Jump to content

Dynamic Foveated Rendering - Everything in one page


mbucchia

Recommended Posts

Yes these issues are only DCS, and as mentioned only ED can fix them.

You can try Pavlov VR if you have it on Steam, it works perfectly and it's literally impossible to see the focus area's edges.

  • Thanks 2

This account is now inactive and not monitored.

Link to comment
Share on other sites

7 hours ago, markturner1960 said:

Going to re post these questions from the original DFR thread as it seems to have got lost down on page 2: 

Can I just clarify what the recommended ( from MBucchia ideally as I guess he would know!) way to go about supersampling is? Above, someone mentioned setting everything at one, and only ramping up the resolution in the QVFR app.....seems sensible from a theory point of view? . Currently, I am running 1.3 in the occulus app and some increased supersampling in the QVFR app only, none anywhere else......Guess I can try it and see but just wondered on the tech knowledge .

In the same theme, same applies for changing the tangent multiplyer in ODT....can this be handled with the same effect by QVFR or not?

Another query...I also see mention of people making a profile in OTT tool and setting the monitor resolution to as low as possible in the profile.......Is there "Official" word on wether this is recommended? As I also read that using OTT is not a great idea as it is an old, unsupported app now.....

 

Its so confusing ! 

It doesn't have to be confusing.

image.png

QVFR will always _output_ to the resolution dictated by the platform (accounting for the focus region supersampling if needed), in this case whatever you have set in Oculus app. You can always check the log file for QVFR, it will tell you exactly what is happening:

2023-09-21 16:06:03 -0700: Recommended peripheral resolution: 720x792 (0.500x density)
2023-09-21 16:06:03 -0700: Recommended focus resolution: 720x712 (1.000x density)
2023-09-21 16:06:03 -0700:   Stereo pixel count was: 4561920 (1440x1584)
2023-09-21 16:06:03 -0700:   Quad views pixel count is: 2165760
2023-09-21 16:06:03 -0700:   Savings: -52.5%

(sorry these numbers are ridiculously low but you get the point)

About FOV tangent, I can't say for sure whether that works, but you also probably don't need it. QVFR makes your peripheral region low-enough that FOV tangent shouldn't help that much.

Monitor resolution shouldn't matter very much unless you have a very old system and a GPU with very low VRAM. I assume your 3090 has 12 GB so it's probably fine to not worry about it.

 


Edited by mbucchia
  • Thanks 1

This account is now inactive and not monitored.

Link to comment
Share on other sites

6 hours ago, mbucchia said:

Yes these issues are only DCS, and as mentioned only ED can fix them.

You can try Pavlov VR if you have it on Steam, it works perfectly and it's literally impossible to see the focus area's edges.

I have posted a request for this to be investigated in the wish list forum.  I suggest others do the same or bump the post to high heaven.

System spec: Intel i9 13900KF @ stock,  Gigabyte GeForce RTX 4090 Gaming OC 24GB GDDR6X, Gigabyte Z690 UD DDR4, Corsair Vengeance RGB PRO SL 32GB (2 x 16GB) DDR4 3600MHz C18, Samsung 980 EVO 500 GB NVME M.2 SSD (system drive), Samsung 970 EVO 1 TB NVME M.2 SSD (games drive), Cooler Master ML360 Illusion CPU Cooler, Asus XG43UQ Monitor, Oculus Quest Pro, Thrustmaster Warthog HOTAS, MFG Crosswind Rudder Pedals

Link to comment
Share on other sites

28 minutes ago, Hippo said:

I have posted a request for this to be investigated in the wish list forum.  I suggest others do the same or bump the post to high heaven.

It would probably help if you could take screenshots (OpenXR Toolkit can capture this effect). Something like looking directly at the sun makes it obvious IIRC?

  • Thanks 3

This account is now inactive and not monitored.

Link to comment
Share on other sites

On 10/21/2023 at 7:01 AM, Hippo said:

 

EDIT Again:  Almost certainly down to DLAA.  Shimmers in the periphery eliminated vs MSAAx4.  Needs a mission restart (at least), i.e. changing the setting in-mission doesn't appear to do it.   In normal usage (i.e. not using DFR) I can discern no difference in quality or performance between MSAAx4 and DLAA.

 

 

Wait, I'm confused. Is the shimmering better or worse with MSAA? 

