New geo-11 driver

[3DNovice]

I wonder if this could be leveraged to make Geo-11 even faster?

PlayStation 3 emulator increases its CPU performance by 30% with AVX-512
https://www.guru3d.com/news-story/plays ... x-512.html

https://whatcookie.github.io/posts/why- ... for-rpcs3/

[costiq]

Geo-11: Amazing news!

Congratulations and Thank you bo3bber & davegl1234!

Being able to experience again the Real 3D, on a big 3DTV screen, using the latest video drivers, is the best gift 2-eyed people got this year!

Can't wait to see this baby grow and get the universal acclaim it deserves.

[bo3bber]

I wonder if this could be leveraged to make Geo-11 even faster?

PlayStation 3 emulator increases its CPU performance by 30% with AVX-512
https://www.guru3d.com/news-story/plays ... x-512.html

https://whatcookie.github.io/posts/why- ... for-rpcs3/

Too much of a mess to even consider. Not having support on all CPUs means fragmented code and some people with worse experience. Also mostly I think that is not our bottleneck, but we'll have to do better profiling of geo-11.

https://www.extremetech.com/computing/3 ... nful-death

[bo3bber]

I was having wildly inconsistent results with/without the driver hack, so I was asked to report my tests here so I tested Splitgate (a really cool, free Unreal Engine 4 DX11 game)
Basically, I recently noticed that I don't need the driver hack for a bunch of DX11 games/apps, but I can count on Unreal Engine 4 games to crash without the hack (unless, as shown below, I use SBS output instead of DX9/DX11 3DV output)



Full details and logs here: https://airtable.com/shrjrcX7OhDRDoBam
I'll keep reporting more as I test

That's pretty interesting. Since it doesn't happen with SBS, that suggests that UE4 is somehow sensitive to the dx9 path being available. UE3 had 3D Vision support for a long time, and I wonder if there is some legacy code that sees 3D Vision and flips out. What driver is that tested against? Once we get past 452.06 its dx9 only, and might change the results.

[bo3bber]

Norton 360 (used for PC security on my PC) continues trying to quarantine new d3d11.dll file. I have not had a problem with Norton interfering in 3D fix installations before. Norton claims it has WS.Reputation threat.

Yeah, that's just a rubbish message. WS.Reputation means it hasn't the seen the dll before. Well, duh. Antivirus is mostly crap nowadays, worse than the problem they claim to solve.

[bo3bber]

geo-11 outputs sequential frame stereo? Is this correct? Is single frame stereo forseen for the future?

It is single frame stereo. By frame sequential the OP is referring to Nvidia 3D Vision output for 3D Vision monitors, which shows the (synchronized, not sequential) left and right frames by alternating the opacity of the glasses, and each eye getting 60Hz. Active 3D.

Perfect! In that case it was only a misunderstanding on my part.

Yes, different terms. frame-sequential is part of the 3D standards, and was always standard for all NVidia 3D Vision hardware. You are probably thinking of Alternate Eye Rendering, which is a different approach to generating 3D. AER is good for cheating some of the complexities, but is generally not preferable if we have geometric 3D available.

[3DJ]

I was having wildly inconsistent results with/without the driver hack, so I was asked to report my tests here so I tested Splitgate (a really cool, free Unreal Engine 4 DX11 game)
Basically, I recently noticed that I don't need the driver hack for a bunch of DX11 games/apps, but I can count on Unreal Engine 4 games to crash without the hack (unless, as shown below, I use SBS output instead of DX9/DX11 3DV output)



Full details and logs here: https://airtable.com/shrjrcX7OhDRDoBam
I'll keep reporting more as I test

That's pretty interesting. Since it doesn't happen with SBS, that suggests that UE4 is somehow sensitive to the dx9 path being available. UE3 had 3D Vision support for a long time, and I wonder if there is some legacy code that sees 3D Vision and flips out. What driver is that tested against? Once we get past 452.06 its dx9 only, and might change the results.

All tests were done in 452.06. I added more specs and logs for each test here: https://airtable.com/shrjrcX7OhDRDoBam

[3DJ]

Also, tested another game with notoriously awful performance in 3D and here are the results!

Before (just turning on 3D vision, with or without the driver hack):

After (with Geo-11):


However, it's very unstable. It consistently crashes after a few minutes of playtime or even on startup.
I've attached the logs in the same spreadsheet: https://airtable.com/shrjrcX7OhDRDoBam

[Lysander]

Some questions came up on discord, capturing them and answers by masterotaku here for the record:

"- Is driver 425 still best for performance or is that only for regular NV 3D V?
Only for regular 3D Vision.

- Supposed to use dx9 games too or only dx11?
Only for DX11 games.

- Still supposed to install NV 3D Vision (some people said they don't have to install it anymore?)
If you use a 3D Vision monitor, you need the 3D Vision driver, to make the glasses and monitor work (but you don't need old drivers or GPUs).

- Still supposed to use helixmod fixes or are those only for normal NV 3DV?
The helixmod fixes are the base, and you're supposed to install parts of geo-11 on top of that (read the instructions that are around). However, games may require some tweaking to correct some behavior differences. That's why I'm uploading some geo-11 specific fixes for now.

- Is the stereo vision still has good looking as normal NV 3D Vision?
Visually, if everything goes right, the same (with more freedom about separation). Performance, between x2 and x3 better usually, in CPU limited situations.

- Is this for people with NV 3D Vision glasses and Lightboost monitors too or just for the 1% of NV 3D Vision users with SBS projector TVs?
For both. But compatibility for 3D Vision monitors still has room to improve. Just give it some time.

When you hear talks about "DX9", we mean outputting the DX11 game in a "DX9" window, to make the glasses work with drivers newer than 452.06. For now it's generally less compatible than outputting in DX11, which is more like 3D Vision does for DX11 games.
There are things to solve about video formats, fullscreen states, etc. It's a work in progress.

SBS/TAB/interlaced should be the most compatible and stable modes for now."

[ParadiseDecay]

Struggling to get any game to work, does anyone have a Video Guide? Cheers.

[RAGEdemon]

Hi guys,

My 3D display only has DLP-Link, which uses full resolution page flipping. It works great via 3D Vision.

Unfortunately, it can't do SBS nor TaB, otherwise it has no problems activating and syncing the glasses via the DLP-Link protocol if a page-flipping format is provided.

Is there a way to natively output a page-flipped format to my device which is GPU/driver agnostic?

The closest solution I can currently find referenced is "direct_mode=nvidia_dx9", which is said to not be native, nor high in compatibility.

TL;DR: Is there a way for the driver to natively output full resolution L-R-L-R-L-R-L-R-L-R-L-R please

Thank you.

[whyme466]

Note that TAB is generally better for 3D display than SBS, since it preserves precision in the stereo baseline (horizontal) axis. SBS should never be used in an interlaced display device (like LG passive 3DTVs), because it introduces an unnecessary horizontal resolution reduction, beyond the loss from interlaced lines.

[bo3bber]

All tests were done in 452.06. I added more specs and logs for each test here: https://airtable.com/shrjrcX7OhDRDoBam

Strange. I'm definitely wondering if there isn't leftover UE3 3D Vision code somehow.

BTW- 3DV turned off will always crash both dx9 and dx11 output modes. We don't have any error handling for no-3DV just yet.

[bo3bber]

Hi guys,

My 3D display only has DLP-Link, which uses full resolution page flipping. It works great via 3D Vision.

Unfortunately, it can't do SBS nor TaB, otherwise it has no problems activating and syncing the glasses via the DLP-Link protocol if a page-flipping format is provided.

Is there a way to natively output a page-flipped format to my device which is GPU/driver agnostic?

The closest solution I can currently find referenced is "direct_mode=nvidia_dx9", which is said to not be native, nor high in compatibility.

TL;DR: Is there a way for the driver to natively output full resolution L-R-L-R-L-R-L-R-L-R-L-R please

Thank you.

Does that qualify as being frame-sequential output? And presumably that works with NVidia 3D Vision for output? If that's true then the DX9 output mode will work here. It uses the DirectMode output of the driver, so it's still the nvidia backend that drives the hardware. I don't know what you mean by 'native'. The game is dx11, converted by geo-11 to internal full-SBS, then output via a DX9 window in 3DV Direct Mode. All these are fully supported modes, because the DX9 path is used for 3D Movies.

[RAGEdemon]

Thanks bo3b.

Indeed, it is frame sequential output.

I don't know how the driver works, so I hope you'll forgive my ignorance:

I was under the impression that the DX9 path was as yet 'buggy', and that for 'best compatibility', to use the 'native' SBS/TaB output.

