I think this thread is in large parts the same as http://forum.beyond3d.com/showthread.php?t=48244 Perhaps the discussions should be combined.
Optical disc formats & next gen consoles.
- Thread starter Mendel
- Start date
Sony ARE an electronics company, there modus operandi . Well widespread 3D LCD adoption will lead to?. The PS3 can already do 4K movies. But it's all about timelines.
But then on the other hand I wonder if next gen consoles could use jpeg2000 or some other better compression algorithm to compress source material and do realtime decompress/recompression to gpu friendly format on the fly. Might save quite a nice amount of disc space and IO bandwidth.
Edit. this is ofcourse guessing on my part, I'm just guessing rage uses gpu friendly compression on disc which doesn't pack data as efficiently as it could be due to lacking cpu power to do de/recompression cycle
Last edited by a moderator:
Rage is already using a very high level of compression on the textures, and it suffers quite a lot of quality degradation (at least on the X360) so higher ratios aren't preferred. We want a reasonable level of image quality, damn it 
But based on the criticism of the game by some people even here, texture resolution will still have to be increased with next gen games - which means even more data.
But based on the criticism of the game by some people even here, texture resolution will still have to be increased with next gen games - which means even more data.
Rage is already using a very high level of compression on the textures, and it suffers quite a lot of quality degradation (at least on the X360) so higher ratios aren't preferred. We want a reasonable level of image quality, damn it
But based on the criticism of the game by some people even here, texture resolution will still have to be increased with next gen games - which means even more data.
Yes, but if the compression used is GPU friendly it has very different requirements as opposed to formats that are not usable for random access. i.e. jpeg2000 makes much smaller files(with same quality) as compared to jpeg which again is smaller than gpu friendly formats.
GPU friendly formats can be random accessed for each pixel whereas jpeg2000 needs to be uncompressed for whole image to access any pixel with full quality. This is the tradeoff why jpeg2000 files can be much smaller with same quality as GPU friendly formats.
Assuming rage uses gpu friendly format to store source data they can do much better diskspace wise with other compression. This ofcourse then has negative impact on cpu side which needs to decompress jpeg2000 and recompress to gpu friendly format. GPU friendly compression is needed to save both bandwidth and main ram. Perhaps next gen consoled can do this re/decompression without it becoming a bottle neck.
The Xbox 360 has done great without a standard hard drive. It allows for a cheap model later in the life cycle.
I agree with Laa-Yosh. Give a developer like John Carmack massive disc space and he will exploit it. As I remember things, Carmack wanted Geometry Compression in his next engine. This is why I think Microsoft will go with GE holograpic storage technique. The physical drive will be virtually the same as a standard Blu-Ray (hence cheap), but the magic is how the optical discs store information as holograms.
It really depends on what the source format is and how much it it being compressed(and I don't know that). But for jpeg2000 this is interesting
http://faculty.gvsu.edu/aboufade/web/wavelets/student_work/EF/comparison.html
Oh, and jpeg2000 decoding has gotten much better since 2001, jpeg on the other hand not
So there is hope of making source bitmaps smaller assuming we have cpu bandwidth to do the required processing.
I think you don't get my point. What I want is not higher compression ratios, but nealry lossless compression instead
What you are not getting is that depending what is the current algorithm for compression used it might be possible to get same quality with much smaller file size or better quality with same file size. The tradeoff being how much cpu power is used to get the source to memory and to new format that is gpu friendly. But all this is speculation as I don't know what compression rage uses. I'm just assuming rage uses gpu friendly format that is not optimized for size but random access(texturing by gpu).
There is a limit where you can't compress further without losing information. This is a certainty, no matter how advanced your algorithm is... and there's only room for improvements with lossy compression anyway.
I'd prefer to have enough storage so that no game has to go over that limit, or at least not by as much as we'll probably see with Rage.
As for Rage, there's been a rather detailed PDF, here's the link:
http://software.intel.com/en-us/articles/real-time-texture-streaming-decompression/
I'd prefer to have enough storage so that no game has to go over that limit, or at least not by as much as we'll probably see with Rage.
