This isn't an announcement (yet). It's a request we need to be making strongly to Microsoft.

The lack of Stereoscopic 3D support at the operating system level is getting beyond a joke. By the time Direct3D 11 games become mainstream, Stereoscopic 3D monitors will be mainstream too. With the existence of monitors like iZ3D and Zalman's that can compete directly with 2D monitors, Direct3D needs to change.

MTBS3D needs to contact all the 3D hardware developers, and get them to send a joint submission to Microsoft with their requirements. Hardware manufacturers do this all the time, it is a perfectly normal thing to do, and Microsoft understands and expects that, when developing their Direct3D API. I assume Microsoft are in constant contact with hardware developers like Nvidia and ATI about their requirements, and also in constant contact with game developers about their requirements. It's time we start being more vocal about our requirements.

I mean, if several companies were manufacturing monitors with a certain resolution, and Direct3D didn't provide a way to create surfaces with that resolution, then the monitor manufacturer would quite legitimately be writing to Microsoft about their requirements, and not accepting any suggestion that it is OK for games to just use part of the monitor.

So what are our requirements?

Well, our main requirement is that we expect ALL Direct3D 11 games/applications to work in stereo on a stereoscopic monitor, and in mono on a monoscopic monitor, regardless of what game/application it is, and regardless of what graphics card they have.

Since all Direct3D 10 and above features are compulsory, that means either a graphics card/driver is Direct3D 11.1 or it isn't. Meaning that if we get stereoscopic support added as a Direct3D 11.1 feature, then it will work on ALL Direct3D 11.1 graphics cards. I anticipate that there would not need to be any hardware changes since Direct3D 10 already allows rendering to multiple buffers at the same time, it would just be a driver-level change. People buying a Direct3D 11.1 graphics card would just know that it would work in stereo.

Getting it to work on ALL games, means that when a Direct3D 11.1 game/application creates a swapchain it receives either 2 buffers or 1 buffer, depending on the kind of monitor, and has no choice in the matter. If the game is informed that it must render to two buffers then it will render to 2 buffers, and if it is informed that it must render to one buffer, then it will render to 1 buffer. That way all games would be stereoscopic 3D, unless they actively go out of their way not to be.

This is not an unfair demand to place on game developers, since they already have to start learning again from scratch how to target Direct3D 11, and the only additional requirement would be adding a loop and learning how to create an off-axis projection matrix (Although DXUT could easily handle that for them). Game developers would not have to learn anything about the various 3D formats, since that would all be handled by the Direct3D driver for them, they would just render the same scene to two different render targets from different viewports, and Direct3D takes care of the rest.

Getting it to work on ALL monitors would be only slightly trickier. Microsoft would have to make Direct3D call another driver made by the monitor manufacturer to get a pixel shader to either convert the left and right buffer to a single buffer (eg Zalman, 3D DLP) or a buffer for each VGA port (eg iZ3D, TDVision). Or it would tell it to swap buffers every frame, setting the VGA pin appropriately.

We also require a Direct3D control panel option to turn stereoscopic 3D on or off, and to choose between kinds of stereoscopic 3D that don't require a special monitor (eg. anaglyph).

I say Direct3D 11.1, but that is just a guess. Ideally it would be in Direct3D 11.0. Even more ideally, Microsoft would also produce a Direct3D 10.2 specification that adds the Stereoscopic requirement, for those cards that can't support Direct3D 11.

We really need to get started on this campaign... The continual hacks that we are used to to get 3D to work are not the way of the future. These hacks will still be needed for current games and hardware of course, but once Direct3D 11 becomes mainstream, that will be a thing of the past, as they rightly should be.

This campaign potentially gets us everything we want, at little or no cost.

And by getting Microsoft on board, the other important result this campaign is likely to bring is S3D on a future XBox. And where the XBox goes the Playstation and Wii will follow.

EDIT: I should add that OpenGL ALREADY has this feature, but because it is not compulsory, and because OpenGL was designed for business applications rather than games, graphics card manufacturers only implemented it on their most expensive cards targetted at businesses. In Direct3D 11, it would be compulsory and all Direct3D 11 cards would have to support it.

If you know of any additional requirements that we have, or anyone wants to have a go at drawing up a draft specification of our requirements, then please post your comments.

And let me know if you have any contacts at Microsoft that we can pitch this to. We also need to contact all the 3D monitor manufacturers to see if they have any special requirements, and to get them to sign on to a joint submission.

EDIT: The monitor-manufacturer-supplied driver would also handle converting from the game's resolution to the screen's native resolution, where resolution is an issue. Of course, it is better if the game lets the user choose the resolution they want to run in, but it wouldn't matter if the game only works in a certain resolution.
Note that the monitor driver is not a 3D driver in the sense you are used to, all it does is take a left and right view and convert them into a single view or the views needed for each VGA port, by supplying a pixel shader.

