Madword said:
Rendering of graphics/displays is already done using onLive, and even application services like Terminal Server and Citrix, unfortunately as onlive has shown (and anyone who uses terminal services/citrix) its just not fast enough yet for high quality graphics. But in terms of data, graphics is certainly a higher bandwidth than small amounts of processed data. |
The data is actually easily a lot more then a lossy compressed video stream.
For example it could be great for animating large AI crowds all with individual behaviour.
Let's say 2000 npcs running around, dead rising / dynasty warriors style. All need position updates with orientation, speed vectors, animation state. Let's say you can get that down to 16 bytes per npc after compression and stick to 30 fps, equals 7.3 mbps just for that.
A detailed lighting map for a HD game, or volumetric fog for a large area is not just a little bit of data. The lighting map can be downloaded in advance, but don't you want your fog, smoke explosions, fluids to move, huge amount of data. And even with lighting there are better real time solutions available now that take all dynamic lights into account as well. We are finally getting rid of pre-baked lighting, and now we're going to bring it back? For example, how is it going to work with destructable terrain?
It's great for non realtime calculations that you download in the background like level generation or procedural terrain as long as you don't allow the player to move too fast.
It would be awesome if you could draw a track on a google earth map and the server distributes that job to create a detailed custom made real world race track in a few seconds, then download that track in half a minute and go.
You could talk to NPC's with Kinect, the server does advanced speech recognition and natural language processing and synthesizes a human like voice with custom responses. That's definitely a very expensive task with low data transfer. Pretty difficult to make though.
Anyway the bandwidth is a bigger issue then the latency for most things.







