This is true and possible. Consoles could cloud-compute non-time-critical events. Again, that's not how you started this conversation though. If we get past the idea of realtime distributed computing of a console's realtime game requirements over several units scattered across the internet, which is impossible and will remain so without incredible interconnections that don't introduce 100+ ms lag, there's potential for consoles to create richer worlds than a single box can manage. However, that'll be limited to certain game types and won't be appropriate for many game types, at which point you are basically talking about persistent worlds rather than realtime distributed computing. And that's something far easier to manage with servers than relying on other people's consoles. For one, some of us switch off our consoles. For another, before the console could work on your game it'd need to be populated with sufficient data, so you'd need a system that searches for available console nodes, distributes jobs, uploads maybe hundreds or thousands of megabytes data for that console to work on (you can't limit it to whoever owns the game or you'd have no idea how much performance was available) which'll have to be supplied by the game-playing consoles IO, so not only is it streaming graphics for the player but also streaming graphics to nodes 1 through 4 to create algorithmic cities,, and be able to replace that node immediately when the owner switches the console on to play their own game.
Things like Gaikai are server based. Any distributed workloads are handled locally. Seti/Folding are non-realtime problems. The infrastructure isn't there for true distributed computing and won't be even starting for a decade IMO. The latency in distributing through internet servers isn't going to go away no matter how fast your final residential BB connections are, and that means 100+ ms delays on data from any node. Designing a game around such random packets is going to massively limit what jobs you can use distributed computing for.
thanks for the your answer
I think about two different situation,
the first is when a lot of consoles are always online and with game data inside the internal disk (but the whole disk can be hosted in the cloud)
and the second when the critical mass of always online console is not yet here
Both the situations could, maybe, be solved with a dynamic balancing between the console cloud, the servers and the local console;
things that need fast interactions with the player are best locally, in the near zone the servers and for distant places, the xbox cloud
A sort of LOD with fast interactive reactions zone rather Z distance
aside of this I see how well can perform on the geometry the xbox cloud, at the load of level the XC can provide with heavy tessellation, deformation, computed animations and so on, there're a lot of effects that don't depends so much on latency, such as water simulation, snow/rain simulation, and given 5-10 TF from server+xb-cloud, you can calculate new shadow maps (night/day or other cases), modify partially textures in almost realtime (maybe is possible to slowly feed the 5.5 GB with some assets from the cloud as you proceed in the level)
you know, this should be a new approach and a lot of problem will jump from new things, and there's some type of games that fits better just like open worlds games, or racing games too, or games based on turns as RPG where even 200 ms of latency is not a problem, and others that needs some deep thoughts and smarter approach as FPS, 3d platform and so on
.
