Nvidia 3080 or 3080 TI 3D vision support using modded drivers

[bo3bber]

Gentlemen, I think we have Lift-off, thanks to Vadim and his amazing iZ3D team:


Download links have been sent to Bo3b, DSS, and Helifax.

The developers write:

Hi, this what I found <link>
But I removed madCodeHook library (http://madshi.net/) from it, because we can't distribute their source code. You can rewrite part of code that use it to something like EasyHook https://easyhook.github.io/

I don't know what this event means to us, but I hope it is what we were all hoping for to some degree...

Additional message:
=================
Shahzad:
Got it - Amazing!
Our community developers will likely have questions. Would you mind terribly if they contacted you?
Perhaps you would kindly suggest a preferred communication method you might consider good for you?

iZ3D team:
I can try to answer some small questions. But unfortunately, I won't have much time in the near future. You can write to me here or on Github if you will develop there.

Shahzad:
Gentlemen, Thank you once again; we are forever in your debt. We shall be very careful to not burden you - I think our devs will prefer Github too.
I shall pass along the kind message

My nickname there <xxx>, you can just mention me in issues/discussions

Vadim:
Good 3D luck!!!



Note: I am trying to keep the identities of the iZ3D team private, as revealing them on the open Internet would be unethical/bad form. Github ID sent to our kind devs...

Brilliant! I've downloaded a copy and will look at it later today. The part about replacing the hook library will be easy. We have historically used Deviare for this in 3Dmigoto, but can also use either EasyHook or even Detours now.

It sounds like he'd prefer to not be contacted too much, so we'll make it point to only ask for help if necessary.

Since they don't seem to have a preference, we will most likely add this to my Github alongside 3Dmigoto.

[RAGEdemon]

Cool; makes sense, as you are also one of the most active.

A number of community members were wondering about supported S3D hardware. A quick peruse presented this long list, including native AMD HD3D support:



Interestingly, Z800 and VR920 are HMDs; the progenitors to modern VR headsets.

Perhaps our talented developers will incorporate more in some fashion, such as modern VR via HelixVision, to this already extensive list, perhaps sometime down the line...

[BazzaLB]

Out of curiosity, what do you guys use to determine the parameters required by the Nvidia DLL calls in order to successfully hook them?

[bo3bber]

I've started a new thread for discussions regarding iZ3D: viewtopic.php?f=105&t=25360

Let's move further iz3d specific discussion there, so as to not alter this thread's main purpose.

Also worth noting that even if we get iz3d working with dx11- all 600 of the game fixes we already have won't work there, and would need to be redone. So there is value in continuing to look into whether we can revive 3D Vision.

[bo3bber]

Thank you for your response! Using a 3Dmigoto to inject the dlls sounds really good, have you tried this yet?

It may be necessary to also inject nvd3dum.dll via your loader.

Therefore, I would like to add the following results from further experiments:

- 456.38 nvapi.dll + nvapi64.dll are "compatible" with driver 452.06 without the driver crashing AND 3DV still working in all DX versions (I have also replaced them in system32 and syswow64, too)

--> this means the 456 nvapis still call 3DV "normally" !

- certain parts of the code inside nvwgf2um.dll and x version can be modified without crashing the driver AND 3DV still working (I have inserted large portions of zeros 0000 via hex-editor to test this, they would still work)

--> the dll may allow to "insert" the dead functions to be restored that control 3DV

- 3DV will NOT activate whithout nvd3dum.dll being present, this is true even for DX 10/11 games

--> this suggest the 3DV code is "parted" between nvd3dum.dll and nvwg2fum.dll / nvd3dum.dll and nvwg2fumx.dll
--> we are lucky the 3DV code is still present in 456 nvd3dum.dll
--> the _cfg files (nvd3dum_cfg.dll, nvwgf2um_cfg.dll) are apparently not involved with 3DV

- inside the dll, I noticed lines with "3D3D3D3D3D3D3D3D3D" --> probably some sort of marker, probably to highlight 3DV code?

Considering the above results, I believe the default 3DV route for stereo_api_calls works like this:

Code: Select all

   
     
         calls          calls            calls
nvapi.dll --> nvd3dum.dll --> nvwgf2um.dll --> NvSCPAPI.dll --> enables 3DV
                               |
                               |
                              here the chain brakes in 456, therefore NvSCPAPI.dll will not be called 

I will try now to debug the dlls to find the exact functions controlling the activation for 3DV. Also, I will continue research to the redirection / dll forwarding approach we have discussed earlier.

I have not had a chance to try this approach yet, but your additional experiment here suggests this has a chance of success. Being able to drop in the 456.38 nvapi into a working 456.02 definitely says the first part of the chain is still working. I can hopefully take a look here in the next week.

