Roboblitz: 50 Mb Xbox Live Arcade using UE3
- Thread starter basanti
- Start date
Procedural Textures (on the 360)
Very interesting technology...
Full PR: http://www.gamesindustry.biz/press_release.php?aid=15093
Site: http://www.allegorithmic.com/v2/news.htm
Very interesting technology...
Full PR: http://www.gamesindustry.biz/press_release.php?aid=15093
Site: http://www.allegorithmic.com/v2/news.htm
Oops, just found http://www.beyond3d.com/forum/showthread.php?t=28875 in the games forum...
Martin Eddy
Regular
Sorry to resurrect a 5 month old thread but this game was mentioned in this thread and I thought it a good idea to show what the game looks like here rather than derail the other thread.
Not bad looking for an XBLA title.
Not bad looking for an XBLA title.
I agree it looks good, but I suspect the title can get away with this kind of technology more easily than games with other (more 'natural') settings. Does that make any sense? 
There's some very cool stuff here though: http://www.allegorithmic.com/gallery/profx_list.php
Have a look at that demo on the right too (' ProFX2 Rendered materials').
There's some very cool stuff here though: http://www.allegorithmic.com/gallery/profx_list.php
Have a look at that demo on the right too (' ProFX2 Rendered materials').
Last edited by a moderator:
One more tidbit: 'As an example, the 73 full materials (about 200 maps: diffuse, specular, normal, etc.) in the GDC demo level, representing 80MB of DXT compressed content, are being generated in 4 seconds at load-time. The ProFX description file weighs only 280KB!'
Edit - According to their forum (post from yesterday!): http://www.roboblitz.com/forum/topic.php?tid=37
Edit - According to their forum (post from yesterday!): http://www.roboblitz.com/forum/topic.php?tid=37
The good news is, that's because we're content complete and have sent the game off for 3rd party testing and certification.
Last edited by a moderator:
Martin Eddy
Regular
I agree it looks good, but I suspect the title can get away with this kind of technology more easily than games with other (more 'natural') settings. Does that make any sense?
Of course, materials like checkerplate steel are a lot more consistently repetitive than say bark on a tree.
Yeah, I read that one long time ago as well, but thought it is mostly company PR fluff. I awlays thought that there would be a trade of in time, that is using procedurals would take of course much sorter time to load but much longer to generate, that in the end it would take longer time to load and generate a procedural than load and decompress a normal compressed texture.
But I guess with how powerfull the CPUs in the consoles now are while the disc reading speeds are similar as previous gen that using procedurals will both lead to less space on the disc and faster loading, with the only negative now being that it takes more CPU power to generate a procedural than to decompress a texture and might therefore not be so good to do it on the fly...
True but it seems the Middleware they're using can also generate convincing textures of natural objects as well (see above) http://www.allegorithmic.com/gallery/profx_list.php
Procedural is the future and I think it's about time devs start utilizing this tech as it will also enable artists to rid themselves of the static repetition on existing textures without investing the time creating these individual textures that extend large enough to be unique over large areas, or unique across multiple cloned objects. Procedurals will enable not only smaller overall game sizes (in mb) but also more unique looking games overall.
Applying this same methodology and ideal to game character behaviors and the use of physics instead of canned animations will again enable smaller sizes and better experiences as well. Ironic seeing as ps3 seems to have an edge in this ability to generate procedural content over 360, yet it doesn't need it as much as 360 does wrt optical storage. Perhaps they could have got away with a dvd9 for ps3 afterall.
Well, the 50 MB limit for XBLA titles aside, this doesn't seem a very good tradeoff. DXT textures compress typically at least 2x even with something simple like zlib, so without procedural generation they'd had to read 40 MB, which, incidentally, is what you can expect to read in 4 seconds off a DVD on a Xbox 360.
If I understood that quote from the site correctly, those 80MB were already compressed...
That's how I read it too.
What tradoff? They're gaining almost 200 times the space on disc (40MB down to 280KB) and losing... what?
Last edited by a moderator:
MasterDisaster
Regular
procedural textures are surely the future
but we are talking here of pre-processed texture synth (infact there's a 4 sec of loading while the cpu makes the textures)
the real revolution will be there when some, if not most, of the textures will be generated in realtime, this can give an incredible detail at every level of zoom, and will makes useless texture filters as AF
and more important, the texture can change il realtime reacting to some elements WHILE the gpu is texturing (same can be done with whole fragments and vertex data)
the machine is ready to this, MemExport, cache sharing/locking and core locking functions are here, we have to wait till the right tool wil permit to dev to do the "big step jump"
but we are talking here of pre-processed texture synth (infact there's a 4 sec of loading while the cpu makes the textures)
the real revolution will be there when some, if not most, of the textures will be generated in realtime, this can give an incredible detail at every level of zoom, and will makes useless texture filters as AF
and more important, the texture can change il realtime reacting to some elements WHILE the gpu is texturing (same can be done with whole fragments and vertex data)
the machine is ready to this, MemExport, cache sharing/locking and core locking functions are here, we have to wait till the right tool wil permit to dev to do the "big step jump"
It doesn't matter how the textures are generated, if you want to avoid blurring AND pixel flicker, you still need AF.
Similar threads
