Any explanation for move to Deferred Shading in 2.5.1 from Devs?

[Mars Exulte]

That's not always possible, though. If you keep the same engine and are merely refining it, adding tweaks, etc, yes. That's very limiting, though. You can add effects and potentially streamline performance through optimisation. This is a new graphics engine altogether, though. When you went from Il2:1946 to Il2:BoS, the hardware requirements increased with the improvement in visual and physics fidelity.

 

As for 'improvements received', the new Caucasus and new maps are being designed for this going forward. Normandy is an outlier, made by an external party and not incorporating a lot of the new tech properly/at all (yet?). For example SpeedTrees is on Caucasus, but not Normandy (it's using old school manually created and placed assets) as a result it draws more power. It has never run as well as the other two maps (NTTR is open desert mostly, so FPS friendly by nature. Caucasus works well, however).

 

You're thinking short term, right NOW, and seemingly in relation to a specific map. Normandy was made prior to all this and isn't a good reference point. The benefits come for products designed to take advantage of them, future environments like Persian Gulf, urban environments, etc. If/when Normandy gets updated to some of this stuff, it'll probably pick it up as well, as the large number of trees and light sources present are exactly the sort of thing this is geared toward improving.

 

This engine has to be 'good enough' going forward for the next 10-15 years probably before they do another overhaul or it's pointless. In five years my 1080ti will be a mid range PC, in ten years, it will be in a trash can somewhere.

 

If they don't future proof, the graphics that were already very dated would long since have become a hindrance to the game's saleability. This isn't an accounting database or server backend, it's a videogame you somehow have to entice people to keep playing and paying for. Going from 35k polys on the Su-25T to 300k polys on the MiG-29 drastically increased overhead.

 

People won't play it all if they don't move things forward, much less pay $80 for a module in 2020 that looks like it was made in 2000. The benefits will come to future plans, especially as they optimise it. It's early days yet, and it's been a rough launch due to the scale of the changes. It'll improve, and make things possible going forward that wouldn't have worked on the old engine.

[Charlie-1]

That's not always possible, though. If you keep the same engine and are merely refining it, adding tweaks, etc, yes. That's very limiting, though. You can add effects and potentially streamline performance through optimisation. This is a new graphics engine altogether, though. When you went from Il2:1946 to Il2:BoS, the hardware requirements increased with the improvement in visual and physics fidelity.

 

As for 'improvements received', the new Caucasus and new maps are being designed for this going forward. Normandy is an outlier, made by an external party and not incorporating a lot of the new tech properly/at all (yet?). For example SpeedTrees is on Caucasus, but not Normandy (it's using old school manually created and placed assets) as a result it draws more power. It has never run as well as the other two maps (NTTR is open desert mostly, so FPS friendly by nature. Caucasus works well, however).

 

You're thinking short term, right NOW, and seemingly in relation to a specific map. Normandy was made prior to all this and isn't a good reference point. The benefits come for products designed to take advantage of them, future environments like Persian Gulf, urban environments, etc. If/when Normandy gets updated to some of this stuff, it'll probably pick it up as well, as the large number of trees and light sources present are exactly the sort of thing this is geared toward improving.

 

This engine has to be 'good enough' going forward for the next 10-15 years probably before they do another overhaul or it's pointless. In five years my 1080ti will be a mid range PC, in ten years, it will be in a trash can somewhere.

 

If they don't future proof, the graphics that were already very dated would long since have become a hindrance to the game's saleability. This isn't an accounting database or server backend, it's a videogame you somehow have to entice people to keep playing and paying for. Going from 35k polys on the Su-25T to 300k polys on the MiG-29 drastically increased overhead.

 

People won't play it all if they don't move things forward, much less pay $80 for a module in 2020 that looks like it was made in 2000. The benefits will come to future plans, especially as they optimise it. It's early days yet, and it's been a rough launch due to the scale of the changes. It'll improve, and make things possible going forward that wouldn't have worked on the old engine.

 

I appreciate what you are saying and agree with you. Standing still isn't an option. What I am yet to hear is why ED chose DS as the best way forward? I presume they made some statement a while back or maybe we're not privileged to that information and never find out. I've said it before on here that there are other video games / sims around with much better photo-real graphics already with much better performance, but won't go there.

