Beta Testing 2.08

[STiFU 585]

29 minutes ago, Abusimplea said:

even in missions where the author added "fake SSAO" on some edges

I guess mission authors should refrain from adding these grime decals now that we finally have SSAO.

[peter_spy 1415]

4 hours ago, STiFU said:

I guess mission authors should refrain from adding these grime decals now that we finally have SSAO.

That kinda depends, AO is a generic effect and as such it won't replace dirt accumulation, water stains, etc.

[grayman 2861]

We'll continue to use stains because they provide effects other than edge occlusion.

[STiFU 585]

2 hours ago, peter_spy said:

That kinda depends, AO is a generic effect and as such it won't replace dirt accumulation, water stains, etc.

Sure! I was just getting at the Fake-AO decals. I've seen people use such decals on every step of a staircase. That's stuff we probably won't need anymore.

[HMart 290]

5 hours ago, STiFU said:

Sure! I was just getting at the Fake-AO decals. I've seen people use such decals on every step of a staircase. That's stuff we probably won't need anymore.

For stuff like geometry corners I don't think spending time on fake AO is now a good idea, large patch's of grime/dirt yes but small fake AO will just get superseded by the AO darkening.

Btw AO is a high end effect like you know and it will take a percentage of performance away from people, depending on the card it could be considerable and they will have to turn the effect off, so if you remove the hand made AO and the SAO, they will get less graphical fidelity, is a trade off that imo you guys have to think about.   

Edited May 12 by HMart

[peter_spy 1415]

1 hour ago, STiFU said:

I've seen people use such decals on every step of a staircase. That's stuff we probably won't need anymore.

Oh, absolutely, that's a bad idea in general, as it could increase drawcall count quite substantially.

[lowenz 554]

About the SSAO: https://gpuopen.com/fidelityfx-cacao/

[cabalistic 575]

47 minutes ago, lowenz said:

About the SSAO: https://gpuopen.com/fidelityfx-cacao/

I've seen it, but there's only an implementation for DX12, and given the sheer size of the shader file, that would be a monumental task to port Also, it's using compute shaders, which we don't (yet) support in the engine. That's also what stopped me from using Intel's ASSAO, which seems conceptually similar, and going for the older, but simpler SAO.

[HMart 290]

48 minutes ago, lowenz said:

About the SSAO: https://gpuopen.com/fidelityfx-cacao/

The current AO even being based on Nvidia code, imo runs reasonably well on my AMD gpu (RX 570X), I bet is not running as fast as it could, for obvious reasons but again runs well enough "in my case".

Also how many here have RDNA based AMD GPU's? I have a GCN based one. 

Cabalistic if it was you that implemented the TDM AO, what was the reason to prefer Nvidias SAO above lets say vendor agnostic Crytek SSAO? Documentation? Performance? Quality?  Just curious. 

[cabalistic 575]

Nvidia's SAO isn't really specific to Nvidia cards. All the techniques it uses are in some way also present in the FidelityFX implementation or Intel's ASSAO, those just add a couple additional advanced optimizations

Reasons for SAO: it was well documented, had an OpenGL sample implementation, good license, was reasonably simple to implement in a reasonable amount of code (definitely much less than the afore-mentioned ones!) and offers reasonable quality at reasonable performance - not fully state of the art, but also significantly better than some basic implementations that you often find in tutorials (and which I started with in my first attempt)

[lowenz 554]

Current AO with "Low" setting has the same 10% performance hit on NV and AMD GPUs having maxed out the rest (Soft Shadows on Medium, no Bloom).

Edited May 12 by lowenz

[peter_spy 1415]

Speaking about performance, any improvements for shadow maps for 2.08? Not sure if you did any comparison profiling with stencil vs. shadowmaps, but on my rigs stencil maps are much less taxing for both CPUs and GPUs. I think CPUs in particular are doing a better and better job with stencil shadows, with subsequent generations.

[cabalistic 575]

I think we'll have to consider shadow maps an ongoing experimental feature.

One significant hurdle (in my opinion) is that not all light types are implemented in shadow maps, which means that the engine will fall back to stencil for some lights and thus still generates stencil shadow volumes even when you select shadow maps. So in a way, you currently get the worst of both worlds with shadow maps

[lowenz 554]

Move them to the experimental section, no problem!

And yes, they're taxing the system and it's why I've accepted to set them to "medium" with the minimum radius (0.5).

It'a a good choice for everybody with a recent but not "top of the line" system, together with SSAO set to "low".

These 2 can really gracefully soften the scene.

Edited May 13 by lowenz

[peter_spy 1415]

1 hour ago, lowenz said:

These 2 can really gracefully soften the scene.

