Announcement

Collapse
No announcement yet.

index pd vs ss?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    #21
    Originally posted by pngflyer View Post
    Hey Taz,

    What would you recommend for a rift S user? I realise it has a lower resolution than the reverb. I have a PD set at 1.0 in DCS, but I’m considering installing Oculus Tray Tool so I can set the super sampling and maybe get a little more clarity...
    I don't have Rift S anymore and it would really depend on your system and preference. I myself prefer higher SS than anti-aliasing. Others don't.
    Supersampling is type of anti-aliasing afterall and I don't like how MSAA makes everything blurry. But others can't stand the jaggies so you just have to find the setting that works best for you. But Oculus Tray Tool is good.

    When you test these, you may use benchmarking tools to gauge different settings but remember those tools themselves take quite a bit of CPU. 10-15% on my end.

    Comment


      #22
      Cheers, thankyou Taz. Will try this out.

      Comment


        #23
        Originally posted by Taz1004 View Post
        No, that's not how it works. VR compositor doesn't do any super sampling. It's not a render engine. It's not like DCS renders first, SteamVR takes that frame and then super sample it. That would be not only inefficient, also ineffective.

        Simply put, VR compositor "requests" super sampling. DCS will render at the resolution multiplied by the request and its own PD setting. Then compositor will simply downsample to native resolution.



        Also incorrect as Reverb owners will tell you their 100% is the native resolution.
        Goodness...

        1. Reverb owners have to use the Windows Mixed Reality plugin for SteamVR. The plugin handles all the necessary api requests first before Steam VR handles the rest. A reverb user will only see the end result, that's why it shows their native resolution at 100% in Steam VR.
        2. There is always the need to postprocess lens distortion for a flat frame from a game camera for any kind of VR HMD. Therefore you always need more pixels in a rectangular frame that will be lost during this operation because of the lens matched distortion. This is a basic VR process to display a flat picture on a flat panel that is only being watched through convex lesnes.
        3. The PD value in DCS acts as an override in combination with SteamVR. The openVR api asks the game to render a given resolution, but DCS returns a much higher resolution frame if the PD value is higher than 1.0. This is wasting frametime, because it obviously won't fit and needs to be resampled. Letting SteamVR managing resolution and supersampling instead is the much more efficient way. It also has an extra option of an advanced supersample filter under Video Settings. Why don't you test this for yourself? Try to set a PD of 0.5 in DCS and look what happens. The result will be a super blocky low res image, but still supersampled to a high resolution frame with your settings in SteamVR 100% 150% whatsoever...
        The difference might be if DCS is using the oculus api, here the PD value seems to hook directly into it. I can't test this here.

        SteamVR and it's openVR api is a image/pixel processing engine. The word "render" can be widely interpreted here. Just a simplified explanation from the api documentation: "One real-world example of an application is a game engine like Unity. Unity calls OpenVR API to get the position and orientation of any attached VR headset and apply them to the Main Camera. Unity then sends the camera image to OpenVR. OpenVR does some operations on the image and then displays it to the real headset screen." Call it what you want...
        Last edited 10-15-2020, 11:57 AM.
        i9 10920X @4.8GHz, 3090 OC, 64 GB RAM @3600, Valve Index, HOTAS & Rudder: all Virpil with FFB base mod (G940)

        Comment


          #24
          Originally posted by Alec Delorean View Post
          Goodness...

          1. Reverb owners have to use the Windows Mixed Reality plugin for SteamVR. The plugin handles all the necessary api requests first before Steam VR handles the rest. A reverb user will only see the end result, that's why it shows their native resolution at 100% in Steam VR.
          2. There is always the need to postprocess lens distortion for a flat frame from a game camera for any kind of VR HMD. Therefore you always need more pixels in a rectangular frame that will be lost during this operation because of the lens matched distortion. This is a basic VR process to display a flat picture on a flat panel that is only being watched through convex lesnes.
          3. The PD value in DCS acts as an override in combination with SteamVR. The openVR api asks the game to render a given resolution, but DCS returns a much higher resolution frame if the PD value is higher than 1.0. This is wasting frametime, because it obviously won't fit and needs to be resampled. Letting SteamVR managing resolution and supersampling instead is the much more efficient way. It also has an extra option of an advanced supersample filter under Video Settings. Why don't you test this for yourself? Try to set a PD of 0.5 in DCS and look what happens. The result will be a super blocky low res image, but still supersampled to a high resolution frame with your settings in SteamVR 100% 150% whatsoever...
          The difference might be if DCS is using the oculus api, here the PD value seems to hook directly into it. I can't test this here.