"...I just wanna fly; put your arms around me baby, put your arms around me baby" - Sugar Ray

RTX 3090, Ryzen 7 5800X3D, MSI MPG B550 Gaming mobo, 64 GB DDR4 RAM, 970 EVO Plus NVMe M.2 SSD 2TB game install drive, Oculus Quest Pro via link cable, Standalone DCS beta.

Link to comment
Share on other sites

On 9/26/2023 at 11:20 AM, recoilfx said:

FPS gain with QVFR drops exponentially when scenery detail goes up. @mbucchia says it's a CPU issue due to how DCS renders. I have a 4090/5800x3d build, and I choose to run without QVFR because any mission with details runs worse on QVFR.

Hmm, this is the first time I've heard this. @mbucchia, are you able to confirm? 

"...I just wanna fly; put your arms around me baby, put your arms around me baby" - Sugar Ray

RTX 3090, Ryzen 7 5800X3D, MSI MPG B550 Gaming mobo, 64 GB DDR4 RAM, 970 EVO Plus NVMe M.2 SSD 2TB game install drive, Oculus Quest Pro via link cable, Standalone DCS beta.

Link to comment
Share on other sites

3 hours ago, FupDuck said:

Hmm, this is the first time I've heard this. @mbucchia, are you able to confirm? 

yes @recoilfx is correct. With quad views you save a lot of on your GPU, but it requires more CPU to submit the 2 extra views. This can be optimized in the game engine, but not trivial (and it doesn't look like DCS is equipped to do this optimization).


Edited by mbucchia

This account is now inactive and not monitored.

Link to comment
Share on other sites

3 hours ago, FupDuck said:

Wait, I'm confused. Is the shimmering better or worse with MSAA? 

IMHO - the shimmering is worse with MSAA.  I was using MSAAx4 + QVFR and was aware of shimmering in the peripheral area, which goes away entirely if I use DLAA.  I found the shimmering too distracting with MSAAx4 so was using a larger focus area to migitate this somewhat, leading to only a modest performance improvement.  With DLAA I have reduced the focus region to 0.25, a reduction in pixel count of ~75%, and a very significant performance improvement.

As to that last sentence.  Ahem.  I wrote that after very limited testing.  I would say that the DLAA visuals aren't quite as crisp, and there is artifacting/ smearing with certain fast motion (e.g. aircraft flybys can leave trails).  Still, I am so far very pleased with the performance / quality I am getting with DLAA + QVFR and it is what I am using.

System spec: Intel i9 13900KF @ stock,  Gigabyte GeForce RTX 4090 Gaming OC 24GB GDDR6X, Gigabyte Z690 UD DDR4, Corsair Vengeance RGB PRO SL 32GB (2 x 16GB) DDR4 3600MHz C18, Samsung 980 EVO 500 GB NVME M.2 SSD (system drive), Samsung 970 EVO 1 TB NVME M.2 SSD (games drive), Cooler Master ML360 Illusion CPU Cooler, Asus XG43UQ Monitor, Oculus Quest Pro, Thrustmaster Warthog HOTAS, MFG Crosswind Rudder Pedals

Link to comment
Share on other sites

15 hours ago, FupDuck said:

Hmm, this is the first time I've heard this. @mbucchia, are you able to confirm? 

Thank you for bringing up @recoilfx 's (and recoilfx for making it) comment.  I did some more (quick and limited) testing and perhaps the increase in performance from QVFR wasn't as great as I first thought.  I tested with DLAA + QVFR, DLAA only, and MSAAx4 only, with these settings and mission (up to end of first turn) and see the results below.

Only with QVFR can I maintain 72 fps throughout.  The GPU util is visibly lower with QVFR, but not by as much as I was expecting.  I can't hit 72 fps at all with MSAAx4.

CPU core 28 (and perhaps 11 and 12) is visibly busier with QVFR.  Core 9 or 10 is usually the main(? most busy) core.

There is no shimmering with DLAA, whereas it is quite visible and distracting with MSAA, including flickering ground shadows.

@recoilfx I'd be interested if you could point me to a mission and settings where you see worse performance with QVFR vs no QVFR.

 

image.png


Edited by Hippo

System spec: Intel i9 13900KF @ stock,  Gigabyte GeForce RTX 4090 Gaming OC 24GB GDDR6X, Gigabyte Z690 UD DDR4, Corsair Vengeance RGB PRO SL 32GB (2 x 16GB) DDR4 3600MHz C18, Samsung 980 EVO 500 GB NVME M.2 SSD (system drive), Samsung 970 EVO 1 TB NVME M.2 SSD (games drive), Cooler Master ML360 Illusion CPU Cooler, Asus XG43UQ Monitor, Oculus Quest Pro, Thrustmaster Warthog HOTAS, MFG Crosswind Rudder Pedals

Link to comment
Share on other sites

21 hours ago, Hippo said:

IMHO - the shimmering is worse with MSAA.  I was using MSAAx4 + QVFR and was aware of shimmering in the peripheral area, which goes away entirely if I use DLAA.  I found the shimmering too distracting with MSAAx4 so was using a larger focus area to migitate this somewhat, leading to only a modest performance improvement.  With DLAA I have reduced the focus region to 0.25, a reduction in pixel count of ~75%, and a very significant performance improvement.

As to that last sentence.  Ahem.  I wrote that after very limited testing.  I would say that the DLAA visuals aren't quite as crisp, and there is artifacting/ smearing with certain fast motion (e.g. aircraft flybys can leave trails).  Still, I am so far very pleased with the performance / quality I am getting with DLAA + QVFR and it is what I am using.

I've found that running the foveate region in QVFR at >150% means that you can turn off MSAA completely - and yes, the shimmering goes away.  The fps boost that you get from turning off MSAA goes into running the foveate region at higher super-sampling.  However, at least you're only super-sampling a smaller portion of the screen rather than applying MSAA across the whole thing..

System specs:
i9-11900KF OC to 5500Mhz with water block- MSI RTX4090 OC - ASUS ROG HERO VIII - 64GB DDR4 @3600MHz- 1TB SSD OS - 2TB SSD GAME- 1000W PSU
Pimax 8kx - TM Pendular Pedals - TM Warthog Throttle - TM Waterhog HOTAS - TM F/A-18 Grip - WinWing Pheonix MIP/UPC - Homemade Landing Gear control.  Andre's SimShaker driving two subwoofers and seat kickers for seat and control column vibration - sketchpad linked to OpenKneeboard - VoiceAttack with VIACOM Pro

 

 

Link to comment
Share on other sites

1 hour ago, GayCavalier said:

I've found that running the foveate region in QVFR at >150% means that you can turn off MSAA completely - and yes, the shimmering goes away.  The fps boost that you get from turning off MSAA goes into running the foveate region at higher super-sampling.  However, at least you're only super-sampling a smaller portion of the screen rather than applying MSAA across the whole thing..

MSAA, by definition, doesn't supersample the whole picture (that's what SSAA does). Just the edges of objects, where "jaggies" are expected to happen.

Also note there's this setting in the NVidia Control Panel (and I assume there's an AMD equivalent). It will clean up edges of transparent textures, like the trees, and I see virtually no performance hit from using it. (Which I have been for years. This together with MSAA are just as good as SSAA, but much faster.)
 

image.png

Link to comment
Share on other sites

I was trying to check out the other AA options unfortunately my pedals ceased working , new cable will be here tomorrow so hopefully I can resume here soon. At least I am very much hoping it is just the cable. My luck continues LOL. Seems I spend more time trying to get all my hardware working proper than I do anything else.

Don B

EVGA Z390 Dark MB | i9 9900k CPU @ 5.1 GHz | Gigabyte 4090 OC | 64 GB Corsair Vengeance 3200 MHz CL16 | Corsair H150i Pro Cooler |Virpil CM3 Stick w/ Alpha Prime Grip 200mm ext| Virpil CM3 Throttle | VPC Rotor TCS Base w/ Alpha-L Grip| Point Control V2|Varjo Aero|

Link to comment
Share on other sites

On 10/25/2023 at 6:57 AM, Hippo said:

Thank you for bringing up @recoilfx 's (and recoilfx for making it) comment.  I did some more (quick and limited) testing and perhaps the increase in performance from QVFR wasn't as great as I first thought.  I tested with DLAA + QVFR, DLAA only, and MSAAx4 only, with these settings and mission (up to end of first turn) and see the results below.

Only with QVFR can I maintain 72 fps throughout.  The GPU util is visibly lower with QVFR, but not by as much as I was expecting.  I can't hit 72 fps at all with MSAAx4.

CPU core 28 (and perhaps 11 and 12) is visibly busier with QVFR.  Core 9 or 10 is usually the main(? most busy) core.

