Lack of effective AA option in VR

[FoxTwo]

9 hours ago, St4rgun said:

Another solution would be VRSS

 

Unfortunately not. Not just because of vendor specific technology, but because it requires DX11 (ok, currently, but moving to vulkan) and forward rendering (nope).

[St4rgun]

It would be so nice to have some hints about the development status of the Vulkan engine rewrite.

Nowadays the newsletters contain detailed info during the development of the modules (like the AH-64D), showing WIP screenshots and sharing useful and interesting descriptions.

 

But the engine rewrite / improvement process is a big black hole. It is always at the "it will be ready, whet it will be done", and the "if will have any new information to share, then we will" state.

The very futuristic approach could be some progress indicator bar with 100% meaning the introduction of the new Vulkan engine with all the new features, so we could see that currently the process is at 10%, 50% or 90%. Or maybe 1%, but it's unlikely because earlier they mentioned Q3 2021 for the estimated deadline.

I can't imagine that the dev team working without milestones, it would be interesting to read about these as well.

[rubbra]

The problem with your desire for status updates is that an overhaul of a code base/Significant library and architecture change is a very complex beast and the various steps along the way will be largely meaningless to anyone who doesn’t have access and knowledge of the DCS codebase.

It’s rather different with a new plane - get it flying, and add some functionality (trivialising it, I know, but hopefully you get what I mean). That can be translated into meaningful progress reports that we all understand.

We don’t expect Microsoft to tell us how their progress towards their next OS is going until it’s nearly ready, do we? That’s kinda what we’re talking about here - the operating system underneath all the modules.


Sent from my iPhone using Tapatalk

[St4rgun]

I disagree, it is possible to have some meaningful milestones which are understandable for the general public (i.e. lighting is tested and working, AA technology under testing etc.), or for at least for those players who takes care about the underlying engine's performance / feature list.

It's frustrating that at the future sometime we will get the word "it is done, 3.0.0 is here" or something, but without any previous hints about how far we are from that point at least approximately.

[eaglecash867]

Vulkan implementation has been talked about on these forums for years like its the cure for cancer or something.  There's a possibility that ED may be finding that, at least for DCS, it isn't the magic pill some have been hanging their hopes on.  If they were seeing encouraging results, they would probably tell us.  On the other hand, I think we've all seen examples of how ED being transparent with "we're looking into it" statements have often been taken as promises by some who are then very vocal about "YOU PROMISED!".  So, its easy to understand why they might not want to say anything until they're sure they can deliver.

 

I'm running 4X MSAA on mine, and although its not perfect, its MUCH better than running none.

[St4rgun]

There were high hopes in Vulkan implementation as a cure for performance issues in DCS here in the forums for years indeed.

But I think about it as a great possibility to refactor parts of the rendering engine (which I'm pretty sure done before several times in the background, at least when 2.5 then 2.7 came out).

 

As ED wrote about the aliasing of the distant clouds with 2.7's new cloud system as "known bug" which they are trying to eliminate, but did not done yet gave me the hope that they are rebuilding some parts of the rendering core completely while implementing Vulkan to handle the AA issue once for all.

 

Right now the last official words from ED is a planned deadline of Q3 2021 for Vulkan to introduce, but without any further details. That's why I was curious about some of the PLANNED new functionalities, but I know we won't get any details unfortunately.

Edited September 6, 2021 by St4rgun

[St4rgun]

Looking at the new Mosquito trailer I became excited but very sad at the same time.

If only I could see this quality in my G2...

It actually seems a rendered high quality animation instead an in-game footage compared to the VR picture quality of DCS right now.

 

Not even a small trace of shimmering, everything is fluid and perfect. Simply breathtaking.

 

ED, at the very moment when your graphic engine will be able to deliver THIS quality in VR, then the battle is over with "other simulators" once for all.

Please do your best to make this dream come true!

Thank you.

[FupDuck]

On 8/29/2021 at 3:38 PM, Enduro14 said:

Roll your head left and right, you will see jaggies disappear and reappear.  Something odd about this graphics engine.

I've noticed this as well. Tilt your head about 45° to either side, and all the shimmering and jaggies disappear. I wonder if this fact could point towards a solution.

[FoxTwo]

Back to the topic at hand.

 

Something to consider is that FSR at render scales higher than 1 are effectively an improved scaling solution and may be an effective anti aliasing technique with minimal overhead.

