Jump to content

Al-Azraq

Members
  • Posts

    388
  • Joined

  • Last visited

Posts posted by Al-Azraq

  1. On 3/15/2023 at 12:15 PM, YoYo said:

    OpenXR and your settings in Tools will overwrite settings in SteamVR as runtime so:

    1/ use OpenXR Developer Tools to set your settings on 2D screen (easiest method) or

    2/ go to settings of OXR in SteamVR (in 3D, when you have goggles on your head) and use controller to see OpenXR settings or

    3/ you can use mbucchia's Toolkit in 3D as well.

     

    Thanks YoYo.

    So if I get rid of OpenXR Tools, then it will not overwrite my SteamVR settings and I will be able to control everything from there which is what I want.

    I will try later!

  2. Hey guys, I like to use SteamVR as the default OpenXR runtime as I want to keep using the per-app resolution settings. However, it seems that the selected resolution in SteamVR does not get applied and fpsVR reports 100% resolution.

    Is this a reporting problem from fpsVR or is SteamVR uncapable of changing the resolution through OpenXR?

    Also I have another question: if SteamVR is set as the default OpenXR runtime, is SteamVR in charge of reprojection?

    Thanks!

  3. What a monster of a GPU, but terrible pricing.

    Makes me consider storing my HP Reverb G2 and come back two years later when I upgrade my 3070 Ti to enjoy VR at 100% resolution. Maybe I just do that and back to TrackIR for now. Happy to know that we can brute force VR now, but that doesn't mean ED has still to do their part and optimise the game.

    • Like 3
  4. 46 minutes ago, Wali763 said:

    Looking at the pics it seems like the CPU is holding back the GPU.

    So my guess would be, that in scenarios with low CPU-load, one will get a decent uplift in fps, but in scenarios with high CPU-demand the 4090 could perform the same as a 3090.

    Having invested so much money, Id be quite unhappy, still being stuck at/around 60fps. But using motionsmoothing, 60fps would still mean 45fps since 90 is too far away. Of courese if you have 90Hz HMD.

    Anyway, this is one more reason, that multi-thread-support should be releasing as soon as possible!

    We have been saying that for a while now, even Nineline said to wait for multithreading before spending that money on a GPU but here we are. Sure the 4090 will offer a huge boost when not CPU bound (I think that in the air) but I wouldn't like spending 2.000 € just to have improvements just under certain circumstances.

    • Like 1
  5. I don't think Heatblur is aiming to a particular Tranche or block but they will model all those capabilities, systems, and hardware they can get reliable information of. I find this approach the correct one because the representation of these features will be realistic, you will just not be able to say (this is a Tranche 2 Block 10). If they aim for an specific Tranche or Block, they will be limiting themselves.

    What if they decide to do a Tranche 2 Block 10 but can get information on the Brimstone or Meteor missile? Should they leave these stuff out because they are not for the specific version modelled?

    In my opinion, Heatblur is taking the right approach here.

    • Like 6
  6. On 5/10/2022 at 5:39 PM, Burt said:

    Any news on when wmr4steam is gonna fix the issue that made me go to another version I think I had to switch to LMK version I believe it was that or ilk. Im having no issues at the moment with that version but just wondering if anybody knows of any progress about this.

    Thanks

     

    I went back to the regular version and it is fine for me.

    • Thanks 1
  7. 15 hours ago, Toumal said:

    4) The systems simulation literally is the content in this game. You don't just buy an aircraft mesh, you purchase a complete cockpit and most of the systems in it, along with the documentation and all that.

    This is a great point.

    Every game needs gameplay loops and DCS gameplay loop is system managing. It is what keeps the player engaged throughout the session and it is what the player will try to improve on.

    • Like 1
  8. You are missing a big point here which is: we enjoy learning how to operate these planes, we enjoy flying them, we enjoy fighting with them. Yes, we know this is just a game at the end of the day but we enjoy the time we spend in it.

    I think that simulators are just not for you, and I don't think you should try to convince us that what we do is pointless, because it is not as long as we have fun.

    • Like 3
  9. 14 hours ago, Gypsy 1-1 said:

    Well, it's more realistic that way - make sure you have AF turned to 16x and in VR it's always going to be an issue. Generally I orient myself on the geometry of the ship, the wake and such. 

    EDIT: Real carrier decks are usually pretty dark and markings, especially on the older ones like the Forrestal Class tend to be worn out as a cruise progresses. Here is one video example, albeit low quality: 

     

     

    You are right, it looks pretty dark in that video. I guess it depends on how much time has passed since it was last repainted.

    I would clear for an IFLOLS in DCS looking as clear as in that video!

  10. 2 hours ago, Gypsy 1-1 said:

    Yeah the LOD is terrible - it only really becomes visible (or barely visible) close in. As well as the fact that you can see the FLOLS LOD from 360° around the ship...

    Yeah, that as well.

    Also, am I the only one having many difficulties spotting the deck yellow and white lines? It is just a black deck for me until I'm very close. I have a Reverb G2.

  11. On 3/30/2022 at 7:19 PM, Gypsy 1-1 said:

    So, when can we expect the much needed fixes to the Forrestal FLOLS appearence and brightness as well as LOD and the damage taking issue on a trap?

    I play on VR and the FLOLS is just very bright to the point I really can't see where the ball is exactly.

    • Like 2
  12. 5 minutes ago, DeltaMike said:

    I was just keeping a low profile while you and Edmuss did all the heavy lifting lol.  Yall did such a service for the DCS community, your efforts and generosity with your knowledge are greatly appreciated.

    @Al-Azraq

    I was wondering if it's more of a latency issue.  All we can measure is render time; once steam lets go of it, who knows what kind of shenanigans are going on.  Not sure how we could test that, except I would predict more motion sickness in SVR/WMR than with OpenXR

     

     

    I think everything is related, input latency included. Lower FPS and frame times result in better latency which are everything for stability.

    I would pay money to Digital Foundry to make a DCS analysis including a VR deep review.

  13. 12 minutes ago, DeltaMike said:

    Well well well.  First impressions:

    1. WHOA!
    2. Haven't done any benchmarking or anything but the amount of judder on the 3-9 line is dramatically reduced (I have motion smoothing off).   There's still judder but it looks like in the movies, more of a blurring than a stuttering.  I'll take it. 
    3. I get the feeling there is something fundamentally fubar about the WMR/SVR interface that render times and FPS isn't quite getting at.  I'm showing 50fps on the deck on the PG map (which is nothing to sneeze at) but the amount of stuttering is but a fraction of what I would have expected at that frame rate.  Likewise and maybe even more obvious when you do a snap roll.  Put it this way, at 50fps in OpenVR I'd be turning on motion reprojection as fast as my fingers could fly.  Really no need here.
    4. Which is good, because motion reprojection in OpenXR isn't any better than WMR's version imo (based on flying around in MSFS.  Which is a hot mess.)
    5. I don't miss ReShade, don't feel like I particularly need it here.  Maybe it's just placebo but it just kinda looks better.  I agree, shimmering is less. I've been spending a lot of time over Vegas benchmarking recently, and I must say, NTTR is looking good and it's just weirdly smooth.  
    6. I did exactly what Nikoel said and it worked on the FIRST TRY.  This is literally the only thing that's ever worked on the first try for me, in VR, ever.  Read the instructions, folks. 
    7. Dear Steam VR.  I think we should start seeing other people.  

    Thanks for my shadows back, Nikoel.  Hate to sound like a wuss, but dangit.  I want shadows. 

     

    Your FPS are the same but you gained smoothness and stability because as shown in the benchmark a couple of posts ago, 1% lows and minimum FPS are much much better which means frame time is stable.

    I'll explain better: it is better to draw a frame at a constant pace throughout a second, than rendering 40 during the first half of it and 10 during the second half of it. The output for both would be 50 FPS, but as they are distributed equally the smoothness in game will be much much better in the first case.

    • Like 1
  14. 3 hours ago, mhe said:

    @BIGNEWYHi there, we know there is not much of a timeline for Vulkan and Multicore support, but do you happen to know whether the Vulkan renderer rewrite will entail OpenXR support for the VR portion of the sim? Thank you! 

    I’ve read Bignewy or Nineline say that OpenXR support will come eventually. Also in the 15/10/2021 newsletter when they updated the multithreading status, VR support was among the remaining tasks. I don’t think OpenXR support will come with multithreading but with Vulkan, but I’m hopeful for performance improvements with it as well.

  15. Guys, DCS just takes all the VRAM that it is available (allocation) and then uses what it needs. It doesn't mean that it uses everything, it just reserves it for itself just in case.

    But yes, it is an inefficient process and DCS is full of memory leaks.

  16. 5 hours ago, DeltaMike said:

    1.. Neither.  They both do the same thing.

    The big picture is, performance correlates with pixel count.  Pixel count increases linearly with Steam resolution, and with the square of pixel density.  For a G2, the formula is 

    (3100x3100)*g*a*p^2 

    Where g = Steam General Resolution, a = Per App resolution, p = pixel density

    Doesn't matter how you get there, long as the numbers add up.  Best to just adjust one, and set the others to 1.0 or 100%.  With Steam resolution being the more granular parameter. 