As for Rage, there's been a rather detailed PDF, here's the link:
http://software.intel.com/en-us/articles/real-time-texture-streaming-decompression/
On his blog Jon Olick (Naughty Dog & ID software programmer) posted some thoughts on high resolution textures.
http://olickspulpit.blogspot.com/2011/02/high-resolution-textures-field-guide.html
http://olickspulpit.blogspot.com/2011/02/high-resolution-textures-field-guide.html
Do you know how big the Rage high res pack for PC install ended up being ? If Carmack has his way with virtual voxel, we probably will need HVD. Also didn't Carmack mentioned virtual texturing isn't good for open world games like GTA, makes me wonder why and how it will effect its adoption. If we didn't get the capacity improvement, we probably just go to multiple discs like Xbox 360 version of Rage.
I remember the highest prototype bluray is 16-layers for 400 GB, not sure if that can be made into consumer product or if there are other development since. I am really hoping next gen consoles can house multiple 3.5" hdd instead of the single 2.5" hdd we have now. All installed with online activation, no need to insert disc anymore.
Rage isn't just virtual texturing, it's also unique texturing. The two work well together but they're not the same. Every game could benefit from virtual texturing because it conserves texture memory, even GTA.
I still don't expect voxels to take off, far too many limitations, combined together they'd seriously limit the possible game settings, features and even genres. And nowadays we see a lot of crossovers, just look at ME3 - it has on-rails driving sessions and large open spaces on top of the corridor shooting parts. If they were to change to voxels they'd have to throw out all the monumental stuff, they couldn't do engine cutscenes that take place in space, and so on.
It simply isn't worth to invest in a voxel engine - not even for id, now that they're aiming to run 2-3 game projects in paralel.
I still don't expect voxels to take off, far too many limitations, combined together they'd seriously limit the possible game settings, features and even genres. And nowadays we see a lot of crossovers, just look at ME3 - it has on-rails driving sessions and large open spaces on top of the corridor shooting parts. If they were to change to voxels they'd have to throw out all the monumental stuff, they couldn't do engine cutscenes that take place in space, and so on.
It simply isn't worth to invest in a voxel engine - not even for id, now that they're aiming to run 2-3 game projects in paralel.
I think I didn't make it explicit enough, so here is one more try
Current pipeline most likely is
load block of texture from disc, texture it to screen
what I'm proposing is
load block of texture from disc in size optimized form - decompress block, recompress block in gpu friendly format - texture it to screen.
Compression algorithms used by gpu's have very different requirements compared to size optimized formats. The main difference is that for gpu's it must be possible to access invidual pixels(or rather really small blocks) from texture without decompressing the whole image. As for size optimized stuff such as jpeg2000 all the pixels in image must be decompressed, invidual pixels or blocks cannot be accessed.
Above should make it possible to shrink the source asset on disc without loosing quality. It just introduces extra step of de and recompression needing cpu cycles(which most likely is a NO NO on current architectures which are cpu starved to begin with). It might be that this de/recompression is just too resource intensive in next gen too and we are stuck with whatever format gpu's can access natively. In here small(ish) cache on flash+ram storing the gpu optimized blocks might do wonders for cpu requirement.
If next gen really does need more than 50GB then BD-XL would allow 100GB and 128GB (yes, it's already on market, not just on laboratories).
http://www.cd-info.com/blu-ray/bd-xl/index.html
I think games with 50GB sizes would be a definitive no no for digital downloads. For example I have 250GB/month datacap due to comcast.
Last edited by a moderator:
after next gen we are around 2020+ and it's a pretty good question if digital distribution has already killed physical media near extinction for anything but legacy compatibility.
Haha you guys are so fast I can´t even seem to ninja edit before I have two replies Anyways I was thinking of slightly over 200nm wavelenghts. Can´t remember where I read it but I seem to recall going below 200nm is asking for trouble with regards to materials for the laser.
Anyways I was thinking of slightly over 200nm wavelenghts. Can´t remember where I read it but I seem to recall going below 200nm is asking for trouble with regards to materials for the laser.