I think your call path description there is pretty good, and I did not know it needed nvd3dum.dll as well. Good to know.

Out of curiosity, what do you guys use to determine the parameters required by the Nvidia DLL calls in order to successfully hook them?

Not sure I understand this question, but the calls are all wrapped in 3Dmigoto using the nvapi.dll that we ship with it. We have the API specification because that's what nvidia ships with their nvapi.h file, so all parameters there are known.

There are some secret calls, and the interface to nvd3dum.dll will be totally unknown in terms of interface, but with C++ we can just specify a typedef that has the parameters that we can see from assembly, so it can call through at least and not impact the inputs.

[skyrimer]

Incredible work RAGEDemon, now let's hope that our shaderhackers can build a working solution with this, but things are looking bright for once in a long time around here.

[BazzaLB]

Out of curiosity, what do you guys use to determine the parameters required by the Nvidia DLL calls in order to successfully hook them?

Not sure I understand this question, but the calls are all wrapped in 3Dmigoto using the nvapi.dll that we ship with it. We have the API specification because that's what nvidia ships with their nvapi.h file, so all parameters there are known.

There are some secret calls, and the interface to nvd3dum.dll will be totally unknown in terms of interface, but with C++ we can just specify a typedef that has the parameters that we can see from assembly, so it can call through at least and not impact the inputs.

You did indeed understand the question. You disassemble. Thanks for the reply

[skyrimer]

There's a new topic on the driver subforum for iz3d, I think we should keep this thread for the nvidia drivers and move all iz3d conversations there.

[skyrimer]

There's a new topic on the driver subforum for iz3d, I think we should keep this thread for the nvidia drivers and move all iz3d conversations there.

[jurandyrafael]

Whats the use of the new driver if we redirect it to the old driver dll???

Because, as we now have found out together, only the previous versions of nvwgf2umx.dll contain the stereo activation code required to activate NvSCPAPI.dll and NvSCPAPI64.dll.

So theoretically it should be possible to make the new driver have both versions of nvwgf2umx.dll at the same time - the new 456.38 one for regular, "non-stereo" calls - the second one from the previous driver 452 just for stereo calls.

Following this approach, I was able to make redirections on 456.38 from nvd3dum.dll --> NvSCPAPI.dll --> abcdefg.dll (previous, but renamed version of nvd3dum.dll) for testing purposes and it worked like a charm (manipulate calls coming from NvSCPAPI.dll).

When I then deleted the abcdefg.dll (while at the same time having the nvd3dum.dll still in the folder), the game still worked normally but without 3DV. Pasting the abcdefg.dll in again, stereo kicked back in. Deleting it again, stereo was off again. This proves it is possible to make redirections from NvSCPAPI.dll --> abcdefg.dll

This clearly only worked because on 456.38, nvd3dum.dll still contains the stereo activation code whereas nvwgf2umx.dll does not, as bo3bber already told us before.

But luckily, this is not true for nvapi64.dll (As far as we have found out yet).

So IMHO, stereo activation with DX11 will work, if (and thats a big if) we will be able to redirect the stereo activation calls (and only them) from the highest instance (this should be nvapi64.dll? ) to a previous nvwgf2umx.dll while at the same time leave the "non-stereo"-calls to go to the current nvwgf2umx.dll.

Is this route correct?:

usual, non-stereo calls: nvapi64.dll ? --> 456.38 nvwgf2umx.dll

stereo-calls only: nvapi64.dll ? --> 452.06 nvwgf2umx.dll --> NvSCPAPI64.dll --> 452.06 nvwgf2umx.dll (or is it the other way around, I am not sure yet)

Question: What file(s) make the call to NvSCPAPI.dll and NvSCPAPI64.dll other than nvd3dum(x).dll and nvwgf2um(x).dll?

I would like to point out that I am not an software engineer, but I will give my best to help you to solve this issue. If you have any documentation on the nvidia Stereo API available, I would be grateful if you could share it with me.

Also, if I am wrong with any of the above, please correct me. These are only theoretical assumptions based on the research I did so far.

I was only able to find this about API documentation, but you probably have already seen it:

Stereoscopic 3D APIs - NVIDIA Developer Documentation - https://docs.nvidia.com/gameworks/conte ... eoapi.html

[AlexTOPMAN]

khm... anybody home? ...

[skyrimer]

Not looking good by the lack of updates....

[BazzaLB]

I think bo3bber is looking at redirecting calls to an older dll but this stuff takes time, especially as it involves reverse engineering and I am sure he makes no promises. This stuff aint easy. Patience young padawin.

