One of the most interesting talks at London’s GDC (Games Developers Conference) this week came form one of the lesser known companies called Allegorithmic, who claim they will be able to reduce texture file sizes in games by up to 70%.
Their new programs, that they hope development artists will soon be using as an industry standard, are called ProFX and MaP Zone 2. Their ambition is to keep the graphical quality of game textures at the same standards as current games, whilst dramatically reducing the amount of data required for the game to work.
The implications of such a technology would be far reaching. As the current trend of digital distribution gains momentum a huge emphasis is being placed on games being made smaller and thus downloadable quicker. Their claim is that the current tool of choice for most games artists, Adobe Photoshop, is not ideally suited to making textures for games.
I was doubtful of this technology; however the company ran a demo that persuaded me otherwise. In the demo they had a bathroom full of beautiful textures, then with the flick of a button the bathroom took a more hellish look – all the while the textures looked the equal of Half Life 2.
The next demo was of a game that is due to come out for the XBOX Live Arcade called ‘Roboblitz’. Due to the requirement to get the game under 50MB, the developers needed to keep the textures as small in filesize as possible. Using the new texture system the overall size for all the textures was less than 280KB – watching the game (which runs on the Unreal 3 engine) I was amazed.
Confused by the fact that I hadn’t heard about this technology before, I spoke to one of the men behind it directly - Dr Sébastien Deguy. He assured me that there were no catches with his system, that if a game contained 1GB of textures he would be able to reduce that to 300MB and lose no quality. When I asked why everyone wasn’t using the program at the moment he explained it was due to people needing to be retrained in learning a new system. He was optimistic however, that soon all games companies will be using their new texture tools.
So what are the implications for you and I? In terms of traditionally packaged games that come in boxes, there probably won’t be much difference. Dr Deguy argues that if textures are smaller in file size and easier to create, then next-generation companies will be able to create even more textures for the games. We may then see a big leap forward in how richly detailed games are in the future as they triple the variety of textures the game includes.
The biggest impact however will be the benefits this will have to digital distribution. Games with texture quality and diversity matching Half Life 2 may soon be available in minutes of downloading rather than hours – for gamers this can only be a good thing.
I hear small arms is using this amazing tech to get it's 1gb of assets under the 50mb limit of xbox live arcade.
This new technology could also be a big boost for streaming in games, currently i know a lot developers complain they're being held back by read speeds on optical drives, with this technology they could get around that current bottle neck and load times for games greatly reduced.
So what, this 'RoboBlitz' is the first game to use this technology?
I think these people are on the right track if loading and downloading times are reduced by this
I call bullshit. It's extremely unlikely they've discovered some amazing new algorithm that magically makes all textures 70% smaller. More likely it's something to do with vector storage or procedurally generated stuff. Compression is a very, very, very mature field and magical discoveries just don't happen. If it were really as revolutionary as claimed here (70% smaller textures across the board) you'd be hearing a HELL of a lot about it. Only an XBLA game is using it too? Come on.
This paragraph in particular cries out as utter crap:
Confused by the fact that I hadn’t heard about this technology before, I spoke to one of the men behind it directly - Dr Sébastien Deguy. He assured me that there were no catches with his system, that if a game contained 1GB of textures he would be able to reduce that to 300MB and lose no quality. When I asked why everyone wasn’t using the program at the moment he explained it was due to people needing to be retrained in learning a new system. He was optimistic however, that soon all games companies will be using their new texture tools.
I'll believe it when I see it... if it could really be applied to textures, it could apply to all photos, which would make a big difference to digital cameras, etc. And going on from that, why couldn't it be applied to movies (e.g. for MJPEG movies).
I don’t understand why it requires retraining to learn a new system. I gather it’s not just a more efficient algorithm then? Does it require the textures be created in a particular way?
I just had a look at his website, looks like it's just advanced procedural techniques, nothing exciting in the slightest.
Not sure why you're getting all excited over this lestat, Microsoft's been pimping this method of game asset generation and storage since the first 360 dev kits came out.
Mech wrote:
I call bullshit. It's extremely unlikely they've discovered some amazing new algorithm that magically makes all textures 70% smaller. More likely it's something to do with vector storage or procedurally generated stuff. Compression is a very, very, very mature field and magical discoveries just don't happen. If it were really as revolutionary as claimed here (70% smaller textures across the board) you'd be hearing a HELL of a lot about it. Only an XBLA game is using it too? Come on.
I knew you'd be happy about this news.
A real game uses it mech, it's not bullshit, also compression algorithms are being improved all the time, VC-1 H264AVC say hi. I would think their software does use some very intellengent form of vector/procedural generation to replicate the image. What ever it's doing it's working a lot better than consine transformed based image compression.
I don’t understand why it requires retraining to learn a new system. I gather it’s not just a more efficient algorithm then? Does it require the textures be created in a particular way?
It requires new tools to be used in content creating and obviously for developers to integrate the decompression into their game. They have a website in that article you can go to and download demo tools to play with.
I just had a look at his website, looks like it's just advanced procedural techniques, nothing exciting in the slightest.
Oh, that answers my question.
But if they’ve created some good procedural methods that are fast, small, and yield impressive results then it’s still good news isn’t it? Microsoft may have been pimping such methodologies, but has anyone so far actually done much in this regard? Making it easier for others to implement procedurally generated textures is worthwhile, and if it works well then it will benefit everyone regardless of the platform…unless you want it to not work in order to justify BR, but that would be shooting yourself in the foot to spite your nose, which makes no sense at all.
mech wrote:I just had a look at his website, looks like it's just advanced procedural techniques, nothing exciting in the slightest.