[St4rgun]

On 9/24/2021 at 4:31 AM, Sandmonkee said:

I've noticed this as well. Tilt your head about 45° to either side, and all the shimmering and jaggies disappear. I wonder if this fact could point towards a solution.

Unfortunately I still can't confirm this observation.

 

With my G2 and MSAA 2x I tried every possible roll angle with my head left and right with small increments but nothing happens with the shimmering.

Of course when the majority of the horizontal/vertical lines became diagonal at around 45° then the pixelation (jaggieness) of the lines will be smaller due to pixel arrangement, but that's natural phenomenon of the aliasing effect.

 

What is interesting that in the world not EVERYTHING is aliasing (shimmering), like the buildings (which are part of the map), roads, cars on the roads, rivers, shoreline, etc are FINE without shimmering.

But every object which seems as transparent textures like electricity pylons, lighting poles, fences and distant objects like airplanes on the ground or objects/buildings which are NOT part of the original map, but placed in the editor (like the subbase complex in the Oilfield campaign of the Mi8 module) shimmering badly from the distance.

 

At least for me the new clouds became MUCH better since introducing them in 2.7, but there's still some aliasing at the distant clouds at the horizon and every objects aliased which are seen BEFORE the cloud layer. But for me the heavy shimmering and jittering is gone (at least with some presets).

 

I observed that the transparent AA settings in nVidia panel CAN reduce shimmering but only for some objects: the contours of the trees and the painted yellow lines at the airports are NOT shimmering with that setting.

 

Q3 is near its end, so I'm eagerly waiting for any news about Vulkan as always... 

Edited September 25, 2021 by St4rgun

[St4rgun]

Dear ED devs, the AA in VR is still pretty much half functioning, for example in the current OB the border lines of any water bodies are shimmering and aliasing badly at distance despite of the MSAA 2x setting. Is there a hope that something can be done with the AA issues in this graphic engine, or it won't be touched until the Vulkan rewrite?

[edmuss]

I don't see that much shimmer, fences and narrow roads are worst at certain angles but on the whole the AA works. If you have an Nvidia GPU then enable MFAA in the control panel to double up the effect of the MSAA without any performance hit.

I'd say that the AA in vr works just fine, there is noticeable reduction in the aliasing with each setting. The shimmering is reduced with increased MSAA but not entirely eliminated.

[zildac]

I don't see that much shimmer, fences and narrow roads are worst at certain angles but on the whole the AA works. If you have an Nvidia GPU then enable MFAA in the control panel to double up the effect of the MSAA without any performance hit.
I'd say that the AA in vr works just fine, there is noticeable reduction in the aliasing with each setting. The shimmering is reduced with increased MSAA but not entirely eliminated.

I wish I could say the same, even with 2xMSAA, MFAA enabled in nvcpl the shimmer is pretty awful frankly. Jump into the Hornet instant action, ready on the ramp and take a look at the parked Hornet to your right. The undercarriage, and all the model edges are aliased awfully and shimmer. It's made somewhat worse by the lighting due to time of day on fairness, but it's pretty horrific.

[edmuss]

I don't have the hornet middle in afraid so can't compare like for like

Possibly the worst I've seen is the instant action ka50 take off (from the back of the ship) if you look to the adjacent boat all of the handrails and rigging is dancing. That said I've not run that one through shiver running with MFAA enabled.

[St4rgun]

It's obvious that the AA problem is not only present on a selected few's machines in VR, but a lot of users experiencing it. That's why I'm asking if it can be resolved in the current engine, or this is a side effect of deferred shading and WON'T be solved (or not worth any effort) as long as they're working on the new Vulkan engine.

I'm pretty sure, that ED will do their best for the Vulkan rewrite to avoid all these issues, and I say take all the time it needs to eliminate those once for all.

But one outcome simply CAN'T be accepted: if the new graphic engine will still have this AA problem. That is a huge no go for me. So it would be nice to have some words from the devs about this topic...

[Svsmokey]

I know of no VR AA solution in the current engine , irrespective of headset resolution or graphics card maker . I use MSAA×2 , but that only ameliorates the problem under some circumstances , and decreases sharpness . I don't even fly the Syria map i purchased because of the aliasing .

 

[St4rgun]

22 hours ago, edmuss said:

If you have an Nvidia GPU then enable MFAA in the control panel to double up the effect of the MSAA without any performance hit.

I've already tried that, but MFAA has some side effect at least for me: very slight and annoying jittering motion in the headset. It is so slight that it can only be felt and not "seen", but turning it off stabilizes the picture. What helped for me right now is the transparent AA setting (Antialiasing - Transparency / 4x (supersample)). it elimitaes the aliasing of the shorelines.

But the edges of the buildings, the contour of the planes, the edges of somewhat reflective surfaces like the HUD lens etc. are still aliasing badly.

I'm still waiting for any feedback from the devs that they are indeed aware of the AA issues and they have some plan for the solution (or already implemented it). Unfortunately we have very litte info on the whole Vulkan development, so every small feedback is highly appreciated.

[edmuss]

Thinking about it, there could also be a link to physiological tolerance as well.  For example I don't see the 60Hz refresh flicker in the G2 yet for some people it's a instant headache inducing vomitfest™!  Perhaps I'm less perceptive of the AA shimmer?  Applying 2xMSAA reduces it for me, adding MFAA almost eliminates it (barring the cloud/horizon interface jitters).

I did test the KA50 Marianas ship take off last night and for me the MFAA makes a big difference to the adjacent ship although it does still shimmer.  Similarly, low level over Syrian cities there is slight shimmer visible but up high I can barely see it.

[St4rgun]

The right evidence is a screenshot: if somewhere the AA is missing then the captured still image will have badly aliased edges without proper smoothing.

[zildac]

The right evidence is a screenshot: if somewhere the AA is missing then the captured still image will have badly aliased edges without proper smoothing.

Not noticeable on the 2D mirror....

[St4rgun]

Check out the following screenshot. Particularly look at the left edges of the pillars in the back and the missing AA of the fence in the front.

The resolution seems quite low, but despite of that the AA of the small blue lamps is perfect in the background at the left edge of the tarmac. At the same distance the left edge of the pillars are WHITE - how can it be, while the sun shines from the right, so that should be in the shadow?

It seems that the lighting, the transparent textures and the applied deferred shading are fighintg each other to have proper AA at some parts while missing at others.

Maybe the shaders can be adjusted? I have no idea. According to my knowledge right now the hardware full screen MSAA of the GPUs are not used because of the deferred shading. It would be interesting to read some short form of "report" or "white paper" from the devs about this whole AA <-> VR topic.

[Lange_666]

3 hours ago, St4rgun said:

It seems that the lighting, the transparent textures and the applied deferred shading are fighintg each other to have proper AA at some parts while missing at others

Same happens on powerlines. Recently i fly the Huey a lot which also means flying low (going under or just over the powerline wires) and slow which makes it highly visible. The powerline wires appear white at first depending on the approach direction and then turn black as you get closer.
However, i can't remember seeing this same effect a few versions ago, only noticed this recently.

[St4rgun]

Dear @BIGNEWY, sorry to bother again with the engine rewrite, just a quick question.

Is the team working on the applicable AA technology for VR to use promising new technologies instead of the existing MSAA?

Previously there was some discussion about the AA / deferred shading difficulties in VR, not to mention the transparent textures' problem. While we're waiting for Vulkan it would be very relaxing to know, that some fancy AA solution is underway for VR.

Of course we don't need any detailed "classified" technology leak, just a YES or a NO. 

[BIGNEWY]

17 minutes ago, St4rgun said:

Dear @BIGNEWY, sorry to bother again with the engine rewrite, just a quick question.

Is the team working on the applicable AA technology for VR to use promising new technologies instead of the existing MSAA?

Previously there was some discussion about the AA / deferred shading difficulties in VR, not to mention the transparent textures' problem. While we're waiting for Vulkan it would be very relaxing to know, that some fancy AA solution is underway for VR.

Of course we don't need any detailed "classified" technology leak, just a YES or a NO. 

Currently work is focused on multicore and vulkan. 

[St4rgun]

2 hours ago, BIGNEWY said:

Currently work is focused on multicore and vulkan. 

Yes, we know it well, but during the Vulkan implementation the team had the opportunity to change the whole graphic pipeline / technology. In the past in some Q/A session they not even excluded the possibility to change from deferred rendering back to forward rendering (Vulkan equally gives the opportunity for both) because of much better performance of forward rendering in VR and the possibility to apply hw assisted MSAA again (which is unusable in deferred rendering).

As we know the VR performance gain (hopefully) got high priority during the rewrite work, and part of that is the AA implementation. It's really technical details I know, maybe it's much less relevant for people who only want to fly the airplanes, but the overall visual quality in VR highly depends ot these decisions.

That's why I asked if there're such plans as part of the Vulkan implementation.

Edited April 1, 2022 by St4rgun
typo