 

I guess I've learned never to pre-order or buy anything from ED again until they are released, established and plenty of info /testimonials available to endorse their state of completion and performance. I would have saved on the Normandy map and probably the Hawk too, which for some reason is an FPS killer.

 

In the meantime, at least have the option to still use the older versions like 1.58 and 2.5 , which is unusual, but very good of ED indeed.

[Sn8ke]

I've spent weeks and weeks playing with settings trying to optimise performance of 2.5.1 beta version to try and get both performance and image quality anything like as good as was before DS was introduced and especially in VR. Going back to V1.58 is like a breath of fresh air and wow for quality and immersion. With this in mind, I'm sure I'm not the only devoted player wondering why the move? Has this been explained anywhere please? :helpsmilie:

 

EDIT: For reference, observations of Deferred Shading so far.

 

1) Detail in image quality worse, cockpit text illegible without zooming in.

2) FPS performance down by 30%

3) Cockpits either too dark or rest of world too bright.

4) Sea is almost black except where suns reflections are.

5) Night lights barely visible

6) Delay re-shading scene when changing views. Especially from map back to cockpit view.

7) Plane surface appear over-shiny

 

I read through this thread today and didn't notice any official comment from ED. I been flying DCS since its inception from the original Flanker series. I've built several computers over the years to match the hardware resource-hungry simulations of today.

 

Today, I have a strong system; an i7 8700K overclocked to 5.3ghz, 32gb DDR4 RAM, a 1080Ti SC, and an ultrafast Samsung M.2 PCIE SSD. During the rollout of 2.5 beta, I had a solid 60 FPS displayed on my 60hz 4K monitor at all times with deferred shading off. With Deferred shading on, as is the only option ion 2.5.1, I have noticed and experienced many of the same things as you:

 

1) Image detail quality went from what I would describe as "Immersive" to "Cartoon-like."

2) FPS down 15 FPS, consistently (25% decrease)

3) There is no balance between cockpit color/lighting, and the outside world. Each is either too light/too dark. No "Sweet spot."

4) Same observation

5) Same observation

6) Stutter and delay when changing views very prominent.

7) With the very unrealistic lighting effects courtesy of deferred shading, the painted A-10C appears to have a chrome-like shine at different angles.

8) Kind of a rough situation to be releasing the F/A-18C into?

[Charlie-1]

On item 1), I'm sure I read doing research that loss of defenition of edges is a downside of DS due to the fact it has to align the second shading render with the first 'geometry' render absolutely exactly, which may explain the fuzzy text on cockpit displays, gauges etc? I don't know, but I'll try and find where I read it.

[SkateZilla]

** Not an Official ED Response **

(But I'll explain it for you without getting too technical)

 

PBR and Deferred Lighting and Deferred Shading are all Separate Things,

 

However, PBR Relies on them to render Accurately, Thus PBR Requires Essentially Deferred Lighting and Shading to be Enabled.

Especially considering all light sources are Dynamic and there are hundreds of thousands of them across the terrains and hudreds in cockpits, and they must all interact w/ the Textures.

 

So, We'll Start w/ Why PBR?:

w/ 1.1.1.1 -> 1.2.16 All Models were rendered Using the Old Rendering which required 7 Surface Properties:

Diffuse: From Diffuse Texture

Specular Color: From Diffuse Texture Merged w/ Specular Texture

Specular Power: From Specular Texture (RED Channel)

Specular Glossiness: From Specular Texture (GREEN Channel)

Reflection Level: From Specular Texture (BLUE Channel)

Reflection Blur: Globally Set in 3DS Max by Material

Ambient Occlusion: Baked onto Diffuse Layer (Unrealistic)

 

w/ 1.5.0/2.0.0 and Deferred Shading / PBR Enabled.

Albedo: from Diffuse Texture

Microsurface (ED Calls it Roughness): from RoughMet Texture (GREEN Channel)

Reflectivity (ED Calls it Metallicness): From RoughMet Texture (BLUE Channel)

Ambient Occlusion: From RoughMet Texture (RED Channel).