By native outputs, I mean only the geo-11 driver is required, as per my understanding of the documentation:

Native modes are:

Code: Select all

   sbs: Any driver.  NVidia 3D Vision driver is not required.
   tab: Any driver.  NVidia 3D Vision driver is not required.   

Semi native:

Code: Select all

nvidia_dx9: Works on any driver. Mostly. Requires NVidia 3D Vision Driver. Works best on 452.06 with upscaling=0. Any driver past 452.06 requires game to be in fullscreen mode.

Non-native:

Code: Select all

nvidia_dx11: Works up until 452.06.  Fullscreen mode does not work.  Requires NVidia 3D Vision Driver.

So you're saying that the DX9 page-flipping/field-seqential mode will work as well as SBS/TaB modes after all? -- That would be brilliant!

[VR3D]

Note that TAB is generally better for 3D display than SBS, since it preserves precision in the stereo baseline (horizontal) axis. SBS should never be used in an interlaced display device (like LG passive 3DTVs), because it introduces an unnecessary horizontal resolution reduction, beyond the loss from interlaced lines.

I did not know that thank you. I was wondering why sbs looked a bit less clear on my LG OLED. I'll try TAB now.

EDIT: Man it's so much sharper in TAB versus SBS on my LG passive OLED. It looks more like the sharpness of 4k 2d but with a little bit of aliasing but well worth the clarity. Thanks again for the tip!

[3DNovice]

...

[whyme466]

I suggest some caution and care in the driver testing process. I accidently used the geo-11 fix files with my 3D Vision driver/OS version of God of War, and the attempted game launch crashed and corrupted my 425.31 driver. Even the 2D version of the game would not launch. I had to reinstall 425.31.

[b4thman]

I need to ask this: Anybody has tested if there is any difference running "Geo-11 with 425.31 insatalled" and "Geo-11 with 452.06 installed" ?

I have read somewhere that Geo-11 does not care about what driver is installed, but not sure if that is true and the performance is the same with those different drivers.

I am having some problems with 425.31, and at this moment I have 452.06 installed

[RAGEdemon]

I need to ask this: Anybody has tested if there is any difference running "Geo-11 with 425.31 insatalled" and "Geo-11 with 452.06 installed" ?

Indeed, I have:

viewtopic.php?t=26268&sid=9fe8081cd7a4f ... db3ca5f443

[b4thman]

Ok, good news it seems, as you conclude that "It looks like we no longer need to rely on 425.31 for performance". So I can stay away from 425.31 (425.31 gives more performance, but also gives me some other problems because of my Windows 10 version). Geo-11 is still not compatible with some games, so 425.31 may be a good choice for anybody who plays those games.

I wish Geo-11 will also improve GPU performace in future updates (I know it may not be possible, but that is a wish). Now my RTX 2080 is clearly the guilty of my bottlenecks.

[bo3bber]

I don't suppose it would be possible to wrap dx8 to geo-11, so the the "Test stereoscopic 3D..." option in NVCP will work again?

Or perhaps a drop in shader fix would work, converting PS2 shaders to PS3? Did dx8 use PS2 shaders? Dunno
I know that helixmod only supported PS3

Will geo-11 be getting a test image, like iz3D has?

Would be worth trying dgVoodoo to convert it to DX9 or even DX11. It should convert pretty well I would expect. You can find the app itself in the 3D Vision install folder.

We aren't sure how we'll package anything just yet. geo-11 by itself is not a complete app, it's just the driver/wrapper. So since it's not an exe, it doesn't really make sense to add a test image to it. Also, with no driver restrictions and being straight dx11, there is a lot less troubleshooting so it's not clear we need one.

I do think it would be a good thing for 3DFM/HelixVision Driver tab, assuming we keep that. But I can't think of any scenarios where that would actually break because it's straight dx11. Weirdo modes like nvidia_dx9 for sure.

[bo3bber]

I suggest some caution and care in the driver testing process. I accidently used the geo-11 fix files with my 3D Vision driver/OS version of God of War, and the attempted game launch crashed and corrupted my 425.31 driver. Even the 2D version of the game would not launch. I had to reinstall 425.31.

I don't know how that would be possible, but there are a lot of moving parts. Were you using the dx9 or dx11 output mode I would assume? Even in those modes, the code only makes non-destructive changes to driver mode, and does not set any parameters in the registry for example.

But advice is still sound- make a image backup of your system.

