DX on your phone, tablet, or other mobile device excluding a laptop. OpenGL for mobile markets uses OpenGL ES. There are many variants of OpenGL for each segment. For PC Gaming, we have OpenGL 4.5 which is feature comparable to DX11. Most game devs still use OpenGL 3.x though so they are shooting themselves in the foot.

 

Khronos Group built Vulkan separate from OpenGL meaning Vulkan and OpenGL are not the same and Vulkan does not use or require OpenGL code. In fact, Vulkan uses the SPIR-V intermediary language to translate ANY language into Vulkan.

You can even throw DX11 and DX12 code at it. It is easier to port a game from DX to Vulkan than it is to port Vulkan to DX.

 

That means that BI CAN implement Vulkan without rewriting their engine or switching away from DX. They can make it so that even if they make DX12 the standard (thereby forcing the move to WinX), non WinX users can still get the benefits and play.

 

These white papers are very informative and scrumptious:

https://www.khronos.org/registry/spir-v/papers/WhitePaper.pdf

  • Like 2

Share this post


Link to post
Share on other sites

I think we should take a break here and wait until Vulkan 1.0 has been released. From that moment it probably takes at least 1.5 to 3 years until good implementations in programs and games make it to the market. If i think of it, it seems actually good that BI is not going full throttle for dx12. Without deep engine adjustments, there probably is not much to be gained, and in the long run, Vulkan is the much wiser choice. From an economic standpoint, game developers must ask themselves whether they wanna be a taker or market maker (serious gaming on something else than Windows). I'd sure like to know what the price tag is between moving (the cost of developers to acquire the knowledge that is) to dx 12 or Vulkan?

Share this post


Link to post
Share on other sites

Tomorrow is Khronos Vulkan meetup http://www.meetup.com/Khronos-Paris-Chapter/events/227540099/. Perhaps someone there will finally announce when they plan to release the damn thing.

 

Android will use it and Oculus most probably too.

http://android-developers.blogspot.cz/2015/08/low-overhead-rendering-with-vulkan.html

 
Some more reading:

https://www.khronos.org/assets/uploads/developers/library/overview/vulkan-overview.pdf

https://developer.nvidia.com/engaging-voyage-vulkan

 

It'd be sad if BI chose another vendor lock-in for another 5 years. Wait, evaluate, incline to standards. <3

  • Like 2

Share this post


Link to post
Share on other sites

 

Perhaps someone there will finally announce when they plan to release the damn thing.

 

Right? They said the release was "imminent". That's really good to hear!  Hope it shreds DX12.

  • Like 2

Share this post


Link to post
Share on other sites

Without deep engine adjustments, there probably is not much to be gained,

Pretty much why I don't expect substantive news either way on this front in Arma 3's case/life cycle -- not when its engine has become "legacy technology" with the engine merge into Enfusion which will not be coming to Arma 3 -- even if Bohemia designers aren't remotely thinking about Arma 4 yet.

Share this post


Link to post
Share on other sites

the real problem for me right is the following: playing BI designed missions arma 3 completely unmodded runs pretty well actually. But adding mods to this always is detrimental to performance. Even if i am careful to only load the mods i think i need for a particular arma mission, it just costs fps. And i feel its seldom one singular mod that tanks, but cummulatively it adds up. I guess we need to relax and hope for the best.

 

Did you guys watch those dx12 demos of that strategy game Ashes of the Singularity? There is a huge difference between dx11 and dx12 performance wise because there are hundreds if not thounsands of ai in each scene. Somehow, those improvements technology wise must make it to the arma series. 

  • Like 1

Share this post


Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in



Sign In Now

×