By using this site, you agree to our Privacy Policy and our Terms of Use. Close
SvennoJ on 19 June 2013
Edit Thread
radha said:
SvennoJ said:
Machiavellian said:
 

If I am stating things that cannot be done then here is your proof of concept

https://en.wikipedia.org/wiki/Wolfenstein:_Ray_Traced

Since MS cloud can do the same thing, I am not speculation all that much.  Also if you just did a search raytrace and azure you would see that its also not as far fetch as you believe.  You do not need GPS transfer speed if you are compressing the video and sending it to the client.  MS cloud can do both.  It can render the scene and compress and send the results at the same time.

The pixar is just an example and whos to say that it cannot be scaled to real time.

It's true Gaika doesn't work like that, or not in that way when it was set up. Gaika doesn't render high end graphics to stay economically viable. Using distributed computing, ie multiple servers to ray trace the video for 1 client is a nice demo, but rather expensive to do for millions of CoD players.

Anyway this is not what MS is promoting. This runs the entire game in the cloud and you end up with no signal on connection hickups. Way more powerful rendering method then Gaika, yet the same problems with latency and compressed visuals. And ofcourse the more servers in your render pipeline, the higher the latency can become.


MS has not specifically stated that they can send processes from the rendering pipeline, they have been ambigous about it and giving that impression, to confuse people, becuase they cant offload anything relevant from the rendering pipeline to the cloud. Data transfer speed from the ps4 CPU to the RAM is 170GBytes per second, a 50 MBits per second internet connection moves 6 Mbytes per second, that is miserable, and upload speed is usuallay half the download speed. Cloud computing works for stuff that does not require real time processing.At 30 frames per second, the frame has to be ready in 0.0333 seconds, thats 33 miliseconds for the entire rendering pipeline, is all BS.

If you really believe they are rendering the entire in 33 milisecond and in the middle of that they are sending anything to the cloud for which you have to wait 20 to 100 miliseconds to come back (just latency and bandwidth limitations, asuming the server take 0 milisecond to get the result), you are completly blinded by you fanboyism.

They can stream assets but you will see popin depending on your connection speed.

In any case, anything that the Xbone can really do with distributes computing my Iphone can too, is not a magical technology it distributed computing, is what the ps3 use to do with folding at home, is all PR bs.

I completely agree with you, I was only pointing out that that Wolfenstein raytrace example is not economically viable and not what MS is after for games, since they still have to continue to work without it.

For now we only have 2 real world examples. Titanfall that also handles NPC behaviour in the cloud during multiplayer, pretty much like an MMO. That makes it easier to keep the games in sync but doesn't free up any significant resources. I don't believe for a second that launch games are maximizing all available cores and memory already, and if it can be off-loaded to cloud, it can be more easily off-loaded to a separate thread.
And there is Forza Drivatards, which actually adds a little extra strain to the console. It always has to upload your replay data for the server to analyze and then download custom AI routines before the race, to be run on the console.

Although MS is not after it right now lag free 30fps gameplay might be do-able gaika style some day. I have ping times of around 18ms to the nearest server. If that server can render and compress in 10ms (200fps rendering is possible, 5ms for render and 5ms for game updates and compression), and I have 1gbps internet that can download 5mbps worth of data (reasonable 720p image stream) in 5ms (200 times faster then 5mbps internet, since that would take 1000ms to download that image data), then I could get a comparable experience. Pipedream for now.

You don't actually need to hit 33ms to get the same experience as we have now though. http://www.eurogamer.net/articles/digitalfoundry-lag-factor-article?page=3 The best we're used to atm is 67ms. So you could get away with 100mbps internet or a bit better visuals. Although you still need to factor in the time it takes to poll the controller and prepare that data to send to the server, plus decoding and preparing the image for output to the display afterwards.

Ofcourse by the time this is possible on a large scale, local hardware can most likely do a better job again with 4K coming up next gen.