 

So, You go from 7 Variables to Control how a Surface Looks to 4 (Counting A.O., But not all models use it).

 

Albedo is the Diffuse and Specular Colors of the Surface.

 

Roughness as ED Calls it, Taking the Place of Reflection Blur (Which is now Dynamic by Pixel instead of Material), Specular Glossiness, Specular Power, Specular

Controls how Rough or Smooth a Surface is, Smoother Surface will scatter light less, and reflect light and surrounding clearly vs not. (Smooth will appear polished/glossy, Rough will appear Matte)

 

Metallicness as ED Calls it, Takes the Place of Reflection Level

Controls how Mirror like a Surface is, Metallic Surface will have a Clear Vibrant Reflection of the Environment and Non Metals will have a faded Reflection of the Environment.

 

Ambient Occlusion, is not a Separate Map and not Baked onto the Diffuse,

Under the Deferred Lighting Engine, the A.O. Map is only drawn when in shadow (ie it will not be drawn on the model if exposed to direct Sunlight/Light Source)

 

 

In Practice Roughness is being used Normally by ED (Controls the Surface and How it scatters light etc)

 

Metallicness however in Practice is Normally a 2 Color Option, White / Black, Metal or Not. Metal Items will have Environment Reflections, Non-Metals wont.

Since ED is using the Metallicness Map as a Gradient (All 256 values), I would Call it by it's normal PBR name, Reflectivity Map. as the Gradient 0-255 controls how Apparent the Reflection is on a smooth surface.

(Examples, RoughMets being used on Canopy Glass w/ 100% Metallic Value, Gradient is being used as a Reflection Value and not Metallic Value, You cant have half metal, A substance is either metal or dieletric)

 

 

PERFORMANCE:

As I Stated in the opening blurb, due to the type of environments we have in DCS and the introduction of PBR, Moving to 2-Pass Shading (Deferring Shading) and 2-Pass Lighting (Deferring Lighting) was needed to use PBR.

 

The Major Improvement is how Dynamic Lights all appear, and how objects render around the dynamic light sources w/ the PBR.

 

The Draw Back w/ Deferred Lighting/Shading has always been Performance w/ Brute force (Spatial) Anti-Aliasing (MSAA, SSAA, EQAA and CSAA Etc), which is why the DCSW2.5 MSAA Limit is 4x instead of 16xQ, etc.

ED is aware of the Performance issues w/ Spatial AA, so hopefully something emerges on that front.

 

For now the obvious work around is to move to PostProcess / Shader AA, (FXAA, MLAA, SMAA etc), which will separate the Anti-Aliasing Process from the Rendering / Shading and Lighting Process.

 

This is why in DCS 2.5 when Deferred is Enabled that Your FPS and Performance are lower than 1.5.0/2.0.x w/ Deferred Disabled.

 

Shader Based AA and High Resolution Screens are making Traditional Spatial AA Obsolete, soon every PC Screen will be 4K Regardless. just as 1080p became the standard after 1024x768.

 

APPEARANCE OF OLD ASSETS:

As Anyone can tell, alot of the Pre-PBR Models look overly shiney/glossy at angles, this is because they have not been Converted to native PBR, instead they use a Converter that converts the Specular File Channels for use w/ PBR. the Values are stored in a JSON File.

Eventually They'll all be replaced w/ Generated RoughMet (Microsurface/Reflectivity) Files.

 

For the Most part, most of the models look presentable in the grand scheme of things w/ the JSON FIles, most of the "Glossy" and Odd Appearances happen on Liveries that use Custom/Unique Specular Files, that do not convert to PBR Very well when using the Conversion Parameter Values in the JSON File.

 

 

BIGHT / DARK VARIANCE:

Outside of the already stated Lighting Adjustments (Night Lighting/Sources etc), Moving to PBR and Rendering Objects more Realistically w/ Realistic Dynamic Lighting also brings with it to changes in how Cockpits Appear.

 

Even When Calibrated w/ simple calibration tools, everyone's Displays are not the same, there are differences in Back Light Brightness, and Contrast at angles etc etc, so , the Same Cockpit might appear DARK for one user and BRIGHT for another, this is the reason for the Gamma Slider.