There is no shimmering with DLAA, whereas it is quite visible and distracting with MSAA, including flickering ground shadows.

@recoilfx I'd be interested if you could point me to a mission and settings where you see worse performance with QVFR vs no QVFR.

 

image.png

 

@HippoThe few benchmark missions I always test are F-16 cold start Nevada instant action, and Huey Mariana Island take off instant acton. You can also create a quick mission to fly over densely populated cities like London over Normandy.

  • Thanks 1
Link to comment
Share on other sites

@mbucchia I've been testing version 1.1.3 with the Pimax Crystal. This enables the use of DFR with eye tracking again with DCS 2.9 however it does have performance problems. Even in the DCS menu the loss of FPS is noticeable.

Here is the Log:

2023-10-27 14:44:24 -0500: PimaxXR - v0.4.4
2023-10-27 14:44:24 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:24 -0500: PVR: 1.25.1
2023-10-27 14:44:24 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:24 -0500: Device is: Pimax Crystal
2023-10-27 14:44:24 -0500: Parallel projection is enabled
2023-10-27 14:44:25 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:25 -0500: PVR: 1.25.1
2023-10-27 14:44:25 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:25 -0500: Requested extension: XR_EXT_eye_gaze_interaction
2023-10-27 14:44:25 -0500: Requested extension: XR_EXT_hand_tracking
2023-10-27 14:44:25 -0500: Requested extension: XR_KHR_visibility_mask
2023-10-27 14:44:25 -0500: Requested extension: XR_KHR_win32_convert_performance_counter_time
2023-10-27 14:44:25 -0500: Device is: Pimax Crystal
2023-10-27 14:44:25 -0500: Parallel projection is enabled
2023-10-27 14:44:25 -0500: Recommended resolution: 4312x5102
2023-10-27 14:44:26 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:26 -0500: PVR: 1.25.1
2023-10-27 14:44:26 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:26 -0500: Device is: Pimax Crystal
2023-10-27 14:44:26 -0500: Parallel projection is enabled
2023-10-27 14:44:27 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:27 -0500: PVR: 1.25.1
2023-10-27 14:44:27 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_D3D11_enable
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_visibility_mask
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_hand_tracking
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_eye_gaze_interaction
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_eye_gaze_interaction
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_hand_tracking
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_visibility_mask
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_win32_convert_performance_counter_time
2023-10-27 14:44:27 -0500: Device is: Pimax Crystal
2023-10-27 14:44:27 -0500: Parallel projection is enabled
2023-10-27 14:44:27 -0500: Recommended resolution: 4312x5102
2023-10-27 14:44:29 -0500: Using D3D11 on adapter: NVIDIA GeForce RTX 4090
2023-10-27 14:44:32 -0500: xrBeginFrame failed with XR_ERROR_CALL_ORDER_INVALID
2023-10-27 14:44:32 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:32 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:33 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:34 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED

...

2023-10-27 14:44:35 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:35 -0500: Maximum number of errors logged. Going silent.

Any Idea of what could be causing the 'xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED' error happen multiple times a second?

Link to comment
Share on other sites

6 minutes ago, f14billy said:

@mbucchia I've been testing version 1.1.3 with the Pimax Crystal. This enables the use of DFR with eye tracking again with DCS 2.9 however it does have performance problems. Even in the DCS menu the loss of FPS is noticeable.

Here is the Log:

2023-10-27 14:44:24 -0500: PimaxXR - v0.4.4
2023-10-27 14:44:24 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:24 -0500: PVR: 1.25.1
2023-10-27 14:44:24 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:24 -0500: Device is: Pimax Crystal
2023-10-27 14:44:24 -0500: Parallel projection is enabled
2023-10-27 14:44:25 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:25 -0500: PVR: 1.25.1
2023-10-27 14:44:25 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:25 -0500: Requested extension: XR_EXT_eye_gaze_interaction
2023-10-27 14:44:25 -0500: Requested extension: XR_EXT_hand_tracking
2023-10-27 14:44:25 -0500: Requested extension: XR_KHR_visibility_mask
2023-10-27 14:44:25 -0500: Requested extension: XR_KHR_win32_convert_performance_counter_time
2023-10-27 14:44:25 -0500: Device is: Pimax Crystal
2023-10-27 14:44:25 -0500: Parallel projection is enabled
2023-10-27 14:44:25 -0500: Recommended resolution: 4312x5102
2023-10-27 14:44:26 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:26 -0500: PVR: 1.25.1
2023-10-27 14:44:26 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:26 -0500: Device is: Pimax Crystal
2023-10-27 14:44:26 -0500: Parallel projection is enabled
2023-10-27 14:44:27 -0500: Pimax Client: 1.16.01.01.15
2023-10-27 14:44:27 -0500: PVR: 1.25.1
2023-10-27 14:44:27 -0500: Application: DCS World; Engine: 
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_D3D11_enable
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_visibility_mask
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_hand_tracking
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_eye_gaze_interaction
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_eye_gaze_interaction
2023-10-27 14:44:27 -0500: Requested extension: XR_EXT_hand_tracking
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_visibility_mask
2023-10-27 14:44:27 -0500: Requested extension: XR_KHR_win32_convert_performance_counter_time
2023-10-27 14:44:27 -0500: Device is: Pimax Crystal
2023-10-27 14:44:27 -0500: Parallel projection is enabled
2023-10-27 14:44:27 -0500: Recommended resolution: 4312x5102
2023-10-27 14:44:29 -0500: Using D3D11 on adapter: NVIDIA GeForce RTX 4090
2023-10-27 14:44:32 -0500: xrBeginFrame failed with XR_ERROR_CALL_ORDER_INVALID
2023-10-27 14:44:32 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:32 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:33 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:34 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED

...

2023-10-27 14:44:35 -0500: xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED
2023-10-27 14:44:35 -0500: Maximum number of errors logged. Going silent.

Any Idea of what could be causing the 'xrGetActionStatePose failed with XR_ERROR_ACTIONSET_NOT_ATTACHED' error happen multiple times a second?

It only happens for the first 100 frames and can be ignored. If you really want to get rid of it, you can enable Hand Controllers in the DCS menu.

Highly doubt this is related to your performance issue (or that QVFR is in general).

This account is now inactive and not monitored.

Link to comment
Share on other sites

On 10/25/2023 at 10:03 PM, GayCavalier said:

I've found that running the foveate region in QVFR at >150% means that you can turn off MSAA completely - and yes, the shimmering goes away.  The fps boost that you get from turning off MSAA goes into running the foveate region at higher super-sampling.  However, at least you're only super-sampling a smaller portion of the screen rather than applying MSAA across the whole thing..

So are you using any anti aliasing at all then? Or just QVFR/

System specs: PC1 :Scan 3XS Ryzen 5900X, 64GB Corsair veng DDR4 3600, EVGA GTX 3090 Win 10, Quest Pro, Samsung Odyssey G9 Neo monitor. Tir5. PC2 ( Helo) Scan 3XS Intel 9900 K, 32 GB Ram, 2080Ti, 50 inch Phillips monitor

 F/A-18C: Rhino FFB base TianHang F16 grip, Winwing MP 1, F-18 throttle, TO & Combat panels, MFG crosswind & DFB Aces  seat :cool:                       

Viper: WinWing MFSSB base with F-16 grip, Winwing F-16 throttle, plus Vipergear ICP. MFG crosswind rudders. 

Helo ( Apache) set up: Virpil collective with AH64D grip, Cyclic : Rhino FFB base & TM F18 grip, MFG crosswind rudders, Total controls AH64 MFD's,  TEDAC Unit. 

 

Link to comment
Share on other sites

@mbucchia Are there any consequences of using nvidia low latency mode and/or VR pre-rendered frames in the control panel with quad view? And should I disable DLAA in DCS? I'm already using link sharpening, it feels weird to use 3 separate sharpening filters stacked like this


Edited by halufpv
Link to comment
Share on other sites

1 hour ago, halufpv said:

@mbucchia Are there any consequences of using nvidia low latency mode and/or VR pre-rendered frames in the control panel with quad view? And should I disable DLAA in DCS? I'm already using link sharpening, it feels weird to use 3 separate sharpening filters stacked like this

 

DLAA is not a Sharpening filter, it’s an anti aliasing technique that smoothens out picture. So in fact it is quite the opposite effect of sharpening I would say. Also Quad views … I would not call that a sharpening filter. It lets you run different resolution in parts of the picture, lower in the part you are not looking (if you have eye tracking) and by that reducing the job for your graphics card.

mbucchia might arrest me on this and have better explanations. But the way I see it you are not using 3 separate sharpening filters.

