Not to bash the game, but when I heard the reference to 32 texture layers I couldn't stop laughing. The texturing from that video looks pretty dull 32 layers or not. From all the hype that's been previously generated by this game, that video sure looks unimpressive. It looks pretty good though.
The First Direct Feed High Quality Killzone Footage!
- Thread starter nazlec
- Start date
marconelly!
Veteran
That video actually looks way more impressive visually and aureally than I thought it would be. The smoke effects alone are overkill... I mean, you can argue that in the light of Half Life 2, Doom 3, etc. this isn't all that (although I would argue this has atmosphere so thick and art direction so good, that it really doesn't matter how many texture layers it renders), but the game looks so much better than anything else in the genre on the PS2, it's not even funny. Those developers deserve kudos for that fact alone.
32 texture comments is probably wrong just like the most other info in that IGN article ('Gorilla' games, fighting against 'aliens' etc.)
32 texture comments is probably wrong just like the most other info in that IGN article ('Gorilla' games, fighting against 'aliens' etc.)
marconelly!
Veteran
Deepak, you need to download and install latest DivX. That will make videos work.
Deepak said:
Quick, easy, but kinda lame: http://www.videolan.org/ . VideoLanClient (VLC) is a media-player, that includes all kinds of codecs. Those codecs are part of VLC, they are not global system codecs. You will be able to watch pretty much every movie, but only in VLC. VLC is freeware, only 800K big and doesn't need to be "installed", you just have to unzip it and you are ready to go.
The better way is to check which codec was used to encode the video clip. You could check that with a tool called "gspot" (download here), when you know what codec is required, go download it here. Don't forget to reboot your PC, after you installed a codec ... just to be on the save side.
Enjoy!
nondescript
Regular
PC-Engine said:
Actually, I agree with you - that layered texture thing didn't seem to be very impressive, nor were the visuals - hard to do justice with a video like that tho. The grenade explosions need some work too.
Maybe I'm just not a fan of any FPS that doesn't have "Quake" or "Half-Life" in the title
Re: Direct links
http://www.gametrailers.com/gt_vault/t_killzone_gp2.html
The videos look like engine demos...
It is very obvious that rag doll physics, animations, AI and some effects aren't finalized or not even implemented in this very early build.
Deepak said:
http://www.gametrailers.com/gt_vault/t_killzone_gp2.html
The videos look like engine demos...
It is very obvious that rag doll physics, animations, AI and some effects aren't finalized or not even implemented in this very early build.
Steve Dave Part Deux said:
The graphics synth can only single-texture, and it doesn't buffer geometry on-chip, so to do more than one layer you WOULD have to re-send the geometry, there's no other way around it.
It's very likely the engine is *capable* of doing 32 layers, but that the actual useage in-game is much less. With even a 32-bpp framebuffer, doing 32 texture layers would probably mean so many rounding errors everything would just turn into a blotchy mess, and we know not all PS2 games actually use 32-bit screen buffers...
Anyway, no matter how many textures are actually used, it is a smart idea.
Guden Oden said:
i believe the original poster's point was that the geometry does not need to be re-T&L-ed. otherwise, it most likely needs to be resend to the rasterizer unit, in the sense that it has to be re-trisetup.
those 32 texture layers do not have to be diffuse maps, so you do not need to get blotchy mess at the end.
The GS can fill 40 screens in one frame (30Mp/s being the baseline required for 640x480), so 32 texture layers sounds plausible for realtime maximum performance.
As Fafalada said in another thread, the hit it takes on bandwidth to resend large meshes numerous times for multitexturing, can be alleviated somewhat by pre-blending the texture on much simpler geometry, and then apply the finished texture to the large model.
GS has the large 4 Mb scratchpad and dual states (two sets of registers) so multitexturing shouldn't have that big a hit on performance.
As Fafalada said in another thread, the hit it takes on bandwidth to resend large meshes numerous times for multitexturing, can be alleviated somewhat by pre-blending the texture on much simpler geometry, and then apply the finished texture to the large model.
GS has the large 4 Mb scratchpad and dual states (two sets of registers) so multitexturing shouldn't have that big a hit on performance.
darkblu said:
I didn't say it would.
The EE doesn't need to re-T&L if you're re-sending the entire finished model (but that requires you keep it around by copying it to main memory, so you'd get a bandwidth hit on the main bus). Assuming you're basically just applying decals to a model, you would only T&L and re-send the additional polys. For example, you wouldn't apply a shiny specular map to a person's head. Assume you only want the character's arms shiny, it might make more sense to not save the entire model away in main memory and simply re-T&L the arms and then upload them to the GS through the direct interconnect between VU1 and GS.
I see little need for 32 layers of opaque texture maps on a model.
Then again, I see little need for 32 layers period, but like I said, it's probably an engine absolute limit and not a situation that happens under a practical scenario.
It's still only 8bit calculation precision - unless you are only doing additions it just doesn't go very far.darkblu said:
Try doing multiple dot3 lights in multipass emulation and see how that works out...
Anyway if someone can point me to a justification for drawing a polygon with 32stages, and then show me evidence of that in Killzone media, I am all for believing it.
But right now the number is either a misprint or an out of context misinterpretation of something the article writer didn't understand.
That works only on trivial blends. Anything where blending value varies per vertex and/or UV coordinates don't match is out of the question (that includes detail maps).Squeak said:
