The status of ViiWok

[FingerFlinger]

That's all great, but you still haven't described how you measured the latency. Certainly the video evidence (unclear as it is) is not sufficient for you to be able to claim 0-5ms.

[Attreyu]

It's deffinately a gross estimation on my part. I couldn't tell the difference at some point (between the TV and the tablet). It's quite exciting really.

[zalo]

Idea time! zalo's idea for latency measurement on the cheep:

1. Make a video in after effects where every other frame is a different solid color (have about 10 discrete colors).

Like this (except it needs to be adjusted for 60fps: https://dl.dropboxusercontent.com/u/8557629/wut.gif )

2. Play the video at full speed on the computer, and stream it to the phablet.

3. Grab a friend with an iPhone 5S.

4. Record a video of the phablet and monitor playing in sync at 120 fps.

5. Look at the color on one frame, and count how many frames it takes for the phablet to reflect this color.

6. Each frame at 120 fps is 8.333ms.

7. Subtract 6-8ms or whatever the latency on your monitor is. Add 6-8ms or whatever latency your phablet screen is.

8. Throw out results because there are probably far better ways of measuring latency (the best of which is Oculus's impending latency tester).

-------------------------------

From what I could see of your videos, latency averaged between 100 and 300 ms. Admirable, certainly, but not 0-5ms. Not even close.

[FingerFlinger]

Seconded!

[Krenzo]

Setup a photodiode in front of each display and feed their output into an oscilloscope. Flash the display from black to white. Measure the time difference between each photodiode's voltage spike on the oscilloscope's display. Even a cheap oscilloscope will give you sub-ms resolution.

[remosito]

wow! quite impressive synchronicity between screen and tablet.

Main problem I see with it though wireless bandwidth just ain't there!

8bit per color full hd is 3Gbps. That's an order of magnitude above current wireless tech that is using general purpose network protocols and not heavily optimized stuff like the wireless hdmi solutions use.... I doubt artifact free compression with such a gain exists

Edit: according to wikipedia lossless video codecs achieve factor 3. Another factor 3 without noticeable artifacts? Is that possible?



as for teethered mode b2)
452MB/s or 452mbps????

what I don't really grasp is how one pipes the gpu output to a usb connector. can this be done in windows?
Not saying it's not possible. Just not knowing and curious...

[zalo]

To Krenzo's post: Ooh, that's a good idea. Use one on the TV and the other on the tablet.

After googling it, I found a blur busters article that uses an arduino instead of an oscilloscope to save money.

[Krenzo]

Sure, you could use an Arduino and count the cycles between each diode's spike and still get sub-ms resolution.

[Benjimoron]

Can't you just use a 60fps camera and then count how many frames out of sync it is? Could be done by suddenly moving the in game character, watch it back frame by frame, each frame until it reacts is 16 millisecs? So if it's what he says then it'd basically be none, if there's a frame or more noticable difference then it's more than he says.

Apologies if I'm not understanding the problem or made a mistake in my calcs.

[Benjimoron]

First test, on the first video. 10 frames difference between the images, 30fps video, therefore 333ms.

Later in the video, approx 3 frames difference, 100ms.



Second video, first test, 29 frames between tablet movement and movement on tablet screen. Basically 1 whole second.

[Attreyu]

Almost half a second ? I don't know about that. Look at the gun's movements on the two displays. But if you say so...

[Benjimoron]

Almost half a second ? I don't know about that. Look at the gun's movements on the two displays. But if you say so...

I don't get why you say you don't know about that. Try playing your own video frame by frame and actually see the results. I've updated with some more above, one better, one worse.

Second video, second test. 8 frames between movements (not including gyro lag), 266ms.

[Attreyu]

as for teethered mode b2)
452MB/s or 452mbps????

what I don't really grasp is how one pipes the gpu output to a usb connector. can this be done in windows?
Not saying it's not possible. Just not knowing and curious...

452 Mbps as in almost 60 MB/s

It's simple: the screen is captured frame by frame (like when you record something, or press Print Screen), compressed a little and sent over the network/USB/WiFi to the receiver which decompresses the package and displays it.

This is not the same thing as using a video connector cable between the GPU and a classical monitor or a TV. Here we are talking about streaming. It's like watching movies on the local network/RDPing into a network node.

[Attreyu]

Almost half a second ? I don't know about that. Look at the gun's movements on the two displays. But if you say so...

I don't get why you say you don't know about that. Try playing your own video frame by frame and actually see the results. I've updated with some more above, one better, one worse.

Because it doesn't "feel" like a half-a-second gap. It's almost instantaneous.

[Benjimoron]

Almost half a second ? I don't know about that. Look at the gun's movements on the two displays. But if you say so...

I don't get why you say you don't know about that. Try playing your own video frame by frame and actually see the results. I've updated with some more above, one better, one worse.

Because it doesn't "feel" like a half-a-second gap. It's almost instantaneous.

