Depth Perception for DCS Mirrors (VR)

[Camble]

On re-reading your post and my response, I see what you mean. It's true, that VR would require stereographic rendering for the mirror. This doesn't necessarily have to cripple performance if done correctly. The biggest problem with the current implementation is that the mirrors act like a stencil over a single viewport rendered in reverse. So when you move your head, in 2D or VR, the image in all mirrors moves as one, which is completely immersion breaking.

[average_pilot]

To be honest I'd be fine with 2D mirrors as long as the camera is placed dynamically at the correct position. Although that would need some culling of parts that shouldn't be rendered. Maybe that's the difficulty for a better implementation.

[Hiob]

2 hours ago, Camble said:

On re-reading your post and my response, I see what you mean. It's true, that VR would require stereographic rendering for the mirror. This doesn't necessarily have to cripple performance if done correctly. The biggest problem with the current implementation is that the mirrors act like a stencil over a single viewport rendered in reverse. So when you move your head, in 2D or VR, the image in all mirrors moves as one, which is completely immersion breaking.

Yes, but I think that is the problem. If you want to do it properly (I think at least) you would need to render a separate POV for each mirror (twice in VR), no?

There are shortcuts and alleviations like rendering in lower details and lower resolution, but still.

Don't get me wrong - I'm all for an improvement of mirrors! Currently I deactivate them usually because they hit performance so badly and still loook awful most of the time.

I just don't see how to improve them in a meaningful way without absolutely tanking the performance......

_____

A first step would/could be to take the best working module and at least make the mirror behaviour consistent through all modules.

Edited April 11 by Hiob

[Camble]

3 hours ago, average_pilot said:

To be honest I'd be fine with 2D mirrors as long as the camera is placed dynamically at the correct position. Although that would need some culling of parts that shouldn't be rendered. Maybe that's the difficulty for a better implementation.

I honestly wouldn't mind that as a temporary measure, a step in the right direction. But still, this isn't just a VR problem. The fact that the viewpoint doesn't move with head movement is one problem. The other is the stereography, but I'd settle for mirrors that don't spaz out when I move my head. The Mi-24 is perfect extreme example.

40 minutes ago, Hiob said:

Yes, but I think that is the problem. If you want to do it properly (I think at least) you would need to render a separate POV for each mirror (twice in VR), no?

There are shortcuts and alleviations like rendering in lower details and lower resolution, but still.

Don't get me wrong - I'm all for an improvement of mirrors! Currently I deactivate them usually because they hit performance so badly and still loook awful most of the time.

I just don't see how to improve them in a meaningful way without absolutely tanking the performance......

_____

A first step would/could be to take the best working module and at least make the mirror behaviour consistent through all modules.

 

There are plenty games with working mirrors, both flat and VR, without a noticeable performance hit. Current DCS mirrors tank performance, and I'm pretty sure it's because they're implemented using a 20 year old approach. Have you ever played Lock-On? Real-time rendering has come a LONG way since then. Convincing reflections can have a negligible performance hit.

I don't really think there is a "best working module", because they are ALL reliant on the current mirror implementation. I also feel that making excuses for the current implementation does nothing to progress sorely needed improvements to the engine.

Edited April 11 by Camble

[average_pilot]

There is little incentive for a more correct implementation because they look about fine in normal displays.

[Hiob]

10 minutes ago, average_pilot said:

There is little incentive for a more correct implementation because they look about fine in normal displays.

Objection! 

In some Aircraft they look about "acceptable" (F-16 ), but most of the time they work completely unrealistic (perspective) or are LEGO low resolution. Even though I could afford the performance tax on a 4090 I mostly deactivate them because they are just distracting!

[draconus]

3 minutes ago, average_pilot said:

they look about fine in normal displays

No, they don't. Obviously we can not benefit from the added VR implementation on monitor but better angle and camera postion simulation would add a lot for anyone using the mirrors with head tracking. Currently they look like mirrors but they work with their own weird physics.

[average_pilot]

Then that gives me some hope of an eventual new implementation.

[Camble]

19 hours ago, average_pilot said:

To be honest I'd be fine with 2D mirrors as long as the camera is placed dynamically at the correct position. Although that would need some culling of parts that shouldn't be rendered. Maybe that's the difficulty for a better implementation.

See the video I posted here

[zerO_crash]

People tend to forget in their flawed comparisons with "games" out there, that DCS, while still being optimized and brought forth to modern standards, is far beyond in terms of sheer data being worked through. A map in DCS alone, is bigger, than a multitude of those games' copies. A module in DCS, can be more than 1/3rd the size of the other products' total install size.

 

Then there is the part of actual simulation of avionics, systems, weapons and all other underlying components. If you do not understand the difference between a cockpit where gauges react to basic input alas speed/altitude (maybe a few other factors), vs. one where you have complete systems modelled (no magical input entered, instead, the whole chain of sensor - cable - computer - cable - component - cable - pilot interface), targeting devices and digital suites (MFDs which simulate software in software - basically more advanced than a complete product elsewhere), with much more, then you simply do not understand what you bought into. 
 

I won't even mention the fact, that DCS trashes flight models in basically any single one "simulator" out there. If not for DCS, I wouldn't even bother with having a high-performance computer, and I can guarantee you that you won't meet anyone more demanding than me. The comparisons drawn here, compare a military-grade (industrial) simulator adjusted for commercial market, to paper-plane sims. There is virtually no single comparison to be made, at any level. If you have a hard time grasping that, I invite to go into module-specific threads (for simplicity, WWII), and have talk there. You'll quickly find out what depth DCS goes into.

 

Case in point; it is a very real problem of performance. The discussion becomes even more relevant, when VR is drawn into the picture. Whilst requesting new features is good, one ought also to understand that currently, we do not have hardware to run DCS maxed out in VR alone. When more powerfull hardware gets released, then it'll definitely be a valid point. Until then however, just toggling mirrors on/off currently, takes a severe FPS hit. I do not see how creating a more demanding system would be beneficial at the moment.

Edited April 12 by zerO_crash

[SharpeXB]

5 hours ago, average_pilot said:

There is little incentive for a more correct implementation because they look about fine in normal displays.

Oh they could look a lot better in 2D. The performance impact could just be handled with a graphics setting. 

[Camble]

12 hours ago, zerO_crash said:

excuses

Whatever the implementation, we can still retain the option to disable or adjust to align with our hardware spec. It's simply a case of priorities. Latest patch: explosive shockwave effects for foliage. So what? We spend most of the time in the cockpit. The priorities are all out of whack. I keep saying this. The mirrors implementation has been around since Lock-On. An upgraded mirror system might actually improve performance.

Edited April 12 by Camble