[DugomFirst]

Error

[TonyMontana]

These are two interesting documents from the developer documentation that could be useful for our purpose:


http://developer.download.nvidia.com/to ... public.pdf

http://developer.download.nvidia.com/wh ... _Guide.pdf

These statements in particular could be useful for further investigations:

"NVAPI is forwards and backwards compatible." --> If this really is true, nvapi should call through even older driver versions.


"NVIDIA suggests that developers use anaglyph mode as often as possible, both in daily
development and for QA purposes. This will reduce the resource contention and
specialty hardware requirements for your testing efforts without compromising on
coverage.
3D Vision Discover mode (Red/Blue glasses and a default driver install) uses the same
code paths as 3D Vision Automatic, except for the final step. In 3D Vision Automatic
mode, the final step is to present the image to each eye, while in 3D Vision Discover
mode the final step is to composite the two images into a single Red/Blue image pair."

So if we were to find a skilled reverse engineer, they would not necessarily need to have access to a full 3DV hardware-setup, instead it would be sufficient to make anaglyph mode working - maybe this could speed up the testing process a little? (Also for anyone who wants to set up a secondary system for testing)

I am pretty sure that an experienced group like CPY / CODEX could patch the dll / create a loader for injecting and re-routing - maybe we should contact them?

[russellk]

These are two interesting documents from the developer documentation that could be useful for our purpose:


http://developer.download.nvidia.com/to ... public.pdf

http://developer.download.nvidia.com/wh ... _Guide.pdf

These statements in particular could be useful for further investigations:

"NVAPI is forwards and backwards compatible." --> If this really is true, nvapi should call through even older driver versions.


"NVIDIA suggests that developers use anaglyph mode as often as possible, both in daily
development and for QA purposes. This will reduce the resource contention and
specialty hardware requirements for your testing efforts without compromising on
coverage.
3D Vision Discover mode (Red/Blue glasses and a default driver install) uses the same
code paths as 3D Vision Automatic, except for the final step. In 3D Vision Automatic
mode, the final step is to present the image to each eye, while in 3D Vision Discover
mode the final step is to composite the two images into a single Red/Blue image pair."

So if we were to find a skilled reverse engineer, they would not necessarily need to have access to a full 3DV hardware-setup, instead it would be sufficient to make anaglyph mode working - maybe this could speed up the testing process a little? (Also for anyone who wants to set up a secondary system for testing)

I am pretty sure that an experienced group like CPY / CODEX could patch the dll to make it work again - maybe we should contact them?

Good effort but I have a feeling that is already known by people like bo3bber. I seem to remember that this is the approach he already uses to get Helixvision working.

Also, I'm not a developer, but this is not just a case of hacking or patching. I believe the code has been completely removed and there are compatibility issues to worry about.

Last but not least, documentation is often wrong or out of date, so those statements you've found could turn out to be completely academic.

Sorry if I'm being negative, I could be wrong on all these things, it's just my opinion.

[TonyMontana]

There's no need to be sorry - in fact you are absolutely right that they stripped out the code from 456+ drivers, hence the api call redirection approach discussed earlier in this thread.

It is actually not my intention to say that it could be "easily" patched / hacked but still I would not exclude it entirely from the list of possibilities.

As a matter of fact, I was able to patch an earlier version of nvwgf2um.dll in such a way it would work with a higher driver version. Apparently, the driver checks the exact byte length of the dlls (among other things) to create internal checksums. If the patched dll does not match the exact number of bytes present in the original, it will not work.

On top of that, bo3bber had the brilliant idea to design a loader that would hook an older version (e.g 452) of nvwgf2um(x).dll directly into a given game to re-route the stereo calls there.

Experiments suggest that the first part of the route is likely still working, now the aim is to re-route the calls from 456 nvapi.dll to 452 dlls.

Default stereo call path description for driver versions 452 and earlier:

Code: Select all

                stereo calls             stereo calls                 stereo calls
452 nvapi(64).dll -----> 452 nvd3dum(x).dll ------> 452 nvwgf2um(x).dll ------> 425.31 NvSCPAPI(64).dll ------> enables 3DV
                                                         |
                                                         |
                                                        here the chain brakes in 456, therefore NvSCPAPI.dll can not be called 

Custom stereo call path description for custom 3Dmigoto / dll loader on driver versions 456+:

Code: Select all

               stereo calls                     stereo calls                  stereo calls
456 nvapi(64).dll ------> 452 or 56 nvd3dum(x).dll ------> 452 nvwgf2um(x).dll ------> 425.31 NvSCPAPI(64).dll ------> enables 3DV
         |						     |				
         |						     |				
         ------>-------re-routed call path ------->-----------
                            
 

Note that both call paths are for stereo api calls only, so all standard calls could continue to use the default 456 driver route.