[sgsrules]

Does this work with DSR? I ussually play games at 1440p on my 720p projector. I gave this a shot with the latest drivers and the witcher 3 using dx9 mode and it works but i could only see the top right quadrant of the game. Everything else was offscreen. So only 720p worked. I'm guessing it might be because DSR only works with d3d11.

[sgsrules]

I rolled back to 452.06 and couldn't get the witcher 3 to work in dx11, screen was all black with 3d enabled. Switching to dx9 works, and dsr also works. For testing purposed i reinstalled the 516 driver and once again dsr wouldn't work the game would launch at the desired resolution but wouldn't downsample to fit in native res so you could only see a corner of the screen.

[zaibaker]

Wonderful news, thank you all the 3D vision community !

I was reading all the comments, it's very hard to follow discussions on this forum.

Any hope to move to a reddit like forum for better visibility?

I was waiting my new HMD (Cinera Edge Pro due to your post Bo3b) to go back in 3D Vision, but i suppose now that i will have to prepare myself earlier and give some feedbacks with my 2080 Ti i bought especially to keep 3D Vision working back in the days .

[BazzaLB]

I was reading all the comments, it's very hard to follow discussions on this forum.

If you think this is bad, you should try and follow discord.. LOL.

[RAGEdemon]

Does this work with DSR? I ussually play games at 1440p on my 720p projector. I gave this a shot with the latest drivers and the witcher 3 using dx9 mode and it works but i could only see the top right quadrant of the game. Everything else was offscreen. So only 720p worked. I'm guessing it might be because DSR only works with d3d11.

DSR works with GTA5's native mode and Psychonauts 2 via the DX11 path. I have not tested DX9. Have you tried toggling the scaling option in the ini file? How about making desktop 1440p before launching game?

Did you disable fullscreen optimisations? Overlays such as Steam/RTSS/Fraps/Reshade etc?

BTW, the latest drivers have AI DSR which produces the insanely good results of 4x DSR (1440p) at 2.25x DSR. You get a free 2X speed boost. It would be awesome to be able to get AI DSR working with geo-11.

(I too DSR my 800p to 1600p - I haven't done DSR testing, however)

[RAGEdemon]

Tested native SBS mode with VR HMD - results are stunning!

Various modes with recommended 452.06 drivers.

Non-native:
DX11 mode works well on a lot of games, assuming 3DV drivers are installed.
DX9 mode has issues - as expected, and stated in the documentation. Psychonauts 2 with Losti's universal fix, for example, just flashed black.

Native modes:
SBS mode: This is currently Half-SBS, and has Perfect output without issues, without 3DVision drivers installed - I tried this using a VR headset + Virtual Desktop in Half-SBS mode. One has to disable VSync and cap the FPS to your HMD, otherwise it will be capped to half your display, e.g. capped at 60fps if you have 120Hz display.

Results are stunning. Bo3b is working on a similar native output for HelixVision, however that will also allow FULL-SBS.

TaB was reversed. Documented that reversing via ini does not work currently. Manually reversing using Virtual Desktop offset the images vertically.

Conclusion thus far:
1. The native output modes are amazing - they just work, without the need for 3DV drivers installed.
2. geo-11 Native half-SBS output to VR virtual screen via Virtual Desktop @ 90fps was jaw dropping, compared to even my 120Hz 300" projector screen.


Rough comparison:

3D Vision display:
-- 60fps max frames are displayed to eyes one after the other
-- 300" screen (usually 27" screen if monitor), far away from the eyes
-- comparatively dark and lacking brilliant colour due to looking through LCD glasses + usually TN panels.
-- Limited to 1080p generally - limited relief available via DSR.

Native SBS output via VR:
-- 90fps displayed to the eyes simultaneously
-- giant screen covering 90 degree+ of FOV right up to the face
-- Bright and vibrant displays of the HMD.
-- Up to the resolution of your VR headset and beyond, using super-sampling.

Note 1: my VR is HP Reverb G2 @ 4k*2k resolution (2160x2160 per eye), which has arguably the best screens currently on the affordable consumer market. If you have an older headset, your results will vary.
I think we can stop worrying about End Of Life 3DVision hardware breaking down - we have a superior alternative solution already - in my humble opinion

Note 2: aside from the DX9 path - iZ3D toyed with the possibility of something called "software page-flipping". It provided a page-flipped signal which ought to be able to be output natively like SBS/TaB is currently, but at full resolution. This will work on 3DV displays without the 3DV driver installed, however left-right images might be swapped if the FPS drops.