          SteamVR and it's openVR api is a image/pixel processing engine. The word "render" can be widely interpreted here. Just a simplified explanation from the api documentation: "One real-world example of an application is a game engine like Unity. Unity calls OpenVR API to get the position and orientation of any attached VR headset and apply them to the Main Camera. Unity then sends the camera image to OpenVR. OpenVR does some operations on the image and then displays it to the real headset screen." Call it what you want...
          It's full of incorrect and irrelevant information but not gonna try to convince you.

          Comment


            #25
            Yeah, like you tried to explain to me in another thread, that Windows, mirroring a screen directly from the front-buffer of the GPU to a second monitor is exactly the same like the extra VR mirror window on the desktop while having VR running. And that this extra window doesn't need extra VRAM to be rendered on the desktop. (Well it does, because it's copied from the VR screen buffer and stored in the front-buffer for the monitor, of course this takes extra space in VRAM. If more windows are open, the more information needs to be stored in the front-buffer)
            i9 10920X @4.8GHz, 3090 OC, 64 GB RAM @3600, Valve Index, HOTAS & Rudder: all Virpil with FFB base mod (G940)

            Comment


              #26
              Originally posted by Alec Delorean View Post
              Yeah, like you tried to explain to me in another thread, that Windows, mirroring a screen directly from the front-buffer of the GPU to a second monitor is exactly the same like the extra VR mirror window on the desktop while having VR running. And that this extra window doesn't need extra VRAM to be rendered on the desktop. (Well it does, because it's copied from the VR screen buffer and stored in the front-buffer for the monitor, of course this takes extra space in VRAM. If more windows are open, the more information needs to be stored in the front-buffer)
              LMAO, you're bringing the other thread into this one too? That one I stopped responding to you just like this one. That there's no reasoning with you and not worth my time. Not because you are right.

              You can continue to believe VR mirror take extra resources.
              Last edited 10-15-2020, 12:28 PM.

              Comment


                #27
                Originally posted by Taz1004 View Post
                LMAO, you're bringing the other thread into this one too? That one I stopped responding to you just like this one. That there's no reasoning with you and not worth my time. Not because you are right.

                You can continue to believe VR mirror take extra resources.
                https://www.bradford.ac.uk/t4-ssis/r...ce/page_07.htm

                Please provide evidence then. Real information how stuff works and how it matches your posts.
                i9 10920X @4.8GHz, 3090 OC, 64 GB RAM @3600, Valve Index, HOTAS & Rudder: all Virpil with FFB base mod (G940)

                Comment


                  #28
                  well this is really bloody helpful. 2 directly contradictory answers
                  7700k @5ghz, 32gb 3200mhz ram, 2080ti, nvme drives, valve index vr

                  Comment


                    #29
                    Originally posted by eatthis View Post
                    well this is really bloody helpful. 2 directly contradictory answers
                    About par , isn't it ?
                    9700k , Aorus Pro wifi , 32gb 3200 mhz CL16 , 1tb EVO 970 , EVGA 1070 FE , Seasonic Prime 850w Gold , Coolermaster H500m , Noctua NH-D15S , CH Hotas on Foxxmounts , CH pedals , RiftS

                    "Hold my beer"

                    Comment


                      #30
                      Originally posted by eatthis View Post
                      well this is really bloody helpful. 2 directly contradictory answers
                      The answer is in his post. If you actually try to read and understand, you'll see it does not make sense at all. For example, Reverb reports 100% in SteamVR because of it's higher native resolution. Not because of WMR. Other lower resolution WMR HMD will have same issue as Index. He's just passing on his own theory as facts. He's done that on every thread.

                      He believes VR mirror display (not cockpit mirror) hurts performance. Enuf said.
                      Last edited 10-15-2020, 07:06 PM.

                      Comment


                        #31
                        So it makes sense to me that each eye on a VR head set needs to be treated as a separate image as each has a different perspective. What i do not understand is why a mirrored display would use more resources... isn't it the same signal just being routed to two places instead of one? If the mirrored screen was a different resolution then I could see this taking more resources for the up or down scaling... but otherwise I am hard press to see where more Vram or processing time would be needed. Full disclosure I am not an expert by any means and I am always looking to learn.