[bo3bber]

These statements in particular could be useful for further investigations:

"NVAPI is forwards and backwards compatible." --> If this really is true, nvapi should call through even older driver versions.

Good reference, thanks for looking into this.

I've read all their docs and we use most of that, but I had forgotten that they commit to forwards and backwards compatible APIs.

It's a bit of a lie of course- they aren't backwards compatible now, they killed the stereo APIs. But the forwards compatible part means that they will at least leave the APIs in place. So in principle if we can get this to work, it has a chance to continue working.

So if we were to find a skilled reverse engineer, they would not necessarily need to have access to a full 3DV hardware-setup, instead it would be sufficient to make anaglyph mode working - maybe this could speed up the testing process a little? (Also for anyone who wants to set up a secondary system for testing)

We setup HelixVision to use Discover mode, because most VR users won't have 3D Vision hardware available. So in general, discover mode is what I target. I convert the image from the buffers before it gets colorized.

I am pretty sure that an experienced group like CPY / CODEX could patch the dll / create a loader for injecting and re-routing - maybe we should contact them?

Interesting idea. I have no idea how to contact them, but if someone is motivated, the modding groups have the chops to make this work, no question. No money or fame in it though, not sure they'd spend the time.

[bo3bber]

Random thought of the day: Anybody have any TriDef contacts?

Given the raging success for contacting iZ3D, if anyone has contact info for TriDef people of any form, please contact them.

[wrecked]

where did everyone go?.

[projektion]

so how does it look like? Is 3D Vision dead for RTX 3xxx?

[whyme466]

@bo3b - if anyone can find current email address for Chris Yewdall (former CEO of TriDef), I would be willing to contact him. When I was active Ignition user, we exchanged several messages, but I only have his old ddd.com email address. He may still be on LinkedIn and Twitter (Professor3D, I believe), but I currently have no access to either platform...

[oneofthe8devilz]

I totally support the notion to contact former Tridef staff members in the pursuit to eventually acquire the source code to that amazing software.

The benefits of that approach:

1. It works with NVIDIA AND AMD GPUs
2. It already supports native SBS output
3. It already supports auto-convergence
4. Unlike many 3DVision users falsely assume, Tridef actually DOES support BOTH "Geometry" 3D generation and "Z-Buffer based" 3D generation
5. It already supports many Directx 9,10 and 11 games

[Lysander]

so how does it look like? Is 3D Vision dead for RTX 3xxx?

The gurus are looking at it, they JUST obtained iZ3D code and new drivers and cards came out like what, 2 months ago? This is hard stuff that takes a long time if possible at all. From following this site what I gather is that there is optimism, but it's just about time. So we need to enjoy existing games, support these guys and have patience - I'm pretty sure everyone here really wants this to happen but only a few have the skills to make it so.

[Necropants]

Yeah it's not like we have the weight of the gaming world behind us on this. Have some patience and cross your fingers. All I know is nvidia have given me a good reason at this point not to buy one of their cards again and AMD sounds like they might even get competitive this gen.

[projektion]

Yes, I also think, keep my old RTX and go for AMD from now on!
NVIDIA has so much money but they dont spend it on their customers!
For NVidia there wouldnt be any problem to keep 3DV in their drivers, but they decided to abonden it.
So there is no need for NVIDIA anymore!

[Necropants]

Worse than abandon, they worked to actively strip it. May not be anything malicious. But they refused to make it open-source, so I wouldn't put it past them either...

[Lysander]

Yeah, I automatically went NV because of 3DV, if we get a solution which will let me use the equipment with AMD or if the whole 3D dies, I'll pick the best solution available when time comes to upgrade. It'll be a while though, as my backlog is huge and new games continue getting fixed

[Chip23]

As people mentioned before, we can't do over 1440p in 3d, and 3080 is barely 15-20% faster at that resolution, so it wouldn't be a big improvement anyway.

How can i do over 1440p in 3D? Monitor Model? Or TV or Projector?

[john105]

As people mentioned before, we can't do over 1440p in 3d, and 3080 is barely 15-20% faster at that resolution, so it wouldn't be a big improvement anyway.

How can i do over 1440p in 3D? Monitor Model? Or TV or Projector?

LG 4K 3D TVs 2016 or older models for example. But I personally think that 4K 3D is not worth it anyway because of performance.

