Asynchronous compute and AMD´s TressFX Hair

[epyon1979]

Asynchronous compute and AMD´s TressFX Hair

Holla, my first post!

I read that CDprojectRED not wanted to use specific properties of the consoles, because the experience should be equal on each system:
Balazs Torok: We always want to provide the best possible experience to all our gamers regardless of the platform and so we are not aiming to develop special graphical features for any of them.
Source:http://gamingbolt.com/the-witcher-3-tech-interview-ps4xbox-one-performance-challenges-dx12-redengine-3-mods-more


But the pc user can use Nvidia´s Hairworks, why should CDRED not implement "TressFX Hair" for AMD systems, like consoles ?


My suggestions:

1.) Implementation of AMD´s TressFX Hair on consoles like Playstation 4

2.) The use of asynchronous compute (a Playstation 4 hardware property) to improve physics, frame-rate and streaming (Perhaps also for TressFx).


Mybe this is to much of work but it could improve the Playstation 4 version.:victory:



Best Greetings!

 

[GodfreyOfBouillon]

Do you want the PS4 to run the game at 10 fps?

 

[Razorig2]

Hypothetically speaking, IF the Witcher 3 was a ps4 exclusive do you think they could have hair works running at 30 fps? Consider that it's optimised fully. I know people bash the consoles but I'm sure later down the years of the ps4 all the untapped potential could be used to run a solid 30fps with hair works.
Look at the difference of sandbox games at the beginning of ps3 life cycle right up to the peak of the ps3, multi platform games like Gta 4 up to Gta 5. The leap in graphics, map scale, detail and I'm sure performance was improved as well. Could the Witcher 3 look and run better if it was made at a later stage for consoles/PC?

 

[epyon1979]

As far as i know, the idea behind asynchronous compute is to fill idle functions of a GPU with code for physics or something else. The Playstation 4 have something called 'volatile' bit which reduces the overhead of running compute and graphics together. This means you can selectively mark all accesses by compute as 'volatile,' and when it's time for compute to read from system memory, it can invalidate, selectively, the lines it uses in the L2 cache . When it comes time to write back the results, it can write back selectively the lines that it uses. So you do not have to flush the entire L2 cache (This is from an interview of lead architect Marc Cerny).

So with this you can implement functions without impacting the performance significantly. So no 10 fps Maybe 27 fps instead of 30 fps.

But this is all in theory. I am not an expert but i would like to know if the Devs would be motivated to use asynchronous compute for TressFX ,better physics and frame-rate.

 

[GodfreyOfBouillon]

As far as i know, the idea behind asynchronous compute is to fill idle functions of a GPU with code for physics or something else. The Playstation 4 have something called 'volatile' bit which reduces the overhead of running compute and graphics together. This means you can then selectively mark all accesses by compute as 'volatile,' and when it's time for compute to read from system memory, it can invalidate, selectively, the lines it uses in the L2 cache . When it comes time to write back the results, it can write back selectively the lines that it uses. So you do not have to flush the entire L2 cache (This is from an interview of lead architect Marc Cerny).

So with this you can impelment functions without impacting the performance significantly. So no 10 fps

But this is all in theory. I am not an expert but i would like to know if the Devs would be motivated to use asynchronous compute for TressFX ,better physics and frame-rate.

Nice post, I didn't knew about that.
But I do think that the framerate/pop in issues are the ones that should be addressed ASAP.

 

[epyon1979]

I agree with you completely.

I thought that asynchronous compute could help "CDprojectRED" to improve such things like frame-rate and pop ins for Playstation 4. Sony has put a lot of work to ensure that this could be a helpful property. The thing with TressFx is just a dream of my

 

[griever0507]

I agree with you completely.

I thought that asynchronous compute could help "CDprojectRED" to improve such things like frame-rate and pop ins for Playstation 4. Sony has put a lot of work to ensure that this could be a helpful property. The thing with TressFx is just a dream of my

So you mean to say The Witcher 3 haven't used the full potential of our PS4 system? This is new to me and thanks for sharing this information. Well hopefully CDPR can try to improve things and follow your suggestion as this will be a big improvement all through out The Wicher 3's performance.

 

[epyon1979]

So you mean to say The Witcher 3 haven't used the full potential of our PS4 system? This is new to me and thanks for sharing this information. Well hopefully CDPR can try to improve things and follow your suggestion as this will be a big improvement all through out The Wicher 3's performance.

You have summarized it well. Beside the standard route of optimising the implemented functions, the shifting of functions like physics, post-process effects,... on to idle GPU resources (which are always there) by using GPU-compute is a way to get higher fps and/or new effects without impacting the performance significantly.

But i read that this is not easy to do. One need to break complex serial workloads into many parallel tasks and calculate these tasks via GPU-compute during idle GPU moments. It is something to master.

I thought if CDprojectRED is ambitious enough this could be a route.

The studio "Q-Games" developer of the game "The Tomorrow Children" are using the asynchronous compute properties of the Playstation 4 and they recommend it.
Quote1:"Really if you’re working on a AMD GCN GPU, I can’t recommend looking into async compute enough!"
Quote2:"Depending on what kind of work you are doing and what dependencies it has, its possible to hack it in later in the dev cycle and still get very good results. But like most things in life, you can often get far better results if you plan for it upfront."

Source: http://gearnuke.com/tomorrow-children-dev-async-compute-benefits-multiplatform-development-across-ps4xbo/

 

[epyon1979]

I want to mention one more point to this thread:

I am sure the Dev`s of CDproject RED are superb, there is no question about this. I dont know what kind of programming techniques they are using, but i would like to give one more information about another studios technique and hope it could be useful.