                        Mathieas

                        Comment


                          #32
                          Originally posted by Mathieas View Post
                          So it makes sense to me that each eye on a VR head set needs to be treated as a separate image as each has a different perspective. What i do not understand is why a mirrored display would use more resources... isn't it the same signal just being routed to two places instead of one? If the mirrored screen was a different resolution then I could see this taking more resources for the up or down scaling... but otherwise I am hard press to see where more Vram or processing time would be needed. Full disclosure I am not an expert by any means and I am always looking to learn.


                          Mathieas
                          Your system spends its resources calculating the frames. Not on displaying it. You can mirror your games to multiple displays and wouldn't hurt fps. You can google it. But this is whole another topic and there's another useless thread about it. Some people will believe that it does and there's nothing you can do to convince them otherwise.
                          Last edited 10-15-2020, 08:55 PM.

                          Comment


                            #33
                            Originally posted by eatthis View Post
                            So leaving pd on 1.0 instead of 1.2 and setting ss to 120% will hafe the same graphical effect but shift the load from the cou to the gpu?


                            No...pd of 1.2 is much higher res then ss of 120%.
                            For example pd of 1.4 equals to a SS of about 200%.

                            The pixel density is calculated multiplying horizontal res * pixel density value and vertical res * pd value.
                            While the steam supersampling is calculating multiplying the total pixel count per SS value.
                            Practical example:

                            Base res 2880 horizontal * 1600 vertical (valve index physical res).

                            Pd of 1.2 equals to

                            3456*1920 = 6.635.520 pixels

                            SS of 120% equals to

                            (2880*1600)*1.2 = 5.529.600 pixels

                            Apart from this difference, there are technicalities that other users can explain better.
                            Usually the standard 100% resolution is different from the real physical res due to lens distortion correction algorithms...so for example for me the valve index reported 100% res is almost 200% of the physical one.
                            So before going crazy with the supersampling values think about your hardware
                            Vincent "Virus" DThe
                            PC: 9900K/RTX2080Ti, 32GB RAM.
                            Joystick bases: Winwing SuperLibra, FSSB R3 Warthog
                            Joystick grips: Winwing F18, TM F-16
                            Throttles: Winwing Super Taurus
                            Hardware: MFG Crosswind Rudder, 4 Thrustmaster Cougar MFD, Logitech G13
                            VR:Valve Index

                            Comment


                              #34
                              Originally posted by Taz1004 View Post
                              LMAO, you're bringing the other thread into this one too? That one I stopped responding to you just like this one. That there's no reasoning with you and not worth my time. Not because you are right.

                              You can continue to believe VR mirror take extra resources.


                              It does...the impact is negligible anyway
                              Vincent "Virus" DThe
                              PC: 9900K/RTX2080Ti, 32GB RAM.
                              Joystick bases: Winwing SuperLibra, FSSB R3 Warthog
                              Joystick grips: Winwing F18, TM F-16
                              Throttles: Winwing Super Taurus
                              Hardware: MFG Crosswind Rudder, 4 Thrustmaster Cougar MFD, Logitech G13
                              VR:Valve Index

                              Comment


                                #35
                                I cannot find fault in Alec Delorians' argument here.
                                It is as I find it, PD is just a nightmare for frame rates.

                                Perhaps the technical reasons may be incorrect but the results are spot on and that is all we need.
                                No need to muddy the waters on this topic as far as I am concerned. 1.0 PD is the max!

                                HP pro Reverb.

                                Current settings:
                                Windows VR setting: IPD is 64.5mm, High image quality, 90Hz refresh rate.
                                Steam: VR SS set to 100%, motionReprojectionMode set to "motionreproduction" and Locked in at 45 Hz display,
                                DCS: Pixel Density 1.0, Forced IPD at 55 (perceived world size), 2 X MSAA, 0 X SSAA.
                                My real IPD is 64.5mm. Prescription VROptition lenses installed.
                                VR Driver system:
                                I9-9900KS 5Ghz CPU. XI Hero motherboard and 2080ti graphics card, 32 gigs Ram 3200 Hz. No OC at the mo.

                                Comment


                                  #36
                                  Originally posted by Rogue Trooper View Post
                                  I cannot find fault in Alec Delorians' argument here.
                                  It is as I find it, PD is just a nightmare for frame rates.