PC: I9 13900K, Asus ROG Strix GeForce RTX 4090 OC, 32 GB RAM@6000Mhz.

Thrustmaster Warthog Hotas. Virpil Base for Joystick. Thrustmaster TPR Pendular Rudderpedals. Realsimulator FSSB-RL MKII ULTRA base + Realsimulator F16SGRH V2 grip

VR: Pimax Crystal, 8KX, HP Reverb G2, Pico 4, Quest 2. Buttkicker Gamer Pro. Next Level Motion Platform V3.

Link to comment
Share on other sites

5 minutes ago, Peedee said:

DLAA is not a Sharpening filter, it’s an anti aliasing technique that smoothens out picture. So in fact it is quite the opposite effect of sharpening I would say. Also Quad views … I would not call that a sharpening filter. It lets you run different resolution in parts of the picture, lower in the part you are not looking (if you have eye tracking) and by that reducing the job for your graphics card.

mbucchia might arrest me on this and have better explanations. But the way I see it you are not using 3 separate sharpening filters.

Quad View does come with a CAS sharpener as default for the center portion, which I'm very much onboard with as it saves us from sharpening the non-center area. DLAA is not a sharpener, but when using it DCS provides a slider right under the option to adjust sharpness, which I can only assume is yet another CAS layer? 

Link to comment
Share on other sites

12 minutes ago, halufpv said:

Quad View does come with a CAS sharpener as default for the center portion, which I'm very much onboard with as it saves us from sharpening the non-center area. DLAA is not a sharpener, but when using it DCS provides a slider right under the option to adjust sharpness, which I can only assume is yet another CAS layer? 

Yup, you are right. I don’t use DLAA so forgot about the sharpening slider under there. 
 

And I forgot about the sharpening as default in Quad view settings.   

 

  • Like 1

PC: I9 13900K, Asus ROG Strix GeForce RTX 4090 OC, 32 GB RAM@6000Mhz.

Thrustmaster Warthog Hotas. Virpil Base for Joystick. Thrustmaster TPR Pendular Rudderpedals. Realsimulator FSSB-RL MKII ULTRA base + Realsimulator F16SGRH V2 grip

VR: Pimax Crystal, 8KX, HP Reverb G2, Pico 4, Quest 2. Buttkicker Gamer Pro. Next Level Motion Platform V3.

Link to comment
Share on other sites


     For DLAA+QVF to give us less blur, a parameter must be raised in the QVF configuration (focus_multiplier). At x1.6 they give a good relationship between performance and focus, the higher you raise it, the sharper everything looks at the cost of losing lots of FPS. For now I'm sticking with MSAA (the old one) because I observe that when applying these new smoothing techniques, frames are increased but image quality is lost and a series of unwanted artifacts/errors such as ghosting and others are also added. . If you look at the radar cursors, for example on the F16 (my favorite plane), they fade as they move across the screen and if you look at the TGP images, they are not sharp at all, they blur and the objects leave traces, on the other hand in MSAA this does not happen.
    At the moment I have not found a configuration that satisfies me, but if you have it and see everything perfect with the new DCS implementations, I invite you to share your experience here.
Link to comment
Share on other sites

On 10/20/2023 at 5:21 PM, mbucchia said:

I can't use partial logs. Need full logs.

You don't need to uninstall anything. Installing a new version overwrites the previous one.

Just released a new build 1.1.3 that fixes the remaining issues with DCS 2.9.

 

so just to clarify - I am running a crystal and just updated to 2.9 and noticed the fps has dropped (assume DFR is not working now) - so I just need to download and run the quad views

1.1.3 and it should work again? (dont need to uninstall anything or reconfig) just install 1.1.3 over the top of the existing install? (existing settings should remain untouched by the update?)

 

also final questions whats the general advice in terms of enabling DLSS and DLAA together? (before this post I was running DLAA by itself with good results)


Edited by mac22
Link to comment
Share on other sites

Hi, I would appreciate some help in understanding what is happening with my system and performance currently.

Following MBucchia reply to my questions further up the thread, I have made some changes to my settings as follows:

I read a thread in the DCS forums yesterday where someone was saying they had better results using single threaded. So thought why not try. It was not as good, so switched back to MT, emptied my FXO & Meta folders, and retried MT, Results were better, as I expected, so I then tried reducing the rendered area in QVFR, from 35, to 25% using the Quad views companion tool. I also turned Turbo on in that tool.