Christian Gyrling from Naughty Dog presented a talk with the title : Parallelizing the Naughty Dog Engine Using Fibers. As i understand this is useful when you are CPU-bound. They described how they managed to archive 60fps in the game "The Last of us remastered".

Source: http://www.gdcvault.com/play/1022186/Parallelizing-the-Naughty-Dog-Engine


So that was all i wanted to say concerning this thread. Thanks

 

[DXMG]

Hypothetically speaking, IF the Witcher 3 was a ps4 exclusive do you think they could have hair works running at 30 fps? Consider that it's optimised fully. I know people bash the consoles but I'm sure later down the years of the ps4 all the untapped potential could be used to run a solid 30fps with hair works.
Look at the difference of sandbox games at the beginning of ps3 life cycle right up to the peak of the ps3, multi platform games like Gta 4 up to Gta 5. The leap in graphics, map scale, detail and I'm sure performance was improved as well. Could the Witcher 3 look and run better if it was made at a later stage for consoles/PC?

it doesnt work that way anymore.... there isnt any "untapped potential" in consoles anymore.

it used to be that way because the consoles used laregly non-standard, often proprietary hardware. There was untapped potential becuase no one was familiar with the instruction set, how the GPU performed, etc.

the PS4 is a PC with an AMD APU in it. In other words, its an x86 computer, running an old and underpowered GPU. There isnt any untapped potential in it.

 

[Gilrond-i-Virdan]

Parallel physics computation with graphics at the same time is supported in Vulkan (and also DX12). See this thread:
http://forums.cdprojektred.com/threads/31539-Witcher-3-and-the-possibility-of-DX12-and-Vulkan-support-post-release

 

[epyon1979]

it doesnt work that way anymore.... there isnt any "untapped potential" in consoles anymore.

it used to be that way because the consoles used laregly non-standard, often proprietary hardware. There was untapped potential becuase no one was familiar with the instruction set, how the GPU performed, etc.

the PS4 is a PC with an AMD APU in it. In other words, its an x86 computer, running an old and underpowered GPU. There isnt any untapped potential in it.

But there is untapped potential. For PC and for the consoles too. This untapped potential are the idle GPU resources. In contrast to the PC and Xbox one, the Playstation 4 has a lot of possibilities (hardware and software wise) to use this resources more efficient, like:

1.) 64 sources of compute commands (New AMD Pc cards do also have this property)
2.) Another bus to the GPU that allows it to read directly from system memory or write directly to system memory, bypassing its own L1 and L2 caches.
3.)The 'volatile' bit (it reduces the overhead of running compute and graphics together on the GPU)
Source: http://www.gamasutra.com/view/feature/191007/inside_the_playstation_4_with_mark_.php?page=2

This things are made to make GPGPU or asynchronous compute more efficient. So on Playstation 4 as on Pc or Xbox one , one can use or try to master GPGPU (to use the untapped potential which is located on the GPU.) But on Playstation 4 it is more efficient.
Source: http://www.redgamingtech.com/ubisoft-gdc-presentation-of-ps4-x1-gpu-cpu-performance/

My point is that there is indeed a lot of untapped potential. But this is hard to master (GPGPU).

---------- Updated at 07:07 AM ----------

Parallel physics computation with graphics at the same time is supported in Vulkan (and also DX12). See this thread:
http://forums.cdprojektred.com/threads/31539-Witcher-3-and-the-possibility-of-DX12-and-Vulkan-support-post-release

Ahh thanks! That is interesting.
I hope CDprojectRED manage the specific Playstation 4 implementation too, because the Ps4-API is made for such things.

 

[davidmunozu]

it doesnt work that way anymore.... there isnt any "untapped potential" in consoles anymore.

it used to be that way because the consoles used laregly non-standard, often proprietary hardware. There was untapped potential becuase no one was familiar with the instruction set, how the GPU performed, etc.

the PS4 is a PC with an AMD APU in it. In other words, its an x86 computer, running an old and underpowered GPU. There isnt any untapped potential in it.

You are really wrong.

 

[mutatedmeow]

wow this is really a nice thread. lots of info about stuff hardware and software alike good to have some hate-free thread every now and then.thank you for sharing your knowledge! i have a question tho what games on ps4 use asynchronous compute?

 

[epyon1979]

wow this is really a nice thread. lots of info about stuff hardware and software alike good to have some hate-free thread every now and then.thank you for sharing your knowledge! i have a question tho what games on ps4 use asynchronous compute?

Thank you!

1.) For now i am aware of the game "The Tomorrow Children" from the studio "Q-Games" . They recommend async-compute.

Quote1:"Really if you’re working on a AMD GCN GPU, I can’t recommend looking into async compute enough!"

Quote2:"Depending on what kind of work you are doing and what dependencies it has, its possible to hack it in later in the dev cycle and still get very good results. But like most things in life, you can often get far better results if you plan for it upfront."

Source: http://gearnuke.com/tomorrow-childre...across-ps4xbo/

2.) The game "Infamous Second Son" from "Sucker Punch Productions". They used some GPGPU stuff (I think for particles effects). But i am not aware if they implement it via asynchronous compute or the traditional rout of GPGPU.

Source: http://www.redgamingtech.com/infamous-second-son-post-mortem-part-2-ps4-performance-compute-particle-system/

I heard and read (but i don’t know where) that asynchronous compute is in coming for future titles.