 

Night Lighting and other things are being tweaked to Fix the Lighting as well as help bring a huge universal sweet spot, so that Cockpits and the Environment would look the same for all users.

Edited May 17, 2018 by SkateZilla

[Charlie-1]

Thank you very much for the detailed response and explanation.

 

Whilst this is soaking in, can you just clarify the point about

 

This is why in DCS 2.5 when Deferred is Enabled that Your FPS and Performance are lower than 1.5.0/2.0.x w/ Deferred Disabled.

 

Is this because MSAA of some degree is enabled behind the scenes automatically with DS on in 2.5?

 

The performance hit is one issue, but the other is a drop in clarity in text on cockpit displays etc, and especially in VR. If my understanding of the above is correct, the MSAA may explain the fuzzy appearance of edges?

[SkateZilla]

Deferred Shading and Spatial AA Have always been an issue, which is why other Engines in the industry have Post Process AA when using Deferred Shading

 

Because Previously, in 1.5/2.0, a user could Run 8x, 8xQ,8xS, 16x, 16xQ MSAA and still have Fluid Frames Per second,

 

now because of performance issues, users are dropping to 2x and Maybe 4x MSAA on high end GPUs.

 

The lack of Higher MSAA in VR, makes image less clear in the VR Headset. You could turn off MSAA and Increase the Pixel Density a few notches to clear up VR Text.

 

I run 4x MSAA on my 1080P Screen to clear up aliasing, however, on my 4K Screen, MSAA is Off Entirely and there isnt no visible aliasing due to pixel size.

Edited May 17, 2018 by SkateZilla

[will-]

But wouldn’t all your issues go away if you could run your settings on high/max? Using a i3 and 16gb of ram will be your downfall.

[Charlie-1]

But wouldn’t all your issues go away if you could run your settings on high/max? Using a i3 and 16gb of ram will be your downfall.

 

Maybe they would, but from what I am seeing my CPU and RAM aren't maxing out at all, but my GPU is. I chose the i3 because of its excellent single core process capability. Besides, DCS ran very well indeed along with other similar sims in VR before the introduction of Deferred Shading, so why should I shell out more money when there are no benefits to see in 2.5.1 or I can play and spend much less money on assets in other games instead?

 

Other users are seeing similar issues to myself with much higher spec machines to.

 

Here's what Steam VR test made of my system https://forums.eagle.ru/attachment.php?attachmentid=183683&d=1525165360

Edited May 17, 2018 by NAKE350

[Whisper]

The Draw Back w/ Deferred Lighting/Shading has always been Performance w/ Brute force (Spatial) Anti-Aliasing (MSAA, SSAA, EQAA and CSAA Etc)

Isn't VR PD downscaling ALSO bringing performance problems in the same order of magnitude than Spatial AA ? (As far as I understood the algorithms used to downscale are very akin to Spatial AA).

In which case this very choice to force usage of DF is a killer for VR users and I don't understand the decision.

[SkateZilla]

Pixel Density is Essentially Super Sampling.

 

I'll prolly check performance in my 2.2 install in VR w/ MSAA off PD at 1.5 vs 2.5 w/ MSAA Off and PD at 1.5 just to verify if PD is also taking a hit.

Edited May 17, 2018 by SkateZilla

[Pikey]

Only talking from experience, devs don't give reasoning for anything here. Occasionally the PM says what path it's taking but there's no discussion and it's high level. ED doesn't work in the same manner as open developers, whether we like it or not. However I fundamentally disagree that the choice of upgrading a technology, even if it's not quite tuned, is a bad one or affects the majority as opposed to the status quo. Last 2.4 play I had they both had equal faults, the terrain looked terrible and I had stutters and invisible tracer fire. The night lighting is being worked on and it's been said before, it's being worked on on Hormuz first, which makes sense as the map they are still creating and spending the time in.

Well as 2.5.1 has singled out a rendering method that negatively affects both performance and image detail amongst the other issues I started my post with, (and I have read a fair bit about the science behind DS before posting) then the issue is 2.5.1 and why I asked the question if the reasoning to use DS has been given by ED anywhere?

