The Xbone

[=sw=warlord]

There was a time people thought it wasn't feasible to put the memory controller on the CPU, there was also a time where people thought offloading the video processing to a dedicated controller was not worth it.
Both of those assumptions have since been proven wrong, with that I'm reasonably sure as connections speed up offloading work to "cloud" will be worth doing more and more.

[Skyline]

It's 3D on the server's end, but all your machine is doing is largely 2D work. By removing the colour information from the transmission, you cut down on packet size; you could even let the 3D world have a low resolution and use the identifiers and 2D painting to mask it with the player none the wiser. It's a new concept

You lose too much information if all your doing is sending 2D data to the client, paint by numbers isn't going to look good and you might as well just do the paint by numbers on the server which then just becomes another streaming service like onLive. Everything is pushing toward higher resolution why would you want to make it a lower resolution (will probably be worse off than the current generation).

[Cortexian]

Unless I can play without an always-on audio/visual recording device pointed at me I still won't even consider buying one.

[DEElekgolo]

( ͡° ͜ʖ ͡°)

[Warsaw]

@Freelancer:
Well me neither, but still.

@Skyline:
Because your local hardware can't necessarily do high-resolution geometry AND high resolution textures AND high fidelity lighting, that's why. The numbers can be more or less densely packed, and you'd have an algorithm that smooths out the final image. The whole point is that you transmit only the bare minimum information needed to draw that 2D image; but you first need to calculate the detailed image before you can decide where to put your info points. This is why I said voxels (which I still maintain are an illusion) might become a thing again. I've seen voxel animation, and the only reason it looked bad was because the game came out in the mid-90s and PCs simply didn't have the horsepower. We now have the horsepower.

[Phopojijo]

Going back to "cloud processing", you could also do things like re-render cubemaps or lightmaps while destruction happens and blast them out to all applicable clients before they're needed.

The problem is that most of this stuff could have just been crammed in spare cycles of a few light-load frames. I don't know...

[Warsaw]

It's all about the detail. If you cram it into the spare cycles during a light load frame, you have to make do with whatever elbow room you have at that moment so you lower your fidelity target as a precaution against overreaching. If you offload it, latency becomes your main concern but that's far more consistent than the local hardware's dynamic load. You can therefore potentially set your targets higher and even scale them with the connection speed.

[Dwood]

Personally I think the bandwidth will only really be able to be sustained by ridiculous services such as google fiber. HOORAY Google fiber.

[Pooky]

@Freelancer:
Well me neither, but still.

@Skyline:
Because your local hardware can't necessarily do high-resolution geometry AND high resolution textures AND high fidelity lighting, that's why. The numbers can be more or less densely packed, and you'd have an algorithm that smooths out the final image. The whole point is that you transmit only the bare minimum information needed to draw that 2D image; but you first need to calculate the detailed image before you can decide where to put your info points. This is why I said voxels (which I still maintain are an illusion) might become a thing again. I've seen voxel animation, and the only reason it looked bad was because the game came out in the mid-90s and PCs simply didn't have the horsepower. We now have the horsepower.

But will it be more fun than Super Mario Brothers?

[Warsaw]

That's not hard to do...