Ikaruga
- Thread starter lethal_illuminati
- Start date
one thing again about the DC graphics chip: PowerVR chips don´t really care how many polygons are overlapping in a given scene since the way they render it they will only render what´s visible as opposed to nearly every other approach on the market. This method is called "tile based rendering", the screen is split up in "tiles" and for each of them the chip determines which polygons are completely visible, which ones partly and the ones that don´t have to be rendered at all. This only requires a very small (64kb) memory window so that a huge amount of fillrate can be saved by this geometry processing tactic. Just use a KyroII-card on a PC and download the "villagemark"-demo from the STM website which has a very high overdraw (overlapping of polygons). Cards that render the scene from back to front (usual approach) will go down on their knees, cards with tile-based rendering will fly. In special occasions like this a KyroII can outperform a Geforce3ti500! (but these occasions are quite rare).
Second, the PowerVR-chips are capable of multitexturing and can apply quite a few textures per pass whereas any multitexturing capabilities on the PS2 have to be coded in software (as has the AA-algorithm, although it is part of the second generation devkit software). So when you use Bump-mapping it comes without a great speed hit on the DC because the effect can be (almost) entirely done in hardware.
To put it short, it heavily depends on the effects used whether a PS2 game can be done on the DC in a similar looking way. But when the PS2´s vector units are used and some clever effects are programmed in software the higher speed and flexibility of the core units will make the difference.
The PS2 does not have too many hardcoded effects. The programmers have to create graphics engines themselves with unique capabilites which leads to many games looking quite different. The DC (and imho the Xbox) on the other hand have graphics chips which come with a lot of pre-programmed effect codes and devkits which descibe a lot of effects. imho this leads to a specific "look" of games on these machines, there´s a certain "feel" which makes you know that you´re playing a DC game. Same thing occured to me with the commodore Amiga, back then the limited color palettes and the continuous use of certain colors by the developers led to games having a distinctive "amiga-esque" look.
that´s just my opinion though
edit: just looked at the aura_for_laura-demo. It´s a typical scenedemo, much of the geometry is cleverly faked though ;-) Some calculations may be stored in advance, the shading doesn´t have to be correct too. In scenes with lots of repeating geometry one can use clever programming to get these "infinity effects", and don´t forget that they save on Z-buffering and other things needed for games by using clever camera angles and such. (I´ve followed the amiga and PC demoscene since about 92 and have seen some clever tricks during this time. Almost all rendering effects that graphic card makers are hyping now have been done in software in a similar fashion three to four years ago. At the university where I´m studying realtime raytracing is the latest cry, currently pushing several billion triangles per second! (fully textured and correctly shaded geometry of course, a field of sunflowers to be exact))
Second, the PowerVR-chips are capable of multitexturing and can apply quite a few textures per pass whereas any multitexturing capabilities on the PS2 have to be coded in software (as has the AA-algorithm, although it is part of the second generation devkit software). So when you use Bump-mapping it comes without a great speed hit on the DC because the effect can be (almost) entirely done in hardware.
To put it short, it heavily depends on the effects used whether a PS2 game can be done on the DC in a similar looking way. But when the PS2´s vector units are used and some clever effects are programmed in software the higher speed and flexibility of the core units will make the difference.
The PS2 does not have too many hardcoded effects. The programmers have to create graphics engines themselves with unique capabilites which leads to many games looking quite different. The DC (and imho the Xbox) on the other hand have graphics chips which come with a lot of pre-programmed effect codes and devkits which descibe a lot of effects. imho this leads to a specific "look" of games on these machines, there´s a certain "feel" which makes you know that you´re playing a DC game. Same thing occured to me with the commodore Amiga, back then the limited color palettes and the continuous use of certain colors by the developers led to games having a distinctive "amiga-esque" look.
that´s just my opinion though
edit: just looked at the aura_for_laura-demo. It´s a typical scenedemo, much of the geometry is cleverly faked though ;-) Some calculations may be stored in advance, the shading doesn´t have to be correct too. In scenes with lots of repeating geometry one can use clever programming to get these "infinity effects", and don´t forget that they save on Z-buffering and other things needed for games by using clever camera angles and such. (I´ve followed the amiga and PC demoscene since about 92 and have seen some clever tricks during this time. Almost all rendering effects that graphic card makers are hyping now have been done in software in a similar fashion three to four years ago. At the university where I´m studying realtime raytracing is the latest cry, currently pushing several billion triangles per second! (fully textured and correctly shaded geometry of course, a field of sunflowers to be exact))
Vash, you said most of the geometry in the Auro for Laura demo is faked?
PS2 doesn't support bump mapping.
Yes, with repetitive geometry you can get away with some clever stuff (for example, the infinite terrain in Star Wars: Rogue Squadron 2 and 3 on the Death Star - there are only like 4 'blocks' of terrain that are transformed, then shifted all over the place for very small performance cost)... but I don't see (m)any places in that demo where that could be used.
And you basically just helped confirm what I've been saying about "tech demos" as a whole. =)
PS2 doesn't support bump mapping.
Yes, with repetitive geometry you can get away with some clever stuff (for example, the infinite terrain in Star Wars: Rogue Squadron 2 and 3 on the Death Star - there are only like 4 'blocks' of terrain that are transformed, then shifted all over the place for very small performance cost)... but I don't see (m)any places in that demo where that could be used.
And you basically just helped confirm what I've been saying about "tech demos" as a whole. =)
The DC's PVR card can do two textures per pass. It needs an extra pass for bump mapping, if you want a diffuse texture to be combined with it. If the bumpy grey-ish look is enough for you, a single pass will do.
On the bump mapping thing... I don't recall seeing it in any DC game (but I have played quite few games, actually). Reading some documentations, I found the DC's bumpmapping is VERY hardcoded, and seems to work only with directional lights (emboss?). I figure that most developers weren't too enthusiastic with such limitation (the usage is quite crippled without point lights). Or it's too slow to be viable?
Can anyone point out a game or even a demo where it's clearly used? And no, Soul Calibur Inferno stage doesn't use bumpmapping, nor per-pixel lighting. A quick look reveals that the "bumps" in the ground texture are always illuminated from the same direction, and for the super-smooth lighting I'm 90% sure that's a super-tesselated mesh (since the background has almost no detail at all). The other 10%goes to a 2nd texture layer with an additive circular gradient that scrolls around to match Inferno's position.
On the bump mapping thing... I don't recall seeing it in any DC game (but I have played quite few games, actually). Reading some documentations, I found the DC's bumpmapping is VERY hardcoded, and seems to work only with directional lights (emboss?). I figure that most developers weren't too enthusiastic with such limitation (the usage is quite crippled without point lights). Or it's too slow to be viable?
Can anyone point out a game or even a demo where it's clearly used? And no, Soul Calibur Inferno stage doesn't use bumpmapping, nor per-pixel lighting. A quick look reveals that the "bumps" in the ground texture are always illuminated from the same direction, and for the super-smooth lighting I'm 90% sure that's a super-tesselated mesh (since the background has almost no detail at all). The other 10%goes to a 2nd texture layer with an additive circular gradient that scrolls around to match Inferno's position.
Medion, there isn't a single Dreamcast game that uses bump mapping. AT ALL.
There are probably a small handful of TECH DEMOS though.
I don't think it's emboss, AFAIR it's a different implementation (EMBM?) but yes, it is pretty restrictive, thanks to the libraries you have to use to get it working.
There are probably a small handful of TECH DEMOS though.
I don't think it's emboss, AFAIR it's a different implementation (EMBM?) but yes, it is pretty restrictive, thanks to the libraries you have to use to get it working.