The glasses will then need to be activated manually. A while back, there was a post where a poster activated them using the minimised test image, though there are probably better ways/hacks...

[EpsilonLyrae]

Great review post!

I really enjoy the quality and clarity of a virtual screen (using a Valve Index)

But I also really like using my OLED monitor...I'd make a brightness trade off in exchange for not having to wear a full vR headset alone, plus the other improvements that come from not having to look through lenses that introduce chromatic aberration, god rays, etc. Maybe when tech progresses to allow for 240hz+BFI on an OLED screen, the scanout will be fast enough that you can sync some active glasses (3d vision2?) to them with minimal cross talk.

[VR3D]

Thanks for that great comparison of current displays. Sounds like VR hmd's are the best experience for 3d. That is comforting to know since VR is only going to get better. Makes sense since it has no light loss due to glasses as well as 2 actual images displayed simultaneously. Also absolutely no ghosting!

As far as 3d formats being displayed in VR, are any of them full resolution to each eye?

[3DNovice]

Sounds like VR hmd's are the best experience for 3d.

Not necessarily, there are clear arguments were a monitor/projector/tv are better in certain games.

As far as 3d formats being displayed in VR, are any of them full resolution to each eye?

When you say full SBS images do you mean if I set it to 4k in game it will be giving me 2 4k images, one per eye or is it halving the resolution?

That would be two 4K images. Full-sbs is a well defined 3D format, but most 3D TVs use HDMI which requires the half-SBS format. Since we can do whatever we want with VR, we choose image quality. Note that there is no video card available anywhere that is going to be able to drive a full-SBS 4K output at reasonable speeds.

quoted from
https://steamcommunity.com/app/1127310/ ... 883511653/

[RAGEdemon]

Thanks for that great comparison of current displays. Sounds like VR hmd's are the best experience for 3d. That is comforting to know since VR is only going to get better. Makes sense since it has no light loss due to glasses as well as 2 actual images displayed simultaneously. Also absolutely no ghosting!

As far as 3d formats being displayed in VR, are any of them full resolution to each eye?

Not yet, however this doesn't seem to be a hard driver limitation. Bo3b (Bo3bber), the developer of HelixVision and one of the developers of this driver, intend to natively output Full-Resolution SBS via HelixVision - ETA is a few weeks, but no pressure on the good chap

[VR3D]

That's exciting news!

When you say full resolution sbs (and tab also I'm assuming), what does that do? Is it let's say on a 1080p display creating a side by side image of 3840x1080? Then when stretched out it becomes 1920x1080 per eye?

On my 4k tv it's nice to have half resolution tab because the performance is better than full resolution 3d. It's not as sharp but at least it's still kind of native resolution of the panel.

[EpsilonLyrae]

Question about geo-11 specifically:

When you output to SBS or TAB, naturally the pixel count is halved. (ie, if you do SBS rendering at 2560x1440, the perceived end result is two separate 1280x1440 images beinged merged into one 2560x1440 frame). What I'm wondering, is the 3d rendering still processing two full resolution images? (ie, a left and right eye 2560x1440 image is processed, then compressed to 1280x1440 before being combined into the final 2560x1440frame), or is it processing the half resolution for each eye and combining them later?

Asked about a different way maybe to help articulate my point better: Are there notable performance differences between full-width modes (nvidia_dx11, nvidia_dx11, the eventual helixvision supported mode) vs a half-width mode (sbs, tab, interlaced)? I think this would be a good question to have an answer to for users that may be deciding what hardware to ultimately use for their own price/performance/clarity/comfort tradeoffs, or for anyone that currently has one piece of hardware but may be curious to try another (ie, a VR user considering 3d vision glasses)

similar wavelength of the above poster, lol

[Necropants]

Thank's to everyone involved and the rest of you still using and supporting this important technology. Future's looking brighter.

Edited because Masterotaku answered my question on discord

[RAGEdemon]

@VR3D, indeed, your understanding is spot on.

@EpsilonLyrae, I believe I asked this question recently, relating to helifax's VK driver, - I believe it is rendering at full resolution that the game is set to, and then half the pixels are being discarded. This is why HelixVision's proposed Full-SBS mode is exciting, as it natively makes full-use of the full rendered image.

A developer would be the best person to answer this properly - I don't know.

