4 ms·
I agree! I don't understand these two separate arms of Google VR. Couldn't they do the processing power on the server side if you have a fiber connection?
by thebiglebrewski 10y ago
I agree! I don't understand these two separate arms of Google VR. Couldn't they do the processing power on the server side if you have a fiber connection?
- elaus 10y agoVR needs 90fps with only a few milliseconds of latency between moving your head and seeing the movement in your headset. Both can't be achieved by streaming from a remote server…
- shostack 10y agoYet... And once you can leverage the power of the cloud for rendering that makes VR a much more portable proposition.
- daeken 10y agoThis isn't really a "yet" case -- it's just physics. Even if you assume that this is all light speed and that congestion, routing latency, etc aren't a factor -- it's still impossible. You need to send your input (sensory and otherwise) to a server, then get a response, 90 times a second. That's 11ms round trip for the entire transaction. Light can only travel 2050mi in that time, meaning a maximum distance of 1025mi each way. And that's the easy part of this -- not to speak of the massive latency that video compression would add (at least 50ms), the transfer time of the data (if you go with a conservative 12Mbps bitrate and a perfect 50Mbit connection, that's another ~3ms there), etc. This would be possible if everyone had a VR render cluster nearby (<100mi) and no network congestion ever happened. Otherwise, not a chance.
- dahauns 10y agoVideo compression shouldn't introduce that kind of latency (I get <10ms with Steam in-home streaming), but other than that you're correct, of course.
- chronial 10y ago90fps gives you 11ms per frame. I have a 20 ms ping to the closest internet servers to me.