By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Gaming - Thoughts on storage and processing power

I was reading through Game Informer's rather long spiel on L.A. Noire today, and they had some very interesting things to say about Rockstar's new MoCap tech. Apparently, they are capturing the outside of the face and/or body whole, as opposed to capturing the skeleton as would be done traditionally and then have animators create the outside "body" of the model using the engine or using CGI. Now, according to Rockstar they have 200 Terrabytes of storage space ready for this capture data. This, according to GI, "allow[s] the player to to finally react to the characters as real actors in a way that even games like Uncharted 2 or Mass Effect haven't acheived." You can have actual, captured facial expressions from real actors, and the more actions they capture, the more realistic it gets. Goodbye Uncanny Valley? Maybe.

The question that this brings up to me is this (and I realize that this has only a tenuous link to what I wrote above, but I thought I would share and it stimulated my little grey cells): will games start to use LESS storage space as technology advances? I know, it sounds kind of crazy but bear with me. As capture technology improves, every item can become more and more independent, and engines will be able to increase flexibility. I'm not sure if this is happening yet, but what if eventually we don't have to have fixed objects consisting in millions of polygons stored on the disk/harddrive/cloud, but a brilliant AI "overbrain" that says, for instance; "This is Madison Ave, New York City, 1926; there's a cop here, punching a man in a fedora hat" and then renders the scene on the fly with nothing but some general size and shape data. No giant hi-rez skyscrapers, no extremely expensive facial expressions and having a hi-def image for the set of the mouth for every letter.

The point is, Rockstar isn't going to put those 200 TB on one disk, be it a DVD or a Blu-Ray. They'll have to have the processors handle what seems to be an unprecident amount of the rendering (it looks great in the screens in the mag, by the way). So, if we can do this now, when we hit that next gen will we need more than 50 GB's of data, as the processors get more and more complex and powerful and handle increasing amounts of rendering without preset objects? Maybe this is old news to everyone else, but it kind of blew my mind to think that game sizes could possibly (and I emphasize "possibly" here) get smaller. This of course would ease the transition to streaming and DD media, as a DVD size game on the current infrastructure is possible to download in a non-brain blowing out amount of time, and streaming would be cake, with less compression as well.

So, am I nuts, way behind the times or just slightly dull? Your thoughts.



Around the Network

I'm not really a tech guy so I have no idea on the matter. But all that stuff they're doing with motion capture is very interesting.



Well the top 3 users of storage space in games are in no particular order:

  • Full motion video
  • Sound
  • Textures

Full motion video is being transitioned away from due to the flexibility, time and money savings of doing everything in-engine. So you could probably erase this one in the future as engines and performance improves.

Sound and Textures aren't an easy one to get around. They are both already highly compressed and its difficult to use less of it depending what type of game being made. Mass Effect for example has a few dozen hours of recorded voice and other sound effects. I have also not seen much progress being made in terms of synthesisation of textures, however the compression of textures has improved over the years. The Xbox 360 has I believe a 10:1 texture compression ratio and the PS2 only had a 6:1 ratio, which means that 30GB of uncompressed textures take up 3GB on the Xbox 360 and 5GB on the PS2. Sound compression is a losing battle unfortunately.

So what does this mean? Most games even in the next generation will not require more than perhaps 6-8GB of space for conventional games. WRPGs more so than JRPGs will require more space as they have a lot of voice acting which they must store on disc.



Tease.

To elaborate a bit on what Squilliam said, FMVs are the main culprit of multi-disc and massive games. Let's take FFXIII as an example, most of the space used on the Bluray are what? Uncompressed FMVs. I've always (since the PS1 era) thought that FMVs are a waste of space and development time, and I thought they would be finally ditched this generation thanks to the real time graphics achieved by consoles... but here comes the Bluray with it's massive space and helps even more to spread their usage. This also applies for audio; I'd prefer to see developers working on new and innovative compression technology that matches uncompressed, instead of them going happy-go-lucky with uncompressed stuff.

Procedural generation, the programming technique that generates stuff randomly from code rather than having assets, has been around for many years, but I'm not sure it's widely supported. Actually, I can't immediately recall any game using it for something, maybe someone can refresh my memory. Perhaps it's harder to implement on games, needing more work or something, and that's why it's not used.

Also, I think the PS2 lacked any type of texture compression on hardware... because I'm sure the Xbox and GC had S3TC on hardware, which gave them a huge edge back in the day.

Today it's been a nerdy day for me hehe...



.kkrieger ( http://www.theprodukkt.com/kkrieger ). 96kb FPS. So yeah Squilliam and fazz you are right. Though I would like people to consider that this 96kb is the size of a game. How much actually logic code goes into a game? it's tiny.

It is not in the current business model to use procedual anything. It's unfortunate. I think with a combination of procedual generation and some hand created patterns we could come up with some really cool stuff. The demo weapons are sorta meh in design. It's just amazing grasphics for a proceduarlay generated game

no we won't see asset usage go down though :(

The motion capturing sounds likey they are going back to mesh animation rather than skeletal animation. Personally I like skeletal animation I suspect this won't go very far. Not that it won't look great. It probably will, but these projects are going to cost far more than it's worth.

The 3 Terra bytes is a great boasting number, but I suspect they will only use apx use 7 samples per animation out of 30 or so. Even after that it will probably be cleaned and end up using a tiny fraction.



Squilliam: On Vgcharts its a commonly accepted practice to twist the bounds of plausibility in order to support your argument or agenda so I think its pretty cool that this gives me the precedent to say whatever I damn well please.