[Charlie-1]

I'm not sure about 2.4, I was comparing both 1.58 and 2.5 with DS off compared to 2.5.1 with DS forced on.

 

Last 2.4 play I had they both had equal faults.

[Schmidtfire]

I have no clue why you want a technology that does not support AA in a game that heavily relies on intricate details in cockpits aswell as VR compatibility. On top of that a very detailed landscape and big cituies, power lines etc. Light interactions are important, but this decision seems driven by creating great looking screenshots of models and not really made for optimized gameplay.

 

Technology must continue to develop and I am really lucky that I can play DCS with high settings, I feel that I bruteforce my way through the engine, as my computer - I7, 1080ti, 32GB RAM, SSD runs good but are strained at times and with loading times trough the roof.

 

While I do agree it looks really good (Im currently in VR, MSAA and 1.5PD ), the tradeoff of loosing a big chunk of the playerbase aswell as more or less killing the little multiplayer that already exists, is a really high stakes game.

 

Deffered Shading might be the way to go in the long-run, but Im wondering if ED didn't pull the trigger a bit early on this one?

[Nate--IRL--]

I have no clue why you want a technology that does not support AA in a game that heavily relies on intricate details in cockpits aswell as VR compatibility. On top of that a very detailed landscape and big cituies, power lines etc.

 

Couldn't agree more, Jaggies and Shimmering textures in the distance have ruined the look of DCS for me. Seriously considering a 1080ti (or one of the upcoming cards) just to run 2x MSAA at acceptable speeds. :(

 

Nate

[Whisper]

Pixel Density is Essentially Super Sampling.

 

I'll prolly check performance in my 2.2 install in VR w/ MSAA off PD at 1.5 vs 2.5 w/ MSAA Off and PD at 1.5 just to verify if PD is also taking a hit.

 

Thanks a lot

[Charlie-1]

From what I can tell DS is early experimental technology in this game and the real workable solution with purpose built models and Shader AA are theoretical and a long way off being practical.

 

It's a shame effort is going that way as IMO it doesn't look much better than running 2.5 with DS off and HDR on, both on a monitor and in VR except the fluorescent trees! If the trees could be improved without DS then why bother?

[Sn8ke]

** Not an Official ED Response **

(But I'll explain it for you without getting too technical)

 

PBR and Deferred Lighting and Deferred Shading are all Separate Things,

 

However, PBR Relies on them to render Accurately, Thus PBR Requires Essentially Deferred Lighting and Shading to be Enabled.

Especially considering all light sources are Dynamic and there are hundreds of thousands of them across the terrains and hudreds in cockpits, and they must all interact w/ the Textures.

 

So, We'll Start w/ Why PBR?:

w/ 1.1.1.1 -> 1.2.16 All Models were rendered Using the Old Rendering which required 7 Surface Properties:

Diffuse: From Diffuse Texture

Specular Color: From Diffuse Texture Merged w/ Specular Texture

Specular Power: From Specular Texture (RED Channel)

Specular Glossiness: From Specular Texture (GREEN Channel)

Reflection Level: From Specular Texture (BLUE Channel)

Reflection Blur: Globally Set in 3DS Max by Material

Ambient Occlusion: Baked onto Diffuse Layer (Unrealistic)

 

w/ 1.5.0/2.0.0 and Deferred Shading / PBR Enabled.

Albedo: from Diffuse Texture

Microsurface (ED Calls it Roughness): from RoughMet Texture (GREEN Channel)

Reflectivity (ED Calls it Metallicness): From RoughMet Texture (BLUE Channel)

Ambient Occlusion: From RoughMet Texture (RED Channel).

 

So, You go from 7 Variables to Control how a Surface Looks to 4 (Counting A.O., But not all models use it).

 

Albedo is the Diffuse and Specular Colors of the Surface.