    There's some debate as to whether adjusting one is better than the other.  If there's a difference in quality, it is subtle indeed.  It's OK to ignore that debate until we come up with some solid data.  Meanwhile, pick one. 

    Resolution is king in DCS. Get a high res headset, and shoot for native. 

    2.  In terms of antialiasing, in decreasing order of performance hit: 

    • Supersampling has theoretical benefits, especially for low resolution headsets, but in DCS at least, eliminating jaggies and shimmering isn't one of them.  Flogs the daylights out of your GPU. 
    • MSAA looks great but it crushes your framerate.  Ditto.  I think MSAA is quite a bit more efficient than supersampling at getting rid of jaggies and shimmering.  
    • FXAA and SMAA accomplish something at least, at about the third of the cost of MSAA.  (SMAA is probably the better technology and the ReShade version of it is pretty impressive imo)

     

    3. It all sucks.  People were kind of butt-hurt that DCS went to deferred rendering, which killed MSAA. But time marches on, and there's new technology we are begging DCS for.  Meanwhile, take a ride around NTTR and then run the same mission in MSFS. Which one is better? I imagine your answer is like mine, "it depends."  Do you want to look at the scenery, or blow it up?  Valid question I think

     

    Good summary! I’m going to update MSFS 2020 today and try it on VR, see how it looks on Nevada.

    And yeah, resolution is what impacts performance more. At first I had almost everything on low but then I noticed that I could increase settings without performance impact as I had some headroom and I’m locked at 45 fps due to reprojection anyways.

    With the PC at my signature, I can run DCS at 80% resolution, medium settings, terrain shadows flat, MSAA off. As we talked before MSAA is increasing pixels which paired with deferred rendering it destroys performance.

  17. 28 minutes ago, DeltaMike said:

    Interesting, so we could do both then.  Supersample via steam or DCS, and then do a pass with TAA.  Interesting.  I'd try that in MSFS, except I'm scared my GPU will catch on fire

    Regarding draw distance, I guess that's been a topic of discussion for years.  Good discussion here for example.  

    When I talk about spotting distance in my tests, I'm referring to the pixel-sized sprites.  Way I look at it, in VR, we have the same probletunity as people with low-res flat screen monitors.  Even though you might have high-res monitors in your headset, when you put em right on your eyeball and blow them up to iMax proportions, each pixel looks pretty big, and I think we do come out ahead by turning AA off.  

    As for once you're close enough to get a 3D object, I still wonder about level of detail as a function of draw distance, but I wouldn't know one way or the other

    Yes sure, you can do both at the same time. TAA as it doesn't render additional pixels it should have a really minimal computational cost but it is not easy to demonstrate because the games that introduce this technique, usually don't offer the option to turn it off because otherwise it breaks lighting, shadows, etc.

    I don't think LOD has an effect for spotting distance as the object gets rendered at low poly, but the shape remains the same if the developer did its job right.

  18. 3 minutes ago, Burt said:

    Honestly I wish I never got into VR . Before I did I was playing on a 34” ultrawide 120hz monitor with a 3090 with TIR and man was it beautiful … so crisp like Wags videos. The clouds , clarity and texture are what makes this sim so superb with 2D. But once I got a G2 and did all the tweaking and start using it more it was all over. Every once in a while I will go back to 2d but the feeling of pulling G’s just ain’t there like it is in VR. I miss the full screen clarity the most and stutter free. 
     

    Yeah same here.

    I miss the clarity and clear image of 2D as well as the sutter free (most of the time) but somehow I just can't come back.

    • Like 1
  19. 34 minutes ago, DeltaMike said:

    OK I set up this test.  Obviously these tanks are a lot closer than an aircraft would be.  The pic is taken at 100/100/1.0 with no anti-aliasing, and it's actually a fairly good representation of what I see through the G2.  The third tank is shimmery but identifiable as a tank.  The fourth, I can't consistently see the gun so I might know it's armor, but wouldn't know it was a tank.  5 and 6 are just blobs, I couldn't tell 6 from any other vehicle.

    SSAAx2 and MSAAx2 look about the same.  

    Running a pixel density of 1.4 (equivalent to Steam at 196%), I think tank #4 looks a little clearer.  Shimmery but I think I could identify it as a tank.

    To my surprise setting steam resolution at 200%, with SuperDooperSampling turned on, doesn't look as good as PD 1.4.  

    I think the supersampling techniques (SSAA, MSAA) have some potential here, as you are increasing the amount of information available to your brain, averaged over time.  FXAA of course does not increase information available and seems to be reducing it actually.  Unclear to me which way TAA swings.

    Biggest concern I have is the resolution of objects as a function of draw distance.  I think it's possible the information just isn't there. 

