M$ shows G800... in DX9 demos... [Archive]

Idiot

22-Mar-2002, 14:43

http://www.watch.impress.co.jp/pc/docs/2002/0322/dx9.htm

3 videos

showing displacement mappings...

http://www.watch.impress.co.jp/pc/docs/2002/0322/video1.mpg
http://www.watch.impress.co.jp/pc/docs/2002/0322/video2.mpg
http://www.watch.impress.co.jp/pc/docs/2002/0322/video3.mpg

and something about 40bit colours

10 bit each RGB 10bit Alpha.

http://www.watch.impress.co.jp/pc/docs/2002/0322/dx9_2.jpg

http://www.watch.impress.co.jp/pc/docs/2002/0322/dx9_3.jpg

nAo

22-Mar-2002, 15:15

Darn..I can't read chinese :-?
Btw..VS2.0 and PS2.0 has slightly changed according to a previous MS power point presentation.
Temporary registers were scaled down from 16 to 12 in VS and PS both. In PS2.0 texture iterators were dual ported, now are not.
And in VS2.0 model there are 'new' registers, as loop and boolean registers. Obviously them should account for the increased flow control capabilities in this vertex shader revision.

ciao,
Marco

Idiot

22-Mar-2002, 15:47

First its Japanese not Chinese

Second, I guess maybe MS is going for the common denominator between the 3/2 possible early DX9 players. The old spec will probabily be what R300 or NV30 end up having, at least part of them.

the mountain thing with clouds/fog is very impressive IHMO.

nAo

22-Mar-2002, 16:20

First its Japanese not Chinese
Sorry :( I know it is Japanese, it was a freudian lapsus.

Second, I guess maybe MS is going for the common denominator between the 3/2 possible early DX9 players. The old spec will probabily be what R300 or NV30 end up having, at least part of them.
Yeah..that's the same thought I had

the mountain thing with clouds/fog is very impressive IHMO.
Can U see it? on my video player is a bit too dark to distinguish different features. I put my hope in better videos..

ciao,
Marco

nggalai

22-Mar-2002, 16:33

Needless to say, I was rather impressed by those video clips. Thanks for the links, Idiot (no offense meant . . . ;) ). Especially video 3 . . . wow. Impressive framerate, at first glance.

Question: how did you come to the G800 conclusion, Idiot? I can't read Japanese, either, so, eh. :)

Hmm. Yes, the shader specs have changed somewhat since the RFC docs that were being distributed. I hope they'll provide the updated specs to the public, soon. Might be interesting.

Again, thanks!

ta,
.rb
________
silversurfer reviews (http://vaporizer.org/reviews/silver-surfer)

pabst

22-Mar-2002, 17:47

He probably came to the conclusion that it was the G800 since it mentions Matrox all over the place and even has the Matrox logo on two of the videos. Unfortunately, I don't speak Japanese and the Alta Vista translantion is practically unintelligible. It does appear towards the end that someone asks if it was done on specialized hardware and the response seems to be that it was some sort of software emulation. No one around here speaks Japanese ???

nggalai

22-Mar-2002, 18:12

Hi there,

(rather bad) English translation: http://forums.matroxusers.com/showthread.php?s=&threadid=32006 and http://forums.matroxusers.com/showthread.php?s=&threadid=32007 .

Seems to have been software emulated. Too bad.

ta,
.rb
________
starcraft replays (http://screplays.com/replays)

Doomtrooper

22-Mar-2002, 18:41

Very cool :D

Nappe1

22-Mar-2002, 19:13

hmmh...

Idiot: did they mention "G800" anywhere?? because G550 really is what is left from original G800 project. And this new chip was codenamed as "parhelia" while the original G800 was known as "Condor II" and was heavily based on G400 core.

Idiot

22-Mar-2002, 19:24

I think you can see the "G800" in their article everywhere, even if you are using English version of IE.