It doesn't really matter what it feels like to you, it's the figures I've put up, that's a scientific test, including working out and multiple results. Just because something feels "almost instantaneous" to you, doesn't mean that it is.

[Benjimoron]

Next two tests, 7 frames and 8 frames (not including gyro lag), approx 250ms.

I'm bored now. You still think this is 0-5ms?

[Attreyu]

It doesn't really matter what it feels like to you, it's the figures I've put up, that's a scientific test, including working out and multiple results. Just because something feels "almost instantaneous" to you, doesn't mean that it is.

I agree! On the other hand, I personally don't give a flying screw about the scientific numbers when I play a game, it has to feel playable. And this did, I could actually play it without issues from the tablet, which is what counts, at least to me...

[Attreyu]

Next two tests, 7 frames and 8 frames (not including gyro lag), approx 250ms.

I'm bored now. You still think this is 0-5ms?

I got bored 3 days ago when the 5th nerd asked about the 5ms latency

[Benjimoron]

It doesn't really matter what it feels like to you, it's the figures I've put up, that's a scientific test, including working out and multiple results. Just because something feels "almost instantaneous" to you, doesn't mean that it is.

I agree! On the other hand, I personally don't give a flying screw about the scientific numbers when I play a game, it has to feel playable. And this did, I could actually play it without issues from the tablet, which is what counts, at least to me...

Ok, but that's not what you posted. If the figure wasn't important to you, why include it, how did you measure it?

[Benjimoron]

Next two tests, 7 frames and 8 frames (not including gyro lag), approx 250ms.

I'm bored now. You still think this is 0-5ms?

I got bored 3 days ago when the 5th nerd asked about the 5ms latency

I'll take that as an acceptance that it's not what you said it was. Some proof would be good, otherwise people can only go on what they see.

[TheHolyChicken]

Almost half a second ? I don't know about that. Look at the gun's movements on the two displays. But if you say so...

I don't get why you say you don't know about that. Try playing your own video frame by frame and actually see the results. I've updated with some more above, one better, one worse.

Because it doesn't "feel" like a half-a-second gap. It's almost instantaneous.

The reason Oculus are releasing a latency tester is because attempting to test latency based on how something "feels" is a completely impractical and non-scientific way of measurement. Human senses aren't acute enough to reliably estimate what the latency is. You need to have an impartial and objective method of testing to determine if you're on the right track.

The same applies here with your streaming. Claiming 0-5ms latency because it "feels pretty instantaneous", without even a shred of evidence to back it up, only serves to completely discredit yourself and all your work. All previous feelings of intrigue and promise have vanished, replaced only by disdain and distrust, and I'm sure that sentiment is not going to be held by me alone.

On the other hand, I personally don't give a flying screw about the scientific numbers when I play a game, it has to feel playable. And this did, I could actually play it without issues from the tablet, which is what counts, at least to me...

Latency is critically, critically important for VR. What might look decent on a screen or a tablet can be a nausea-inducing vomitvest in VR. The latency figures for the wireless streaming alone is already an order of magnitude higher than would be permissible, and that's before we even consider the entire motion-to-photon latency.

[FingerFlinger]

Deleted in favor of a PM

[Attreyu]

Next two tests, 7 frames and 8 frames (not including gyro lag), approx 250ms.

I'm bored now. You still think this is 0-5ms?

I got bored 3 days ago when the 5th nerd asked about the 5ms latency

I'll take that as an acceptance that it's not what you said it was. Some proof would be good, otherwise people can only go on what they see.

I agree. So the minimum that can be observed using scientific methods is around 100ms ? Nice.

[Benjimoron]

I agree. So the minimum that can be observed using scientific methods is around 100ms ? Nice.

Yes it would seem around 100ms was the best result I could find from what you've posted. Important to note though is that was just display lag between the 2 screen. If you include controller (gyros etc) lag then the best result I could find was around 1 second!

[Attreyu]

The reason Oculus are releasing a latency tester is because attempting to test latency based on how something "feels" is a completely impractical and non-scientific way of measurement. Human senses aren't acute enough to reliably estimate what the latency is. You need to have an impartial and objective method of testing to determine if you're on the right track.

I don't agree on the absolute necessity for "scientific" measurements when you're playing games. It's not rocket science. I know I'm on the right track when I can play a shooter wirelessly without obvious latency.

The same applies here with your streaming. Claiming 0-5ms latency because it "feels pretty instantaneous", without even a shred of evidence to back it up, only serves to completely discredit yourself and all your work. All previous feelings of intrigue and promise have vanished, replaced only by disdain and distrust, and I'm sure that sentiment is not going to be held by me alone.

Hold your horses dude, with the whole "discredit yourself and your work". I'm not the creator of Splashtop THD. I just test the thing to see which of the available solutions are best suited to the scope of a wireless HMD, and in what conditions. I have described what I found in pretty specific terms. And my conclusion was that it is very close to what would be needed for a decent gaming experience. It's not without a "shred of evidence", I have played it and recorded it. Stop being agressive.