 

Roughness as ED Calls it, Taking the Place of Reflection Blur (Which is now Dynamic by Pixel instead of Material), Specular Glossiness, Specular Power, Specular

Controls how Rough or Smooth a Surface is, Smoother Surface will scatter light less, and reflect light and surrounding clearly vs not. (Smooth will appear polished/glossy, Rough will appear Matte)

 

Metallicness as ED Calls it, Takes the Place of Reflection Level

Controls how Mirror like a Surface is, Metallic Surface will have a Clear Vibrant Reflection of the Environment and Non Metals will have a faded Reflection of the Environment.

 

Ambient Occlusion, is not a Separate Map and not Baked onto the Diffuse,

Under the Deferred Lighting Engine, the A.O. Map is only drawn when in shadow (ie it will not be drawn on the model if exposed to direct Sunlight/Light Source)

 

 

In Practice Roughness is being used Normally by ED (Controls the Surface and How it scatters light etc)

 

Metallicness however in Practice is Normally a 2 Color Option, White / Black, Metal or Not. Metal Items will have Environment Reflections, Non-Metals wont.

Since ED is using the Metallicness Map as a Gradient (All 256 values), I would Call it by it's normal PBR name, Reflectivity Map. as the Gradient 0-255 controls how Apparent the Reflection is on a smooth surface.

(Examples, RoughMets being used on Canopy Glass w/ 100% Metallic Value, Gradient is being used as a Reflection Value and not Metallic Value, You cant have half metal, A substance is either metal or dieletric)

 

 

PERFORMANCE:

As I Stated in the opening blurb, due to the type of environments we have in DCS and the introduction of PBR, Moving to 2-Pass Shading (Deferring Shading) and 2-Pass Lighting (Deferring Lighting) was needed to use PBR.

 

The Major Improvement is how Dynamic Lights all appear, and how objects render around the dynamic light sources w/ the PBR.

 

The Draw Back w/ Deferred Lighting/Shading has always been Performance w/ Brute force (Spatial) Anti-Aliasing (MSAA, SSAA, EQAA and CSAA Etc), which is why the DCSW2.5 MSAA Limit is 4x instead of 16xQ, etc.

ED is aware of the Performance issues w/ Spatial AA, so hopefully something emerges on that front.

 

For now the obvious work around is to move to PostProcess / Shader AA, (FXAA, MLAA, SMAA etc), which will separate the Anti-Aliasing Process from the Rendering / Shading and Lighting Process.

 

This is why in DCS 2.5 when Deferred is Enabled that Your FPS and Performance are lower than 1.5.0/2.0.x w/ Deferred Disabled.

 

Shader Based AA and High Resolution Screens are making Traditional Spatial AA Obsolete, soon every PC Screen will be 4K Regardless. just as 1080p became the standard after 1024x768.

 

APPEARANCE OF OLD ASSETS:

As Anyone can tell, alot of the Pre-PBR Models look overly shiney/glossy at angles, this is because they have not been Converted to native PBR, instead they use a Converter that converts the Specular File Channels for use w/ PBR. the Values are stored in a JSON File.

Eventually They'll all be replaced w/ Generated RoughMet (Microsurface/Reflectivity) Files.

 

For the Most part, most of the models look presentable in the grand scheme of things w/ the JSON FIles, most of the "Glossy" and Odd Appearances happen on Liveries that use Custom/Unique Specular Files, that do not convert to PBR Very well when using the Conversion Parameter Values in the JSON File.

 

 

BIGHT / DARK VARIANCE:

Outside of the already stated Lighting Adjustments (Night Lighting/Sources etc), Moving to PBR and Rendering Objects more Realistically w/ Realistic Dynamic Lighting also brings with it to changes in how Cockpits Appear.

 

Even When Calibrated w/ simple calibration tools, everyone's Displays are not the same, there are differences in Back Light Brightness, and Contrast at angles etc etc, so , the Same Cockpit might appear DARK for one user and BRIGHT for another, this is the reason for the Gamma Slider.

 

Night Lighting and other things are being tweaked to Fix the Lighting as well as help bring a huge universal sweet spot, so that Cockpits and the Environment would look the same for all users.

 

Thanks for your response!

[Dav IRL]

I think it is fair to say that the trade off is not worth the hassle in terms of both performance and visual. There is not enough of an improvement to warrant the extra demand. The release of the hornet and straits of hormuz will be on hold until the engine takes a smoother step in the right direction, its been a headache this deferred shading business, badly timed with VR introduction and while all modules are pressing ahead full steam we can't see the correct lighting on the switches in the KA50.

[Mars Exulte]

Just like we're in a transitional phase with models before (for example FC3 aircraft being progessively updated 3d models from a few tens of thousands to a few hundred thousand polys) we're in a transition phase for the new render engine. Newer modules look and work better with the new engine because they were made for it. Older ones are going to be gradually updated to the new standard. It is unfortunately not something that you can write a batch file for and let cook over night.

 

 

I also don't see how some of you don't think there's an improvement. A demonstration video in the Harrier in NTTR I made for some friends at certain angles looked like real life. As opposed to 1.5 which looked like a very, VERY dated simulator, even with the texture mods I was running (which helped a lot, but still).

 

 

As for "other games" doing photorealistic rendering "better" with "less performance hit"...

 

 

Are any of them modeling professional level flight simulations? No? Then that's a lot of extra resources they can direct to "just" graphics, then, huh? Go to your local aviation museum, look up into the chassis on a fighter jet display, and note the massive tangles of wires, cables, hydraulic lines, etc. Think about radar, sensors, all those electrical systems. Think about creating a virtual wind tunnel. Hundreds and hundreds of things being simulated, updated, tracked, etc. Now, think about those things all interacting with a couple dozen other things all doing the same thing, and online to boot.

 

 

When you stop and actually THINK about what they're doing here, it's surprising you don't need a Xenon cooled server rack to do it, it all works on a little 2ft x 2ft x 1ft box in your living room. Are you SERIOUSLY going to compare that to Witcher 3, Battlefield, *insert random game* that does VIRTUALLY NOTHING by comparison and say "Well, THIS has better graphics and runs at 90fps"... No shit, Gertrude.

[Charlie-1]

I reckon the transition is likely to be a year or more to port over existing assets and complete?

 

Was the video from the cockpit or outside? I agree DS can make the outside world look amazing, especially in Nevada at sunset, but making a video to show your friends is not quite the same as trying to read gauges and displays in VR in battle, which 2.5.1 makes considerably less enjoyable. Did you record from a replay btw, as that's broken and and should be on the priority list for fixing.

 

DCS 2.5 with HDR and lens flares on things look amazing from the air and even sitting on the runway there's not much difference, except the damn trees, which now look rediculous with DS off because they only have shading on the leaves with DS on. Nevada looks more pink at sunset than 2.5.1, but I think a gamma slider with 2.5 DS off would even that out.

 

I think photoreal is nice to aim for, but not at the expense of performance and practicality. Finding the compromise is subjective, but IMO, 2.5 is not bad at all.

 

Other examples here..

 

https://forums.eagle.ru/attachment.php?attachmentid=183566&d=1525022277

 

https://forums.eagle.ru/attachment.php?attachmentid=183567&d=1525022901

Edited May 18, 2018 by NAKE350

[BitMaster]

The big question is, will this transition phase be over some time or do we have to live with software that develops so fast and wide that it will likely stay in such a state for some time to come ?

 

As each new map and module has it own troubles, it is not hard to believe that there will be tons of things to fix with th SoH and FA18 release. Not to mention all the others in the pipe.

 

I have answered this for myself and do believe it will always be in a phase you would not call "overall stable and usable all around", like RTM in old days.

 

My wish was a "fix before adding new content" after SoH and FA18, but then there is Vulkan ....

 

...sigh, back to my earlier believe, we are in a circle of constant development and have to deal with it, if we like it or not.

 

I also do believe it is hard to do it any other way than how they actually do it.

 

Wishful thinking and daily business are 2 different things

 

 

Bit

[David OC]

The big question is, will this transition phase be over some time or do we have to live with software that develops so fast and wide that it will likely stay in such a state for some time to come ?

 

As each new map and module has it own troubles, it is not hard to believe that there will be tons of things to fix with th SoH and FA18 release. Not to mention all the others in the pipe.

 

I have answered this for myself and do believe it will always be in a phase you would not call "overall stable and usable all around", like RTM in old days.