There are a couple of my personal dream requirements that I deliberately didn't mention. Firstly, not all 3D monitors work in the 2-image way. Some autostereoscopic monitors have up to 9 views. At least one uses a mono image + depth buffer and generates its own views. And some people are not content with a single monitor, and want a VR CAVE. Some people have a head-mounted display with tracker. In the future games will have to render to whatever surface they are told to render to, with whatever projection matrix they are told to use, and will have to deal with inputs from tracked input devices. But there are limits to what we can achieve on a first campaign, and home VR is not at an advanced enough stage for Microsoft to get it right. For now we should pretend that multi-view monitors don't exist, that people only use one monitor at a time, and that they haven't got a head-tracker set up.

Now that'd be awesome. Where do I sign the petition?

We have to create one first.

Unfortunately I don't actually know anything about Direct3D 11 yet, so we need to research what they have planned. We've already left it a bit late, so we might not be able to push them to give us any S3D support until D3D 12. But it is still worth pushing them.

EDIT: I should mention that this is by far the easiest feature of Direct3D 11 for Microsoft to implement, AND it has by far the biggest improvement in impressive graphics.

The Senior Lead Program Manager for Direct3D has (I think) this personal email account:

allison.klein@gmail.com

Assuming she is the Allison Wynne Klein from Princeton who worked for Microsoft Research or computer graphics, and assuming she still checks her gmail account.

I haven't found an official email address yet though.

I don't mean to be a party pooper but since when has microsoft listened to anyone?

EDIT: upon further contemplation I think that every bit of community pressure helps, and that if you don't speak out they are even less likely to introduce it. MTBS needs more people like you, well done!

I am already developing a very exciting initiative with some game developers that can help set the groundwork for this. It's too early to discuss it, but it will require all of your participation when the time is right. I apologize if I am being too vague, but as usual, there is a method to my madness. Please don't misread or over-hype these remarks.

Regards,
Neil

I agree completely. Stereo3d rendering needs to be part of DirectX. I had done some research into S3D drivers and the necessary methods are all way too hackish. There needs to be some standards and who better than Micorsoft to implement them into the next version DirectX. I totally support this effort.

fun fact about microsoft: it was actually AGAINST the creation of the internet or anything like it. It was thanks to about 4 visionary individuals that it was created. See any parallels?

Plus one.

I humbly request that your (legitimately) rambling post be divided into coherent sub-sections, PM if you want me to help (I am too busy with lunch ATM to submit a proposal. )

I agree with the idea, will bring up points later when I can read the whole deal.

Multi-output standards are a must, as we are at the point where a $200 PC can have 6 outputs it is getting ridiculous that DX kills all activity when you have a full-screen game on one monitor.


Edit: I agree on all fronts, but I think that if the push comes from nVidia and AMD, (maybe even Intel), then Microsoft will have no reason not to support it, they need hardware after all.

I know nVidia is onboard with 3D, but they are trying to pimp their cards with it, so I don't know if they are into "standards" that take their 'gimmick' away.

AMD should get onboard because they already have dual output support on Crossfire, that covers 95% of the stereo display types available.

Am I correct in thinking that just about any card from the last 5 years could easily support this with a new driver?

They listen to MAC and Linux... Look at Vista...


@ Topic:

The time will come for an DX sided support of S3D, I think DX11 would be an realistic point. I think DX11 will come with Windows7.

Ill do whatever is necessary that I am able to do to help out. Frankly, I'm surprised that it isn't already supported.

Microsoft actually tried something before. I think it was in DirectX 7 or before. It was removed when they released DirectX 8.
There are remnants of it in the documentation, see here.

That's pretty interesting.

Wow, almost 2 years have passed and the train is still there.


Unfortunately, MRT suffers from multiple quality limitations, its main use is rough shadows and reflections. Current hardware does not support truely orthogonal two-view rendering, that is with every single feature enabled.

Direct3D 11 is a strict superset of Direct3D 10/10.1, so actually any new strereo capabilities could be an extention to 10/10.1/11 API.

Well I think if Sony and Nintendo has success in the 3D space Microsoft may finally wake up to it. DirectX12?

https://www.cmpevents.com/GDCAN10/a.asp ... &id=454031
http://www.gamasutra.com/view/news/2843 ... ing_3D.php

This is more console related, but I think it's pretty cool!

Regards,
Neil

Zargapour may be excited about stereo3D gaming, but he's basically a game studio executive within a large coproration that has not shown any commitment to supporting stereo 3d as of now. Yes, game studios and game developers are the most enthusiastic about stereo 3D gaming because they understand the potential, but we need Direct3D development people and hardware development people wake up to the trend.

Saying that everyone can "shift four inches and render again" is not enough - the graphics hardware is always get used to the last bit even in mono, so brute force approach is not really optimal, especially given that in 5-10 or so years silicon manufacturing will reach a practical limit for optical lithograpy and graphics cards will stop doubling in performance every year.

in about 6 years Intel will be manufacturing in 11nm, after that the lithography can't go smaller (thats what a Intel guy told me on the Cebit).
Let's see what will be done until then

Well, ITRS roadmaps talk about year 2022, but yes, 11 nm is the practical end of current optical photolithography (and probably Moore's Law as well), and still there is no guarantee that the industry will be able to reach this target in volume production even with next-generation lithography methods and different materials. (However the end of Moore's Law has been predicted numerous times in the last 30 years, and the law still holds).