displacement mapping?

[Fred]

So whats the word about which implementation is going to be heralded in the upcoming months.

As I remember, a lot of people were displeased with the cut and dry Matrox approach? Or will that become the standard.

 

[nooneyouknow]

Displacement Mapping in DX9 will not please very many people.

 

[Joe DeFuria]

If history is any indication with "new" features like this, it's going to take 1 or 2 more incarnations of DirectX and hardware before most developers start taking displacement mapping seriously.

I consider the first "pixel shading" hardware to be DX7 level. EMBM and DOT-3 can be considered to be the "first crack" at what we call "pixel shaders." But the combination of the relative inflexibility, and especially the performance level of DX7 hardware, made these features more gimmicky than really useful.

Second generation DX8 and DX8.1 hardware are starting to make pixel shading more of a reality (both because of improved flexibility and performance). I suspect that with DX9 hardware most develoeprs will finally be able to code for pixel shaders as the base-line.

I'm expecting a similar path for displacement mapping. We'll see some tech demos, and maybe a few "gimmicky" applications of it in a handful of games over the next couple years...but I'm betting that most developers won't be able to take it seriously until DX10 or 11. (Because of increased flexibility of future implementations, and increased hardware performance to be able to handle it.)

 

[nooneyouknow]

You got that right Joe.

Except for DX 10. That has been renamed DX9.1 at this point.

 

[Joe DeFuria]

Ack! My, what fun we'd have if it turns out that DX 9.0 closely matches ATI hardware, and DX 9.1 closely matches nVidia hardware.

Then we can watch both camps flip sides when we talk about how "useful" DX 9.1 is compared to DX 9.0.

I thought though, that as far as shaders are concerned, DX 9.0 language was suppossed to be flexible enough to accomodate different hardware implementations. Any info on how DX 9.1/10 differs from 9.0?

 

[nooneyouknow]

DX9.1 will not be like the minimal DX8.1.

I think it is so to allow Microsoft to time the major revisions with major OS releases.

So, DX9.1 will be release in between OS cycles but will be a fairly large upgrade. hard to tell right now but that is what I gather.

EDIT: The only reason there was an DX 8.1 is that MS released how important PS 1.4 was. Remember, they also added PS 1.2 and 1.3, so Nvidia should be happy they did since GF4 needed it :smile:

<font size=-1>[ This Message was edited by: nooneyouknow on 2002-02-28 16:30 ]</font>

 

[Hyp-X]

On 2002-02-28 16:29, nooneyouknow wrote:
Remember, they also added PS 1.2 and 1.3, so Nvidia should be happy they did since GF4 needed it :smile:

Not only nVidia...

 

[Nappe1]

Hyp-X:


<font size=-1>[ This Message was edited by: Nappe1 on 2002-02-28 20:07 ]</font>

 

[nooneyouknow]

Right, ATI, Nvidia, and Matrox needed it :smile:

 

[muted]

how hard would it be for a game developer to convert their Bump maps to Displacement mapping .. it's the same concept .. i don't have the presentation on displacement mapping ...

still it's a shame a lot of game don't even support simple bump mapping

 

[Joe DeFuria]

I think a lot of that comes down to performance. Up until the current crop of cards, (GeForce3+ Radeon 8500+), I don't believe the cards were generally powerful enough to perform bump-mapping in any way that "really" makes a difference. Sure, you could bump-map (or DOt-3) an object or two here or there.

But following Carmack's approach...he doesn't really want to use a feature until it generally makes "everything" better. So that you can apply per pixel lighting globally on all surfaces. That's when we'll get the next "step change" in graphics quality.

As Carmack "strongly suggests", older cards capable of DOT-3 simply don't have the performance capabale applying the effect in such a global manner. That's the reason, IMO, why it isn't as widely used today.

 

[Hyp-X]

On 2002-02-28 21:10, Joe DeFuria wrote:

As Carmack "strongly suggests", older cards capable of DOT-3 simply don't have the performance capabale applying the effect in such a global manner. That's the reason, IMO, why it isn't as widely used today.

Apart from performance, those cards don't have all the features needed, and most of them doesn't have the sufficent memory capacity.

 

[Nexiss]

Felt that this link would interest a few of you. If you check in the gallery, you'll see a few pictures from his per-pixel displacement mapping technique. He posted an IOTD over at flipcode (which is how I found the site) showing the technique for a terrain.


<font size=-1>[ This Message was edited by: Nexiss on 2002-03-01 01:42 ]</font>

<font size=-1>[ This Message was edited by: Nexiss on 2002-03-01 01:42 ]</font>

 

[Xmas]

Well, I'm not too impressed by this technique... the second shot looks quite good, though it exhibits some visual artifacts. The first one doesn't look convincing at all.
But somehow this reminds me of emboss bump mapping...

 

[MfA]

It looks cool, but its very fudged. You'd have a hard time integrating it with normal geometry ...

You cant do per pixel displacement mapping without an iterative search, its just one of those fundamental issues.