[Chtiblue]

How can i do over 1440p in 3D? Monitor Model? Or TV or Projector?
[/quote]
LG 4K 3D TVs 2016 or older models for example. But I personally think that 4K 3D is not worth it anyway because of performance.
[/quote]

Wrong because thanks of upscale or SBS or T&B mode you can play at 1440p (or custom res like 1620p) without noticing quality decrease (on my LG C6V 1440p and above looks the same than 2160p) but with huge performance boost

[floph]

Wrong because thanks of upscale or SBS or T&B mode you can play at 1440p (or custom res like 1620p) without noticing quality decrease (on my LG C6V 1440p and above looks the same than 2160p) but with huge performance boost

Is it more immersive a 55" LG C6V TV than a 150" 3D projector ?
I only tried 3D monitors up to 24" and in terms of immersion my personal opinion is that it doesn't even compare with a 150" diagonal 3D projector. With a 24" it's like looking through a small window, while with a 150" projected image it's like a portal is open and you can go through.

[john105]

Wrong because thanks of upscale or SBS or T&B mode you can play at 1440p (or custom res like 1620p) without noticing quality decrease (on my LG C6V 1440p and above looks the same than 2160p) but with huge performance boost

You are saying that I'm wrong about 4K performance because you play at 1440p or another non-4K custom resolution? Ok, it makes a lot of sense. lol

[Chtiblue]

You are saying that I'm wrong about 4K performance because you play at 1440p

Yep this, thanks to SBS or T&B Tv mode

[Guig2000]

Random thought of the day: Anybody have any TriDef contacts?

Given the raging success for contacting iZ3D, if anyone has contact info for TriDef people of any form, please contact them.

@bo3b - if anyone can find current email address for Chris Yewdall (former CEO of TriDef), I would be willing to contact him. When I was active Ignition user, we exchanged several messages, but I only have his old ddd.com email address. He may still be on LinkedIn and Twitter (Professor3D, I believe), but I currently have no access to either platform...

Maybe Neil have it?
https://www.mtbs3d.com/articles/news/12 ... again.html

[Chip23]

As people mentioned before, we can't do over 1440p in 3d, and 3080 is barely 15-20% faster at that resolution, so it wouldn't be a big improvement anyway.

How can i do over 1440p in 3D? Monitor Model? Or TV or Projector?

LG 4K 3D TVs 2016 or older models for example. But I personally think that 4K 3D is not worth it anyway because of performance.

It's passive 3D on LG TV. Artifitialy polirized light. With no convergence ajusment.
I mean REAL 3d - active - with shutter glasses -- 3D vision and monitor.

[Chip23]

Wrong because thanks of upscale or SBS or T&B mode you can play at 1440p (or custom res like 1620p) without noticing quality decrease (on my LG C6V 1440p and above looks the same than 2160p) but with huge performance boost

Is it more immersive a 55" LG C6V TV than a 150" 3D projector ?
I only tried 3D monitors up to 24" and in terms of immersion my personal opinion is that it doesn't even compare with a 150" diagonal 3D projector. With a 24" it's like looking through a small window, while with a 150" projected image it's like a portal is open and you can go through.

I have 4k Projector with 5M wide screen - a wall in leavin room )
This one SONY VPL-VW570ES. But 3D only FHD and 30Hz

[floph]

Is it more immersive a 55" LG C6V TV than a 150" 3D projector ?
I only tried 3D monitors up to 24" and in terms of immersion my personal opinion is that it doesn't even compare with a 150" diagonal 3D projector. With a 24" it's like looking through a small window, while with a 150" projected image it's like a portal is open and you can go through.

I have 4k Projector with 5M wide screen - a wall in leavin room )
This one SONY VPL-VW570ES. But 3D only FHD and 30Hz

I have the Epson 5030UB 3D projector. It works just fine in FHD at 24 Hz because it has a nice feature - Frame interpolation. It makes it look like 60 Hz.

So have you tried a 55" LG C6V TV to compare which one is more immersive ?

[Chtiblue]

It's passive 3D on LG TV. Artifitialy polirized light. With no convergence ajusment.

???? No convergence adjustment????

Serious misconception here, passive 3D is real 3D too with of course Both convergence & depth adjustment.

[whyme466]

@bo3b - I resurrected my dormant LinkedIn account, and I sent a message to the old TriDef CEO describing the current state of 3D Vision, referenced the incredible helixmod.blogspot library and HelixVision, described how iZ3D has graciously helped this community, and asked him about sharing the old Ignition code. I will share his response...

Update:

Per Chris Yewdall, former DDD CEO (including TriDef Ignition), "Unfortunately the code was owned by the Australian subsidiaries of DDD and the liquidators of those business completed the wind up of both those entities earlier this year so the code no longer exists."