......

other than that what platform did they run the "software" emulation on exactly??? The FPS is pretty good, and i won't mind if pure software rendering on my PC can achieve that kind of effect. In those cases I could just wait until the next "great" graphics chipset.....

muted

22-Mar-2002, 19:48

Well in the last video you can hear the guy say , on matrox displacement mapping hardware

they also mention that DirectX 9 will go into beta stage this summer and will have a final release in 2003 ( i was hoping for an earlier release date )

the first video , reminds me of what JC ( no, not jesus christ ) wants to do for doom3 , very impressive for an originaly "low" polygon model

don't tell me it isn't ..

Oh yeah , more of matrox on saturday

Dave Baumann

22-Mar-2002, 19:54

they also mention that DirectX 9 will go into beta stage this summer and will have a final release in 2003 ( i was hoping for an earlier release date )

Take a look at this report (http://www.theregister.co.uk/content/4/24534.html) from the register:

Exclusive: Microsoft has reshuffled its roadmaps once again, and begun briefing partners and customers to expect an interim upgrade to Windows XP, dubbed XP "SE", in the first quarter of next year.

This buys more time for the Longhorn team to complete the complex task of implementing a native database file store, which Jon Honeyball first revealed to the world at The Register here, last August. In January we exclusively confirmed that the native, SQL Server-derived database would go in Longhorn, with the Blackcomb release - originally earmarked for the transition - pushed out to 2004.

XP SE will be principally a consolidation release. Candidates for inclusion are the essential .NET client-side plumbing: the common language runtime (CLR), Internet Explorer 7.0 and DirectX 9.0, and a mature Bluetooth stack. Microsoft has blown hot and cold on Bluetooth - mostly cold, actually - but with chipsets at the $5 mark, it's going to be ubiquitious in PDAs and phones by 2003.

Microsoft has more to offer to offer in the consumer department, (and DirectX is essentially a consumer technology), and as we observed this week, the Mira tablet hinges on a multi-user, RDP-enabled XP acting as the server. Windows XP consumer edition doesn't come with XP, so this is one obvious area for improvement.

Both of these things seem to tally, it also goes with earlier comments that MS want to make DX revisions a â€˜biggerâ€™ thing.

However, you have to question where this puts hardware that rumoured to be close and of DX9 spec â€“ would they release an â€˜over featuredâ€™ card with no API, or will MS do another interim DX8 release?

Laa-Yosh

22-Mar-2002, 20:11

the first video , reminds me of what JC ( no, not jesus christ ) wants to do for doom3 , very impressive for an originaly "low" polygon model

Nope, what Carmack does is very different - the trick is to generate bump maps for Dot3 lighting from a high poly model.
It does not affect geometry at all, so it needs a fairly detailed model for nice silhouettes (2-3000 faces IMHO).

muted

22-Mar-2002, 20:15

don't tell me it's not similar , he'd be better off with displacement mapping , i think he should look into that .... i hate seeing features unused

Laa-Yosh

22-Mar-2002, 20:20

The idea - moving geometry detail into a texture - is similar. However Carmack is most likely avoiding high poly counts in Doom3.

Using displacement would either result in incorrect shadows (when generating shadow volumes using the un-tesselated model) or cause a serious slowdown (for correct displacement mapping, poly counts would increase at least ten times). It's exactly the same reasons that stop him from using Truform AFAIK.

LeStoffer

22-Mar-2002, 20:36

they also mention that DirectX 9 will go into beta stage this summer and will have a final release in 2003

We need it get this info verified as I have a hard time believing a final release in 2003 with R300 coming out this summer. But something fishy is going on here alright...

However, you have to question where this puts hardware that rumoured to be close and of DX9 spec â€“ would they release an â€˜over featuredâ€™ card with no API, or will MS do another interim DX8 release?

Indeed. If MS do let DX9 slip into next year hardware vendors and game developers would have to shift focus to OpenGL to support the newest features. It could turn out to be a huge gamble on MS' part when you think about the fact that OpenGL has picked up pace with its own development...

As I said: Something fishy is going on alright... :o

Regards LeStoffer

muted

22-Mar-2002, 20:45

yes .. but doesn't MS own opengl... :?

LeStoffer

22-Mar-2002, 21:12

yes .. but doesn't MS own opengl... :?