 

My wish was a "fix before adding new content" after SoH and FA18, but then there is Vulkan ....

 

...sigh, back to my earlier believe, we are in a circle of constant development and have to deal with it, if we like it or not.

 

I also do believe it is hard to do it any other way than how they actually do it.

 

Wishful thinking and daily business are 2 different things

 

Bit

 

I do agree with you on this, the problem is the future tech going forward and which way to go today?

 

They made the right call? Well I hope they have with Vulkan and DS. They do need some long term direction especially with VR technology over the next 5 years moving forward. They need to plan the best they can today for this technology. Plus the different teams needing something to move forward with and having it all come together nicely.

 

I also wish the two 2.5's were more apart in development, tho it has only been 3 and half months and I'm sure ED is the first one that wants the 2.5 locked down more with DS, optimization.

 

This level of programing must be loads of fun...

 

Edited May 18, 2018 by David OC

[Flighter]

Nice post SkateZilla, thanks for your time.

 

I have one question :

 

..... Thus PBR Requires Essentially Deferred Lighting and Shading to be Enabled....

 

Do you say that the Deferred Shading was the ONLY way to go for using the PBR material model ?

 

If so, I strongly disagree.

Agreed, the DS method is "less compromising", one small step closer to ray tracing.

The alternative is a less accurate approach (for sane, short rendering times) but with significantly better performance IF WELL IMPLEMENTED/BALANCED.

The produced PERCEIVED quality with present top notch rendering hardware is

- for SCREEN projection identical for both methods.

- for VR HMD projection the DS method is a killer no good rascal that eats away all your candy and you are left crying.

Emphasis on the word PERCEIVED.

 

As I am sure you are well aware, 3d rendering, starting from prefab pseudo 3d overlays, all the way to full blown ray tracing, reflection, refraction, caustics with infinite bounces and sugar on top, is a series of balancing between complexity/speed of render and optical precision/image quality.

If we bring psycho optics in the equation we get very, very interesting results considering that :

- the average person can hardly identify, let alone verify the effects of more than 10 point lights in a room (see cockpit).

- a dynamically changing scene makes the accuracy of the results 99% impossible to verify in real time.

- the irrelevance of the accuracy is multiplied if the lighting/shading itself is not the main subject.

- Accuracy of ambient lighting is even harder to grasp (next to impossible). Human brain is easily pleased by very small tricks.

 

Our brains only need to be convinced about a scene, they will never analyze it for geometrical precision. This is what DS brings us, do we need it ? And at what price ?

How many fps and how much more shimmering is worth paying for the shadow and the lighting of the 30th little indicator that gets lit under my arm ? Is it going to make a perceivable difference ? Do I care if it is mathematically and analytical geometrically correct ?

 

The way ED took, I think, is not the realistically wise but the theoretically impressive. At least for people with VR setups, this is major step backward.

I believe that the one phrase that needs more focus is "Perceived Quality" as it holds many solutions that maybe were left unexplored.

 

At the end of the day perfection is the use of the best compromises under given circumstances.

 

Cheers

Edited May 21, 2018 by Flighter
typos

[Charlie-1]

 

 

BIGHT / DARK VARIANCE:

Outside of the already stated Lighting Adjustments (Night Lighting/Sources etc), Moving to PBR and Rendering Objects more Realistically w/ Realistic Dynamic Lighting also brings with it to changes in how Cockpits Appear.

 

Even When Calibrated w/ simple calibration tools, everyone's Displays are not the same, there are differences in Back Light Brightness, and Contrast at angles etc etc, so , the Same Cockpit might appear DARK for one user and BRIGHT for another, this is the reason for the Gamma Slider.

 

Night Lighting and other things are being tweaked to Fix the Lighting as well as help bring a huge universal sweet spot, so that Cockpits and the Environment would look the same for all users.

 

Forgive me if this has been stated before and it sounds crazy, but is it possible to use traditional rendering in cockpit panels where detail and sharpness of the image really matter, (especially in VR), but use DS for rendering the outside world? This would potentially solve several of the major issues we're experiencing. WIN WIN!