     

    20220321_090318_MixedReality.jpg

     

    Thanks a lot for more testing! A couple of notes though:

    SSAA doesn't affect the VR image, it is supersampling but for a 2D image. The way of supersampling in VR in DCS is going to SteamVR and setting higher resolutions but even if you put some crazy resolution you won't get rid of the shimmering because even in that case, the upscaled image will have jaggies which will not be entirely removed when downscaled to the headset resolution. To clear the image we will also need an AA solution as well.

    SSAA is a very expensive AA solution because it is not selective and it creates extra pixels for the whole frame. Techniques like MSAA generate extra pixels but only in the relevant areas such as the object borders, that's why MSAA sometimes doesn't apply to every object. So MSAA is sort of supersampling but only in the relevant areas.

    TAA being the most modern one and less taxing is perfect for VR because it uses information from past and present frames without the need of upsampling or creating new pixels (contrary to what MSAA does). Then it reconstructs the image using these pixels knowing where they were placed in past frames and doing some fancy calculations, that's why it needs motion vectors introduced by the developer.

  20. 10 minutes ago, DeltaMike said:

    Good point. 

    It's tempting to say, we are up against a hard limit in terms of information available, based on the number of physical pixels-per-degree.  But, that's the probletunity of supersampling in VR; if you keep your head moving, there's more information at hand than the display can render.  You just have to rely on your brain to stitch it all together.  

    In other words, your CPU makes a map, and your brain makes a map.  All the GPU has to do is, act as an intermediary between the two.

    With that in mind, maybe we should set the matter of shimmering aside and see if supersampling (one way or the other) enhances object identification at intermediate distances.   

     

    I think we can be very sure that supersampling indeed improves object identification as what you achieve with supersampling, it is the same you achieve with antialiasing which is providing more information to the renderer so it can output a more defined and rich image. I even tried to increase SteamVR resolution to 150% to try how the image was improved and it did reduced shimmering quite a bit, but it was still there.

    I really think that the only solution to it is having TAA, because let's not forget that MSAA with deferred rendering will never ever clean all the image. Even with forward rendering MSAA doesn't apply to some 2D objects like fences.

    Finally, I think that shimmering and contact ID go together because if you can clean map objects, then you can clean aircraft however, take also into consideration that identification of enemy planes is not easy in real life either and we may have been spoiled by the zoom function in 2D.

  21. 39 minutes ago, DeltaMike said:

    Ah.  That explains a lot.

    Based on your explanation, ReShade's "TAA" isn't really TAA then.  I will say, it's kind of a nifty hack.  It is depth-aware, so for example you can set it to where your cockpit isn't blurry but stuff "out there" is.  Clearly not made for VR, I imagine it works best with a static camera.  But in the end, all it really does it make the image blurry, with ghosting besides.  

    I hear what you're saying about MSAA.  I feel much the same way about generic supersampling.  As I went from Rift, to Rift S, to G2, seems the cost of supersampling has increased exponentially but the benefit hasn't kept up.  To the point where I'm starting to think it's pointless.  

    MSFS's TAA implementation does work, but overall the scene is kind of a blurry mess.  It's pretty, so let's call it a hot mess.  It's not always that hazy out there, especially out over the desert.  When I say "traffic in sight" what I mean is, I see a blob flashing red and green.  Is it a Cessna?  A 747?  Flying saucer?  No way to be sure.  Which is fine.  Apparently the only thing that crashes in MSFS is MSFS

    Far as DCS is concerned, my feeling is, pretty is as pretty does.  The game plays great in a high-resolution headset.  

    I guess to summarize my feelings on anti-aliasing:

    • Supersampling is the bomb if you have a low-res headset; kind of disappointing given the cost, if you have a high-res HMD. To me, anyway. 
    • MSAA is fine, if you have the GPU to drive it.
    • VR Toolkit or SMAA if fine, if you don't
    • None are necessary to play the game, which plays best at native resolution imo

    Aaaaand I'm spent

     

    Yeah I agree, haven't tried MSFS on VR though as I'm too lazy to set it up but I prefer blurry rather than shimmering. Also I think that reducing shimmering via Supersampling in a high resolution headset like the G2 is maybe one GPU generation away and for now we will just have to live with the shimmering.

    That's why I have been advocating for TAA so much. Sure it tends to blurry the image but if it is well implemented it is not really an issue and it is a very light AA technique. I'll take the blurriness. Also with TAA the door is open for DLSS or FSR 2.0.

    Regarding contact ID, I also have difficulties IDing even in the other sim where I run at 100% SteamVR resolution with the G2. The lack of AA makes the bandits become a mess of pixels until I'm quite close and I can't afford to turn on MSAA at 100% resolution.

×
×
  • Create New...