:D

Oh, in case you weren't making a joke:

An independent consortium, the OpenGL Architecture Review Board, guides the OpenGL specification. With broad industry support, OpenGL is the only truly open, vendor-neutral, multiplatform graphics standard.

Regards, LeStoffer

muted

22-Mar-2002, 22:41

I just know they bought a big chunk of it , but i don't know more

and , did anyone hear anything about the R300 supporting Displacement mapping ?

more is better

Humus

23-Mar-2002, 11:31

MS bought some patents from SGI for a lot of money some time ago, but what they bought never really got published. Lot of people interpreted it as though MS now owns OpenGL, but it would be right out stupid for SGI to sell the rights to it.

Rookie

23-Mar-2002, 11:43

I now doubt who will lauche fully compiant DX9 board first? :-?

duffer

23-Mar-2002, 13:15

One big piece of unfinished DirectX business that Microsoft has promised to deliver soon is a full .Net binding to DirectX. (See Phil Taylor's recent emails in the DirectXDev mailing list.)

Perhaps they will call this .Net binding DirectX 9, and then call the new hardware API DirectX 9.1?

Or maybe it's the other way around -- maybe the Inquirer got it wrong, and DX 9 will support the new hardware, and come out this fall, while DX 9.1 supports .Net and comes out in 2003? That sounds more reasonable.

Unless, of course, MS knows that no DX9-level hardware is actually going to ship this year? (Say it isn't so!)

LeStoffer

23-Mar-2002, 13:43

Or maybe it's the other way around -- maybe the Inquirer got it wrong, and DX 9 will support the new hardware, and come out this fall, while DX 9.1 supports .Net and comes out in 2003? That sounds more reasonable.

Just my thought on the matter.

Regards, LeStoffer

muted

23-Mar-2002, 18:46

Sorry , i 've been reading the inquirer lately .. ya know .. not everything is true

PC-Engine

24-Mar-2002, 11:05

Isn't Farenheit (DX9) OpenGL mixed in with D3D?

Rodéric

24-Mar-2002, 13:32

Farenheit has been canceled eons ago.

Rags

25-Mar-2002, 05:33

http://www.watch.impress.co.jp/pc/docs/2002/0322/video3.mpg

Listen to the voice. Clearly states it was done in Hardware with Matrox hardware.

Rags

pascal

25-Mar-2002, 17:24

My english is not good, but I can hear it very well "...and the Matrox displacement mapping hardware" 8)

pascal

25-Mar-2002, 17:40

IIRC some time ago (1996 or 1997?) Carmack was talking about a new game engine named Trinity (never finished). His vision for this new engine was a completelly outdoor/indoor capable game engine. He described something like a caracter flying for many miles and then landing in the top of a castle and getting inside the castle without any problem.

Will displacement mapping help with it?

edited: See this interview http://www.voodooextreme.com/articles/interview_jc.html

VE: Is Trinity dead? Do you have plans, after Q3 is finished, to resume work on Trinity, or is it dead, for sure?
John: I have always been a little puzzled how "trinity" became a specific thing. At the end of Quake 1, I started doing a bit of future research with morphing level of detail representations and other things. We got tired of just talking about "the next technology", and Michael Abrash suggested following Intel's strategy of naming development work for nearby rivers. Dallas -> trinity river.

It was always just a convenient way of talking about stuff I was looking at that wasn't part of the current technology. I had several completely different experimental rendering frameworks that I worked with under the "trinity" name: full surface unique texture rep, voxel splat, voxel ray cast, deformed environment meshes, lummigraphlets, terrain morphing, etc.
edited2: See this too 8) http://www.planetquake.com/Trinity/infobytype.htmlRendering Engine Features

Will use OpenGL for rendering. (Source: An email from John Carmack to Jim Lowell. 12/8/97)
Will use displacement maps for added detail on close objects. (Source: A note which Brian Hook sent to the Fatcity OpenGL Mailing List. 12/1/97)

...