Stencil shadows + soft shadows (low) + SSAO (low) do the same, but at overall lower performance cost. At least that's what I'm seeing on several different hardware configs. Reasons stated by Cabalistic are one thing, but I suspect there are other factors like how cascading or render distance is set up  

[Araneidae 23]

I don't know if this beta test is an opportunity to look at this, but there seems to have been a map breaking regression, probably introduced in 2.07, which makes it impossible to complete The King of Diamonds without resorting to noclip.  Specifically, there is an key character in an upstairs room who is scripted to wake up and walk out of the room at some point in play, but he is stuck against the chair he is trying to walk away from.  Looking at reports of this problem in the forum (and on a posted playthrough on youtube) I'm pretty sure this started around the time 2.07 came out.

Is there any way this bug can be fixed?

Edited May 13 by Araneidae
typo

[nbohr1more 2008]

1 minute ago, Araneidae said:

I don't know if this beta test is an opportunity to look at this, but there seems to have been a map breaking regression, probably introduced in 2.07, which makes it impossible to complete The King of Diamonds without resorting to noclip.  Specifically, there is an key character in an upstairs room who is scripted to wake up and walk out of the room at some point in play, but he is stuck against the chair he is trying to walk away from.  Looking at reports of this problem in the form (and on a posted playthrough on youtube) I'm pretty sure this started around the time 2.07 came out.

Is there any way this bug can be fixed?

Chair and Bed navigation have been improved in 2.08. Are you able to reproduce this in 2.08 beta?

[Araneidae 23]

2 hours ago, nbohr1more said:

Chair and Bed navigation have been improved in 2.08. Are you able to reproduce this in 2.08 beta?

Yes, that's why I mentioned it here!  Do you want a save game?

[nbohr1more 2008]

3 minutes ago, Araneidae said:

Yes, that's why I mentioned it here!  Do you want a save game?

Yes please. Also, does this happen with both capped and uncapped FPS?

[Araneidae 23]

43 minutes ago, nbohr1more said:

Yes please. Also, does this happen with both capped and uncapped FPS?

That's a nuisance.  I can only upload 0.5MB, and the save game, even after compression, is 2.3MB.  I don't have anywhere else to put it, I'm afraid.

Also, I'm afraid I don't know how to set FPS capping, unless you're referring to the Vsync option, in which case I had it enabled and unsetting it didn't seem to make much difference.

[HMart 290]

21 hours ago, cabalistic said:

Nvidia's SAO isn't really specific to Nvidia cards. All the techniques it uses are in some way also present in the FidelityFX implementation or Intel's ASSAO, those just add a couple additional advanced optimizations

Reasons for SAO: it was well documented, had an OpenGL sample implementation, good license, was reasonably simple to implement in a reasonable amount of code (definitely much less than the afore-mentioned ones!) and offers reasonable quality at reasonable performance - not fully state of the art, but also significantly better than some basic implementations that you often find in tutorials (and which I started with in my first attempt)

Thanks for the explanation I see your point I would do the same. :)  

9 hours ago, cabalistic said:

I think we'll have to consider shadow maps an ongoing experimental feature.

One significant hurdle (in my opinion) is that not all light types are implemented in shadow maps, which means that the engine will fall back to stencil for some lights and thus still generates stencil shadow volumes even when you select shadow maps. So in a way, you currently get the worst of both worlds with shadow maps

 Then it was very well done because I don't remember ever realizing that some shadows were stencil based!

Btw IMO we can't fully compare stencil performance, to shadow maps, why, because shadow maps can cast shadows that stencil can't, like from alpha mapped textures, right now that is not much of a reason, because official alpha textures have the "noshadows" keyword but someone can make custom materials for their missions. So maps will always run a little slower than stencil even if max optimized. 

[duzenko 633]

21 minutes ago, Araneidae said:

That's a nuisance.  I can only upload 0.5MB, and the save game, even after compression, is 2.3MB.  I don't have anywhere else to put it, I'm afraid.

You do have a GMAIL account, right?

[Araneidae 23]

9 minutes ago, duzenko said:

You do have a GMAIL account, right?

No?  Don't understand (nor how it would help, to be honest).

[peter_spy 1415]

4 hours ago, Araneidae said:

there seems to have been a map breaking regression, probably introduced in 2.07, which makes it impossible to complete The King of Diamonds without resorting to noclip.

Did you have similar problem on 2.05 or 2.06? Usually it's a bad approach to fix mapping bugs by changing something in the engine. Some maps have mappers' errors left that manifest much later in conjunction with features that were added correctly. It's better to check the map before you ask to change something in the engine.

Edited May 13 by peter_spy

[JackFarmer 382]

@nbohr1more

@stgatilov

You guys were right, after updating to 2.08 it looks like...

1.  ...crashes during WIP loading have vanished and

2. ...in game crashes in WIP area with fire elementals have vanished as well!

Thank you, after adding final voice lines in the next days I will start beta/optimizing!

Jack

P.S. ....love the new swimming system, although I think I will have to add a few more breath potions as the lake is very, very deep!

P.P.S. ...and I have the feeling, that everything runs much faster now!

 

 

Edited May 14 by JackFarmer