In Open XR toolkit, when you switch on the "Advanced" FPS counter, it displays 3 frame time outputs , app CPU, Rdr CPU & app GPU. Now, these stay under 1 all the time, top 2 fluctuate around 0.5 - 0.7, app GPU pretty much stays at zero. My frames in instant action are staying at 72 95% of the time, occasionally dropping to 36 ( I am running at 72Hz refresh) . Whenever I have turned on the advanced FPS counter in OXRTK previously those values were generally between 12ms and 25ms as I said. So I really dont understand how I get such a huge performance increase.....just from what I did.

My settings are as follows: in Occulus desktop software, slider full right, it's x 1.3 I think. I am using DLSS in DCS and sharpening at 0.65 on the slider. Nothing in OTT or ODT. ( Although I previously (Pre 2.9) was playing with the tangent FoV settings and locking frames at 36 and forcing ASW on, now I use nothing) I have default settings in OXRTK, except for the FPS counter. I just use the QVFR companion, and those settings are as shown:

settinngs4.png

Of Course, I am very happy ! But I dont really understand how I could get such a difference in performance from previously? Any thoughts appreciated! 

Is there a log file I can generate and attach ? How is this done?


Edited by markturner1960

System specs: PC1 :Scan 3XS Ryzen 5900X, 64GB Corsair veng DDR4 3600, EVGA GTX 3090 Win 10, Quest Pro, Samsung Odyssey G9 Neo monitor. Tir5. PC2 ( Helo) Scan 3XS Intel 9900 K, 32 GB Ram, 2080Ti, 50 inch Phillips monitor

 F/A-18C: Rhino FFB base TianHang F16 grip, Winwing MP 1, F-18 throttle, TO & Combat panels, MFG crosswind & DFB Aces  seat :cool:                       

Viper: WinWing MFSSB base with F-16 grip, Winwing F-16 throttle, plus Vipergear ICP. MFG crosswind rudders. 

Helo ( Apache) set up: Virpil collective with AH64D grip, Cyclic : Rhino FFB base & TM F18 grip, MFG crosswind rudders, Total controls AH64 MFD's,  TEDAC Unit. 

 

Link to comment
Share on other sites

3 hours ago, Silver_ said:

     For DLAA+QVF to give us less blur, a parameter must be raised in the QVF configuration (focus_multiplier). At x1.6 they give a good relationship between performance and focus, the higher you raise it, the sharper everything looks at the cost of losing lots of FPS. For now I'm sticking with MSAA (the old one) because I observe that when applying these new smoothing techniques, frames are increased but image quality is lost and a series of unwanted artifacts/errors such as ghosting and others are also added. . If you look at the radar cursors, for example on the F16 (my favorite plane), they fade as they move across the screen and if you look at the TGP images, they are not sharp at all, they blur and the objects leave traces, on the other hand in MSAA this does not happen.
    At the moment I have not found a configuration that satisfies me, but if you have it and see everything perfect with the new DCS implementations, I invite you to share your experience here.

Try setting pixel density to 1.4 or 1.5 and then default settings in QVFR. This gives a good balance on performance and quality with DLSS. The cockpit is sharp for me and landscapes look good. I do however notice differences between maps with regards DLSS artifacts. What CPU and GPU are you using and what headset? 

5800x3drtx407064Gb 3200: 1Tb NVME: Pico 4: Rift S: Quest Pro

Link to comment
Share on other sites

2 hours ago, mac22 said:

so just to clarify - I am running a crystal and just updated to 2.9 and noticed the fps has dropped (assume DFR is not working now) - so I just need to download and run the quad views

1.1.3 and it should work again? (dont need to uninstall anything or reconfig) just install 1.1.3 over the top of the existing install? (existing settings should remain untouched by the update?)

 

also final questions whats the general advice in terms of enabling DLSS and DLAA together? (before this post I was running DLAA by itself with good results)

 

Yes 1.13 will fix it and you can just run the installer and it will replace it. Your settings should remain.

DLSS and DLAA can't be used together. You chose one or the other. I prefer DLAA because I have the pimax crystal for it's visual clarity and DLSS reduces the quality significantly.

Also make sure you set sharpening around 0.6.


Edited by f14billy
Link to comment
Share on other sites

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...