What about a enhanced Doom3 engine with more passes and displacement mapping doing big outdoors 8)

pascal

25-Mar-2002, 20:13

Some Photoshop help needed here: http://www.watch.impress.co.jp/pc/docs/2002/0322/dx9_5.jpg
See the yellow Menu Help in the top left with 14 lines.
One line say "Toggle Clouds Texture"
Other line says "Full Mode (solid)"

Definitelly a Matrox hardware 8)

Could someone invert the yellow to black and apply some sharp filter in the JPG above?

nAo

25-Mar-2002, 20:17

Definitelly a Matrox hardware 8)
What?? Where is the proof?

pascal

25-Mar-2002, 20:48

What ??? Do you believe doing it realtime in software emulating hardware ???

nAo

25-Mar-2002, 20:53

What ??? Do you believe doing it realtime in software emulating hardware ???
Of course I do. One don't need to emulate the hw, one just needs to write the main algorithm via software.
Displacement mapping is not too hard to do, interactive frame rates are possible on current CPUs...

ciao,
Marco

pascal

25-Mar-2002, 20:56

Well, you dont need to emulate all the gates of the hardware, but emulating displacement mapping is not enough.

For example one will need to emulate other algorithms like the filtering using the CPU.

nAo

25-Mar-2002, 21:03

For example one will need to emulate other algorithms like the filtering using the CPU.
You'd better watch some good sw rasterizer. I coded a fast (25 fps, 320x256x8 bit) trilinear-filtering roto-zoomer on my old 25 mhz/460 ns access memory Amiga4000 almost 5 years ago.
With SIMD extensions is even easier to code stuff like that, with tiled textures bilinear filtering could be almost free, hiding operations with mem latency. I can't see why a modern CPU with a fast submemory system can't do it.

ciao,
Marco

pascal

25-Mar-2002, 21:07

Is it possible to use current hardware to do some algorithms and software to do the displacement mapping?

The mountains resolution was too good, probably 1024x768 or better. See the Windows line at the bottom.

nAo

25-Mar-2002, 21:10

Is it possible to use current hardware to do some algorithms and software to do the displacement mapping?
Of course it is. Do all the displacement stuff on cpu and send the generated geometry to the graphics card.

ciao,
Marco

pascal

25-Mar-2002, 21:20

Oh well, it is probably a displacement mapping software emulation using some Matrox 3D card to do the other tasks. :(

Thanks nAo :)

Nappe1

26-Mar-2002, 01:01

nAo: don't jump on gun.

I am 99% sure that those videos weren't running on SW.
Displacement maps with that amount of tesselation on CPU and getting acceptable framerates?? well, I really would like to see that.

I have one pretty simple Displacement Mapping demo running totally on SW and with 1Ghz Athlon TB I get about 14 fps with 18Kpolys tesselation and 320x200x8 displacement map. And eventhough that demo on those videos would have used some parts on HW and some SW, it would have been just too much for CPU.

and don't get me wrong... I understand your point of view, but there is a huge jump from Trilinear filtered rotozoomer (Future Crew had this on their Second Reality btw. Or was it using even Anisotropic?? :) ) to Displacement Mapping (with that amount of tesselation.) done in SW.

and nAo: check out the Soiled Legacy from Resource. That is freaking awesome demo if you ask me. and it is running on good old C64! so get a copy of CCS64 and grab that demo from http://c64.ch and I can promise that you need to pick up your jaw from the floor more than once. :)

Dave Baumann

26-Mar-2002, 01:07

Our 'mole' at GDC had this to say about the Matrox presentation:

Matrox seems to have working silicon during their displacement mapping demos they showed it in realtime (was too fast to be software emulated) they also ran 3D Max on that same machine and showed displacement mapping design for a game in realtime. The machine only crashed once, unfortunatly they immediatly turned off the display so we could not see the BIOS bootscreen. Displacement mapping seems to be very cool but there are quite a few limits imposed on the artwork (to avoid getting gaps in the meshes).

nAo

26-Mar-2002, 01:08

getting acceptable framerates?? well, I really would like to see that.
Umh..that video is pretty bad, but it doesn't show any acceptable framerate :)

and don't get me wrong... I understand your point of view, but there is a huge jump from Trilinear filtered rotozoomer (Future Crew had this on their Second Reality btw. Or was it using even Anisotropic?? :) ) to Displacement Mapping (with that amount of tesselation.) done in SW.

