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

I remember that demo and these were the results
http://www.ppsloan.org/publications/Crassin13Cloud.pdf

Wow, thanks for doing the math.  Yea, like I say, this demo is already several years old, and before current leves of compression (side note: aren't the current twings getting  compression .h265 in software updates?  I know they talk about it on the Xbox One, but I would think the PS4 could do it as well.  It's the one that allows you to stream 4k resolution in the same bandwith as today's 1080p, basically 4x better than today's for streaming.)    So they don't need to send the full frame at 30fps, but a partal map - or prerender of what they local computer needs to render.  There might be 10,000 objects falling, but only 3,000 in your view.  Then the local on bord GPU on the Xbox One would render only what your character sees.

And I'm wondering if this is one of the things they really needed DirectX 12 to do - with it's any CPU can address the GPU expansion.  Thus dismissing a major bottleneck that it would need to process gobs of graphic data.  As others have said, it's one of the things they should have shown at launch (But I don't think it was near ready.)

PS People forget that World of Warcraft use to work on a phone modem.  The servers are doing the work, the info sent to your console can be much less than one assumes.

.h265 can get close to half the file size for video compression compared to .h264. It is very CPU intensive though. The consoles can do it, yet it will take more CPU power. For encoding it can require 10x the cpu power of .h264, decoding doubles the cpu usage. (And the consoles probably already have very efficient built-in .h264 codecs not burndoning the CPU)
.h265 works better by removing more low fequency detail you don't notice clearly while watching moving video. How suited it is to light maps remains to be seen. Blown up accross a wall you might notice weirdly even patches.

I guess the server can limit the detail to what's close to you, yet due to lag it can't exactly limit it to what's in view. Stuff wouldn't be updated or missing when you turn around. The client will have to sort out what's in view like normal after the geometry of the world has been updated. Normally open world games have a cache of the direct area around you with detail reducing the further away from you. It would be interesting to see how they solve fast movement. Pop up already occurs while streaming from HDD, now it will also have to download all the geometry changes other players caused while you speed along.
We don't know how big the levels will be, maybe they'll keep the basic geometry of the whole level in memory while the server continously sends all changes for the whole area. The game doesn't have much in the way of textures so it shouldn't be a problem to allocate a lot of memory to level geometry.