Sakkura :
bit_user :
That's why it's not so simple, and why MIT is fussing about with millimeter waves. I think it's also pretty safe to say that some simple form of video compression will be instrumental. I also expect ATW will migrate into the HMD, itself.
* Presumably, they could halve this by alternately transmitting & displaying frames for the left & right eyes.
Compression is not the answer, since that adds latency too. Foveated rendering would help, as the total rendered resolution would be reduced.
Alternating between left and right eye would probably not be a comfortable solution either.
Well, I say that having seen the insides of x264 (a popular, opensource H.264 encoder) and modifying it. There are techniques which don't offer the same benefits of H.264, but are much cheaper & lower-latency. If you're only looking to get a factor of 2 or 3, it's doable with minimal quality loss and only a few scanlines worth of extra latency.
As for alternating L/R, this is how LCD shutter glasses work. But they traditionally use a far lower framerate. If we're talking about 90 Hz, it seems to me that should be viable. Ideally, you'd render the eyes separately, and send each one as it finishes. I realize this involves a bit more work, especially with Nvidia having made a big fuss about how they can now simultaneously project geometry onto multiple image planes. But the work to actually shade the pixels doesn't really change, whether you render both eyes simultaneously or alternately.
Even if the eyes are rendered simultaneously, the ATW really should happen in the HMD, right before the image is displayed to the respective eye. This would be beneficial no matter whether you're using a wire or not. The ATW can actually happen as the data is received by the HMD, using the extrapolated pose at the time that display is expected to happen.
BTW, I thought of mentioning foveated rendering, but figured that'll probably happen regardless. I think FOVE said it results in approximately a 2x speedup. Of course, the potential exists for greater improvements, but depends on even more accurate eye tracking & modeling.
In summary: the potential for low-latency wireless exists, but it's not trivial and needs to be engineered as an end-to-end system. I expect to see a good effort from at least one of the big players, in 2017.