If one creates a custom resolution of 1280x1440 that you could get the game to accept, assuming the game can generate the correct FOV, then that could be used to create a full performance half-SBS output. I don't believe that is happening, however.


----------
Edit: Quick benchmark in Psychonauts 2 I'm playing with the kids:

Scene 1:
DX11 mode: 95fps
SBS mode: 106fps

Scene 2:
DX11 mode: 76fps
SBS mode: 85fps

Slight improvement, probably due to native vs DX11 mode overhead, however, not the expected 100% improvement if the rendering resolution was also being cut in half.

[EpsilonLyrae]

@VR3D, indeed, your understanding is spot on.

@EpsilonLyrae, I believe I asked this question recently, relating to helifax's VK driver, - I believe it is rendering at full resolution that the game is set to, and then half the pixels are being discarded. This is why HelixVision's proposed Full-SBS mode is exciting, as it natively makes full-use of the full rendered image.

A developer would be the best person to answer this properly - I don't know.

If one creates a custom resolution of 1280x1440 that you could get the game to accept, assuming the game can generate the correct FOV, then that could be used to create a full performance half-SBS output. I don't believe that is happening, however.


----------
Edit: Quick benchmark in Psychonauts 2 I'm playing with the kids:

Scene 1:
DX11 mode: 95fps
SBS mode: 106fps

Scene 2:
DX11 mode: 76fps
SBS mode: 85fps

Slight improvement, probably due to native vs DX11 mode overhead, however, not the expected 100% improvement if the rendering resolution was also being cut in half.

Nice. Thanks for trying this.

Could I ask you to also text directx9 mode of geo-11?

Somewhat related to that: I decided to test the DX9 rendering path of geo-11. I have no 3dvision hardware to speak of, but I wanted to see if it would run without crashing.

Relevant specs:

RTX 3070, driver version 516.42
Windows 10, build 21H2 (all current updates installed as of this post)
RTSS on, injecting into the game as my primary frame limiter (I simply find this more convenient to use than the Nvidia driver function)
Stereoscopic 3D ON in the nvidia driver (important, which means you also need this driver installed)

Got some promising results. The fix appears to work about as well in this mode as it did in SBS or another mode. On-screen, it appears in anaglyph 3D mode- which makes it really, really easy to tell if some functions (such as the auto-depth function of the FFXIV fix, for sorting nameplates, so on and so forth) are working correctly. That element is still broken in the DX9 path of geo-11. However, I'm happy to report that this rendering path still appears to work for this game without any errors. I had made the mistake of launching the game in borderless windowed mode, and it didn't position itself correctly on my desktop, leaving a lot of the game cut off. It was also stuck in a quasi-full screen state where the title bar was not visible. Flipping to Full screen, then back to borderless or window mode allowed me to play in the window mode of my choice.

Only observed oddity is that the frame limiter value from RTSS was halved. I have it set to 60fps, the game was running at 30. Increasing the limiter to 120 capped the framerate in-game to 60. This might be how it works in general, can't say as the last time I was using native 3d-vision functions, I had a GTX 1080 and the DX11 path was still functioning. I was also using Helixvision, and the version of 3DFM that's bundled with it took care of sensible frame limiting options for me. Additionally, the hotkeys to change convergence and separation appeared to be hooking directly into the nvidia 3d vision DX9 driver, rather than being interpreted by geo-11 (the UI for the 3d vision driver and geo-11 are very different). This is probably the expected behavior though, as an accommodation of sorts for users that have proper 3d vision hardware with physical dials and buttons to change these settings.

If I had a pair of ye-olde red and blue specs, this would technically be playable, albeit with the usual cross talk you can expect using such a dated method of stereoscopy.

[BigErn77]

I tried Witcher 3 with the nvidia_dx11 output and I get a black screen (but I can hear the music). I have to Ctrl-T to disable 3d vision to be able to see the menu so I can quit the game. (Stereoscopy works fine with normal 3d vision.)

-3d vision enabled in Nvidia control panel
-hardware = emitter+Asus monitor
-Geforce driver: 452.06
-3DFM 3d fix installed for Witcher 3
-Game is set to fullscreen in settings
-I simply copied over the x64 folder contents and set force_stereo = 2 in d3dx.ini

[RAGEdemon]

@EpsilonLyrae, unfortunately, DX9 doesn't work for me in this game. The game starts, flashes black, the minimises. Only way to exit is CTL+F4.