Not sure why you're getting all excited over this lestat, Microsoft's been pimping this method of game asset generation and storage since the first 360 dev kits came out.
Nothing exciting? The theory is no longer theory but practice mech, i'd say that's pretty exciting. I knew you'd be all grumpy about this news.
Twit wrote:
unless you want it to not work in order to justify BR
A real game uses it mech, it's not bullshit, also compression algorithms are being improved all the time, VC-1 H264AVC say hi. I would think their software does use some very intellengent form of vector/procedural generation to replicate the image. What ever it's doing it's working a lot better than consine transformed based image compression.
VC-1 and co are all well documented and popular algorithms that don't make fancy claims like "will reduce all movies by 70% with the same quality".
Twit, it's a good thing, but it's already been around for a while.
The theory is no longer theory but practice mech, i'd say that's pretty exciting. I knew you'd be all grumpy about this news.
All anyone has to do is download .kkrieger to see what can be done with 96kb. It's been in practice for a while now.
I'm not sure why you're getting caught up in the hype that article's trying to generate. There's nothing new here.
Not grumpy about the "news" but the delivery of it. And you get pissed off at Sony for their hype? Pffft
lestat wrote:What ever it's doing it's working a lot better than consine transformed based image compression.
They're two different compression methods for two different image formats. Comparing them directly to each other is almost as retarded as misspelling cosine in a post that tries to paint yourself as some sort of math expert.
VC-1 and co are all well documented and popular algorithms that don't make fancy claims like "will reduce all movies by 70% with the same quality".
I remember ms making such claims when they were pimping their vc-1 codec and vc-1 does deliver the same quality at 50-70% of the original size of mpeg2.
Twit, it's a good thing, but it's already been around for a while.
All anyone has to do is download .kkrieger to see what can be done with 96kb.
lestat, I'm not sure why you're getting caught up in the hype that article's trying to generate. There's nothing new here.
The difference here mech is they provide a tool that takes any texture image and creates a procedural profile, before procedural techniques were hard coded and limited, their new technology is a very generic approach to creating any type of procedural texture you want from what i can see. So it does really open up the possibilities of using procedural textures for all your textures now.
Ignoring the space issue for a moment, wouldn't procedural textures (which I assume require some form of processing) take more of a hit on the CPU/GPU?
Talez wrote:
They're two different compression methods for two different image formats. Comparing them directly to each other is almost as retarded as misspelling cosine in a post that tries to paint yourself as some sort of math expert.
What the fuck are you on about? The end results is an image, something you can compare. Trust you to dwell on a typo.
Yes shaneus you'd need to use the cpu to decompress/create the images in ram. So graphics memory will still be a limitation since the gpu's can only texture from bitmaps/DXTC textures.
Last edited by lestat on 05 Oct 2006 11:38 am, edited 1 time in total.
Again lestat, that's just dandy, but this in particular is probably the line that irritated me the most.
He assured me that there were no catches with his system, that if a game contained 1GB of textures he would be able to reduce that to 300MB and lose no quality.
This is a lie. Some textures simply don't lend themselves to being procedurally generated well or at all.
Shaneus wrote:Ignoring the space issue for a moment, wouldn't procedural textures (which I assume require some form of processing) take more of a hit on the CPU/GPU?
Not necessarily, what .kkrieger does is generate them before you can play the game and stores them in memory. But for streaming stuff, then yeah definitely.
Mech wrote:
This is a lie. Some textures simply don't lend themselves to being procedurally generated well or at all.
Procedural generation can get way better compression rates than 70%. Like i mentioned the small arms guys have 1gb of texture content under 50mb apparently. So he's probably making the 70% claim with the assumption some textures just can't be successfully converted to a procedural generated image. But honestly when you look at game textures mech a lot of them are very procedural friendly.
Heh the reaction from resident sony fanboys is quite what i expected, I said this was the future and it's coming to fruition.
lestat, stop being such a fucking retard about this.
As stated, this kind of technology is great and all, but the way you and that article is pimping it is totally out of whack. PS3 will probably use this technology as well, although it won't have as much of a need for it. It's not like it's all roses, as Shaneus pointed out it has processing overheads so that for streaming content it's less effective. Just because - WOW - some guy created a decent procedurally generated texture tool, doesn't mean suddenly all games are going to be 70% smaller.
You're either being a shocking fanboy atm, or a troll, or both.
I think you're the one being a retard mech, i can see you're just having a big sulk because what i said in the bluray arguements i've had with you are coming true.
I have been seeing more and more games make use of procedural synthesis techniques. Just play games like TD:unlimited and just cause and see where things are headed. Also for streaming the saints row developers said their biggest bottleneck was optical drive speeds, this greatly reduces the streaming data read load, We have all these cores in cpu's now it's time we put them to use.
Hahahaha, man you just don't get it do you? I don't give a shit about whether this affects Blu-Ray or not.
Edit: and regardless, some smaller textures does not make a game 70% smaller. Other assets take up a bunch of space (especially movies) and aren't affected by this.
And you're totally missing the point, you're acting like this is some amazing breakthrough and that it's totally without any negatives at all. It's misrepresenting the technology majorly, both your take on it and the original article.
Then what do you give a shit about? Real games are now using this technology, quite a few current games are using procedural asset/world generation. Why are you so sceptical? You're always so positive when talking about ps3 potential.
I'm not skeptical, I know it can work, I've even mentioned examples, you're still not getting it.
Again...
mech wrote:[Y]ou're acting like this is some amazing breakthrough and that it's totally without any negatives at all. It's misrepresenting the technology majorly, both your take on it and the original article.
