Re:Google Earth VR
by mdrzn on Wednesday November 16, 2016 @11:28AM
(#12968526)
Why only on HTC Vive? Does this require huge computing power, or is it because of the controllers?
I hope it'll become available on the Google Daydream phones.
[ Parent
| on HN ]
They had to do a crazy amount of optimization just to the Earth engine to make it work in 90fps VR.
Here's a talk one of the devs gave at SIGGRAPH a few months ago about how they made it work: https://vimeo.com/177549565
Weird because once you get it running via the hack it runs perfectly fine. I'm running it on a DK2 and the only issue is controller emulation.
My bet would be the controllers, which would mean a Rift compatible release would be imminent when the Touch controllers are officially released in a month.
Most OpenVR/SteamVR titles more or less "just work" with Rift + Touch. Obviously the different controllers mean input mappings might be weird or lacking an equivalent and the in-game representation of the controller might be wrong, but the basics work as well as you could reasonably expect.
Same with Daydream - odd that they didn't keep earth for the release.
I've been using DK2 + Leap Motion to play some of the simpler Vive only titles. I'll try it and report back.
Please do! I'd like to avoid cluttering my desk with too many more gadgets.
It works on my DK2 when I used this workaround. https://github.com/Shockfire/FakeVive
It's pretty tough to use with the Leap Motion. However - you can run the tours via the mouse on the desktop. (Also - in another comment I mentioned you can edit the tours in a text editor so that might be enough to keep people happy.
Re:Google Earth VR
by thebiglebrewski on Wednesday November 16, 2016 @04:49PM
(#12971905)
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?
VR 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…
Yet... And once you can leverage the power of the cloud for rendering that makes VR a much more portable proposition.
This 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.
Video 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.
90fps gives you 11ms per frame. I have a 20 ms ping to the closest internet servers to me.
The roomscale VR that let's you walk around in cities as a giant is really the killer feature of this tech demo.
I'm confused about this as well - this could be a great application for "casual" vr like daydream.
As a title, sure, but that's about it. As a product, it would be a fundamentally different thing. That's inevitable when one system offers full position tracking and 300 watts of power while the other tracks rotational motion only and runs on 3 watts.