                                  Perhaps the technical reasons may be incorrect but the results are spot on and that is all we need.
                                  No need to muddy the waters on this topic as far as I am concerned. 1.0 PD is the max!
                                  +1. I tried to give the bottom line to the OP back on page 1, but I guess verbosity is what gets peoples' attention. Oh well..
                                  ASRock Z370 Professional Gaming i7, Intel i7 8700K Coffee Lake processor @4.7GHz, Corsair H105 Liquid Cooler, G.SKILL Ripjaws V Series 64GB DDR4 3400 memory, MSI GTX2080Ti Gaming X Trio 11GB video card, Samsung 840PRO 256MB SSD for Windows 10 64-bit OS, Samsung 850PRO 2TB SSD for program files, LG WH14NS40 Blu-Ray burner. HOTAS Warthog, Saitek Pedals, HP Reverb Pro, and Oculus Rift CV1. Partridge and pear tree pending.

                                  Comment


                                    #37
                                    Originally posted by VirusAM View Post
                                    No...pd of 1.2 is much higher res then ss of 120%.
                                    For example pd of 1.4 equals to a SS of about 200%.

                                    The pixel density is calculated multiplying horizontal res * pixel density value and vertical res * pd value.
                                    While the steam supersampling is calculating multiplying the total pixel count per SS value.
                                    Practical example:

                                    Base res 2880 horizontal * 1600 vertical (valve index physical res).

                                    Pd of 1.2 equals to

                                    3456*1920 = 6.635.520 pixels

                                    SS of 120% equals to

                                    (2880*1600)*1.2 = 5.529.600 pixels

                                    Apart from this difference, there are technicalities that other users can explain better.
                                    Usually the standard 100% resolution is different from the real physical res due to lens distortion correction algorithms...so for example for me the valve index reported 100% res is almost 200% of the physical one.
                                    So before going crazy with the supersampling values think about your hardware
                                    Our systems are very similar. Would you mind screen shot ing me your steam and in game settings? (Assuming you’re happy with your performance. You can pm if you like.
                                    Thanks
                                    I9 (5Ghz turbo)2080ti 64Gb 3200 ram. 3 drives. A sata 2tb storage and 2 M.2 drives. 1 is 1tb, 1 is 500gb.
                                    Valve Index, Virpil t50 cm2 stick, t50 base and v3 throttle w mini stick. MFG crosswind pedals.

                                    Comment


                                      #38
                                      Originally posted by Rogue Trooper View Post
                                      I cannot find fault in Alec Delorians' argument here.
                                      It is as I find it, PD is just a nightmare for frame rates.

                                      Perhaps the technical reasons may be incorrect but the results are spot on and that is all we need.
                                      No need to muddy the waters on this topic as far as I am concerned. 1.0 PD is the max!
                                      Maybe you guys aren't reading what he said.

                                      Going higher on the PD setting will start to waste rendered pixels because the panels aren't able to display them 1:1 anyway. This will also cost valuable CPU frametime.

                                      Comment


                                        #39
                                        Originally posted by VirusAM View Post
                                        It does...the impact is negligible anyway
                                        He was claiming VR mirror having separate frame buffer and taking 1.4GB of VRAM. Not negligible. Windows took 1.4GB. Having VR mirror on top of it doesn't.

                                        Comment


                                          #40
                                          @Taz1004

                                          Read again, don't twist what i wrote. I was measuring the whole windows desktop taking 1.4GB VRAM on a ultrawide monitor. Every new window aka "running app" on the desktop, takes a bit more VRAM for the Desktop's screenbuffer. That's how Windows GUI works. Every new opened window needs VRAM to get it's content displayed, because it's a 2D plane with graphical elements which need pixels to be drawn and stored somewhere. It doesn't cost "performance" but it needs resources like VRAM. The more windows aka "apps" are open the less VRAM is left for the game. It's a couple of hundred MB per window on a high resolution monitor.

                                          The need for a bigger rendered frame to counter lens distortion is common VR technology for every headset:
                                          Last edited 10-16-2020, 07:43 AM.
                                          i9 10920X @4.8GHz, 3090 OC, 64 GB RAM @3600, Valve Index, HOTAS & Rudder: all Virpil with FFB base mod (G940)

                                          Comment

                                          Working...
                                          X