New DX9 demo

[Humus]

Humus delivers his first DX9 demo

It's a volumetric lighting demo requiring vertex shaders 2.0, so it should run on Radeon 9700 and Parhelia.

http://esprit.campus.luth.se/~humus/

 

[Neeyik]

Humus - reading the info about this demo on your site. Seeing as I'm not a programmer, I'm probably being thick about this, but is it possible to use multiple passes with VS1.1 to get the same effect or does it have to be done with just one VS2.0 pass? On the subject of 2.0, is it possible that you could tweak your demo to allow software vertex processing, so that one could see the difference between this and hardware acceleration of VS2.0?

Off topic a little, I've noticed that the VS demos in the DX9 SDK show a significant difference in performance between the sofware processing of VS2.0 and 3.0 - I presume there must be a specific reason for this; can anyone feel me in on the details? My usual victims for these sorts of questions are a bit busy at the moment!

 

[Humus]

Humus - reading the info about this demo on your site. Seeing as I'm not a programmer, I'm probably being thick about this, but is it possible to use multiple passes with VS1.1 to get the same effect or does it have to be done with just one VS2.0 pass? On the subject of 2.0, is it possible that you could tweak your demo to allow software vertex processing, so that one could see the difference between this and hardware acceleration of VS2.0?

Off topic a little, I've noticed that the VS demos in the DX9 SDK show a significant difference in performance between the sofware processing of VS2.0 and 3.0 - I presume there must be a specific reason for this; can anyone feel me in on the details? My usual victims for these sorts of questions are a bit busy at the moment!

I was first thinking about doing a multipass fallback, but decided it wasn't worth the effort. It would require me to first render the multipass to an offscreen buffer and then use that buffer in the pixel shader. I haven't finished the render-to-texture code in DX in my framework yet, it would also be quite slow to render five passes over a screensize buffer, then the last pass for outputting the final output to the backbuffer.
For the second query, I did a quick test and it turned out that I get 104 with hardware processing and 7 without. So yes, quite a large difference.

 

[Althornin]

Hmm, Humus, with the latest drivers for the 9700 i get an error in fullscreen mode "Error: Couldnt Create Direct3D Device interface"

I also get problems with your "shadows that rock" demo - it flashes up on the screen for a second, then the screen remains black.

Note: they both work fine in windowed mode - although the shadow demo does the same thing in windowed mode if AA is enabled. But it wont work in Fullscreen at all, neither will the Volumetric one, no matter the AA/Aniso settings.

 

[Volenti]

[de-lurk] works fine here with the second latest driver set, fairly sutble effect.

oh it's based on dm-Fractal from UT

 

[Humus]

Hmm, Humus, with the latest drivers for the 9700 i get an error in fullscreen mode "Error: Couldnt Create Direct3D Device interface"

I also get problems with your "shadows that rock" demo - it flashes up on the screen for a second, then the screen remains black.

Note: they both work fine in windowed mode - although the shadow demo does the same thing in windowed mode if AA is enabled. But it wont work in Fullscreen at all, neither will the Volumetric one, no matter the AA/Aniso settings.

Update your drivers. The black screen bug was fixed in the 6275 driver. Not sure about the other problem, but it may be related.

 

[Althornin]

Update your drivers. The black screen bug was fixed in the 6275 driver. Not sure about the other problem, but it may be related.

as i said, Humus, i AM using the latest Catalyst 3.0a drivers.

On a possibly related side note - all the ATI demos work fine for me, EXCEPT for the NPR hatching demo.

EDIT: Hmm, it seems that the cat 3.0a drivers are only rev 6255.
SO i should be using the ultra beta drivers with the 16bit FSAA support?

 

[Brent]

works great for me, 9700 Pro in WinXP with Cat 3.0 drivers (6.14.01.6255) 7.81 packaging version

at 2x aa and 16x af on my p42.4b @ 2.7 150fsb/ddr dram

122 fps

very nice work Humus!

 

[Andrew]

Very nice.

 

[Randell]

On a possibly related side note - all the ATI demos work fine for me, EXCEPT for the NPR hatching demo.ev 6255.

Try the NPR hatching demo without AA/AF see if it works then.

 

[Althornin]

Try the NPR hatching demo without AA/AF see if it works then.

Ah, that makes it work
Thanks.

 

[Rodéric]

But I don't want to buy a new graphic card just yet !

 

[Hyp-X]

Works fine with 6275.

Is there any reason you can't fallback to software VP on cards that has PS1.0 support but <2.0 vertex shaders?
It shouldn't take much more than 1-2 lines of code...

 

[Doomtrooper]

I think the main reason would be the title..DX9 Demo as their isn't alot of them out there but many, many DX8 including most of them on Humus's site.

 

[Randell]

Try the NPR hatching demo without AA/AF see if it works then.

Ah, that makes it work
Thanks.

np I had the same issue

 

[Humus]

Works fine with 6275.

Is there any reason you can't fallback to software VP on cards that has PS1.0 support but <2.0 vertex shaders?
It shouldn't take much more than 1-2 lines of code...

Well, I would have to redesign my framework a little for that. I'm not sure if it's worth it when it runs at 7 fps with software VP.

 

[epicstruggle]

dam, just returned my 9700pro and you came out with a new demo. darn have to wait till i can scrounge money for a new card.