FC's rotozoomer was the worst part of that demo, a low precision unmipmapped point filtered mess :)

ciao,
Marco

Nappe1

26-Mar-2002, 01:15

nAo: well, the year was also 1992 so what else you expect? ;) 386 was a glorious thing and some rich ppl might have mighty 486 machine, but 286 wasn't rare either.

(btw, 2nd reality runs pretty well on 286 10Mhz with 287 on machine with MCA bus.) :)

nAo: and watch that demo!! :) it really something pretty cool. ;)

nggalai

26-Mar-2002, 08:30

Hi Rags,Listen to the voice. Clearly states it was done in Hardware with Matrox hardware.Whoop, didn't realise there was a soundtrack with the movie (office PC . . .). I stand corrected. :oops:

OK, anybody in for a G800 speculation session? ;)

ta,
.rb
________
easy vape vaporizer (http://vaporizers.net/easy-vape)

Tempest

26-Mar-2002, 11:05

Some Photoshop help needed here: http://www.watch.impress.co.jp/pc/docs/2002/0322/dx9_5.jpg
Could someone invert the yellow to black and apply some sharp filter in the JPG above?
It's nothing too exciting, goes something like this:

FA? - Toggle Text Display
L?? - Dynamic Tessellation (on)
C? - Tessellation Factor (320:1) (not sure about these numbers)
?? - Displacement Scale (1.30:1) (not sure about these numbers either)
YA? - Displacement Interpolation Type (linear)
TA? - Sun Angle 135,000000 degrees
Z/X - Sun Light Factor 1.000
C/V - Sky Light Factor 1,000
1/2 - Toggle Terrain & Sky Radiance Texture (on)
3/4 - Toggle Detail Color & Detail Horizon Map Texture (off) (could be "on" as well...)
5/6 - Toggle Normal Map & Horizon Map Texture (on)
7/8 - Toggle Clouds Texture (on)
W?S - Fill Mode (solid)
R?? - Fog (enabled)

makuta

26-Mar-2002, 11:24

pascal

26-Mar-2002, 13:37

Listen to the video again:
"version pixel shader ... and displacing mapping ... and the Matrox displacing mapping hardware"

Well, was it done using some Matrox hardware partially or totally?

The key is the "version pixel shader". What is the current Matrox hardware supporting pixel shader? None right? :D

edited: Looks like it was a 1024x768 pixel shader with displacing mapping presentation. The totally done by Matrox hardware looks more plausible 8)

pascal

27-Mar-2002, 14:50

From www.matroxusers.comFriday March 22, 2002
Displacement Mapping At GDC
News item posted at 20:37 GMT by Ant
It seems Microsoft have demoed DirectX9 displacement mapping at GDC showing off a couple of demoâ€™s, and what's interesting about that you ask? Well it's displacement mapping for one and for two the demo had the word Matrox plastered over it. A report is over on a Japanese site, they have mpeg videos of the demo's and there are translations of sorts here and here in our forums. From the translation it seems to indicate the demo's are run in software but on one of the videos a voice can be heard to say it is running on Matrox displacement mapping hardware and it sure doesn't look like a G550 to me.
Are you starting to believe yet?

Tuesday March 19, 2002
Matrox To Sneak In A Glimpse Of Parhelia At GDC?
News item posted at 2:26 GMT by VigilAnt
As was posted here before by Ant, Matrox is scheduled to have some presentations at GDC. The most interesting is one by Juan Guardado . If you remember from last year, Juan showed a demo of displacement mapping running on unnamed hardware (now a G550). If you were at the event, you will remember the demo ran at a snail-paced 1-3 fps. Well, this year he is at it again, but this time I would bet we see some real hardware to push a demo of displacement mapping. Barring any unexpected problems, expect a preview of what Parhelia will be capable of.

Discuss this here.

VigilAnt

Matrox had used the software emulation before with a unanounced card.
Two things:
- It was too slow, then probably it is not the same software emulation.
- They used an announced card, maybe they are using another announced card this time too 8)

We need Sherlock Holmes help :)