Latency is critically, critically important for VR. What might look decent on a screen or a tablet can be a nausea-inducing vomitvest in VR. The latency figures for the wireless streaming alone is already many factors higher than would be permissible, and that's before we even consider the entire motion-to-photon latency.

I agree.

[Benjimoron]

To be fair, Theholychicken is absolutely right. You make yourself look untrustworthy when you claim things that are vastly incorrect. If you'd come on here and said that you tested it and have now realised that you made a mistake and what you thought was 5ms was actually 500ms then you might have been alright.

But to make that claim and then say that you didn't test it, you just thought it was probably that as it "felt good", just isn't good enough.

If you were buying something, and the creator claimed it was a certain spec, you'd want to be able to trust them. Now consider how well people can trust you from here on.

[Attreyu]

I agree. So the minimum that can be observed using scientific methods is around 100ms ? Nice.

Yes it would seem around 100ms was the best result I could find from what you've posted. Important to note though is that was just display lag between the 2 screen. If you include controller (gyros etc) lag then the best result I could find was around 1 second!

I've had an exchange of PMs with Palmer and he said that the best that can be achieved using Splashtop is about 50ms, that is, in ideal conditions. So at least in theory, and with current technologies at hand, we can't have proper wireless VR. It's good to know, in order to stop further developments until better WiDi standards reach the market.

About the gyros, I really don't know exactly how they impact, Palmer told me about a couple of ms at best, probably around 10. There are alternatives to using the integrated 6DOF though. I'll test it again using better hardware.

So I guess that settles this and the Vrase issue at the same time, regarding gaming.

I think we can all agree now that using a mobile dispay (smartphone or tablet) for a perfect VR gaming experience is beyond our means at this time. I was curious about it, I tested it thoroughly using the best speed conditions I could find and I don't think it can be expanded to limits under 10ms. Moving on.

[Benjimoron]

Moving on.

Quickly sweep that one under the carpet....

Also movin' on. Cya.

[Attreyu]

To be fair, Theholychicken is absolutely right. You make yourself look untrustworthy when you claim things that are vastly incorrect. If you'd come on here and said that you tested it and have now realised that you made a mistake and what you thought was 5ms was actually 500ms then you might have been alright.

But to make that claim and then say that you didn't test it, you just thought it was probably that as it "felt good", just isn't good enough.

If you were buying something, and the creator claimed it was a certain spec, you'd want to be able to trust them. Now consider how well people can trust you from here on.

You are right. I can say in my deffense that my intention wasn't to publicize anything, it was more like a lack of knowledge and technical data regarding the measurements of latency, during testing of a 3rd party technology. I would never lie about something I would intend to sell, that was constructed exclusively by me. I couldn't anyway, since with an ODT it can be clearly seen if it works or not. And there would be 3rd party, objective testers before any hypotetical launch that would attest to its problems.

So I appologize if I caused any problems. I was excited to see it behave so well. I still am. But the numbers are not low enough and I accept it.

[Zoide]

Attreyu: That's OK. I think we all learned something new. Thanks anyway for showing us your cool projects

Now let's get back to the omni treadmill! I hope you don't mind me advertising in your name, but people should know that they can back your preliminary investigations prior to the actual Indiegogo/Kickstarter campaign:

http://gogetfunding.com/project/viiwok-deva

I put it in a small donation but I think we need more people to join

[Attreyu]

Thank you Andrés ! I promise to make sure to give you a nice discount when it's oficially launched !

As I have explained on the spanish forum, there isn't really a massive need for funding at the moment, it's just a way of seeing if it can gather some interest and as a backup solution.

[TheHolyChicken]

Sorry, perhaps my language was a little too strong. I was just trying to communicate how I felt. The claims that "a cheap 3D HMD is within range, with fullHD resolution and without cables" had me excited and intrigued, and so I had an equally strong negative reaction when the claims were revealed to be baseless.

[JoeReMi]

So Attreyu, just to be clear, do you still think the consumer rift will not ever make it to market due to obsolescence in the face of mobile phone/tablet streaming solutions?

[Attreyu]

I think that streaming solutions will only get better and that mobile devices will eventually prevail, being cheaper (adapters, or even with integrated tablets) - I'm basing my oppinion on the launching of USB 3.0/WiFi Direct devices with high enough transfer rates plus powerful GPUs. That would bring the latency to a minimum eventually. I still think that a final version of the Rift will not be launched and will be available only as incremental updates for some time, because the wireless/display combos in mobile systems will keep the pace high enough. Maybe in time, Oculus will concentrate gradually more in side VR technologies, on software solutions/SDKs maybe, with Carmack now in the team.

VR is not only about gaming, there are fields where absolute minimum latency is not that important.

It remains to be seen.

[PalmerTech]

I still think that a final version of the Rift will not be launched and will be available only as incremental updates for some time, because the wireless/display combos in mobile systems will keep the pace high enough.

VR is not only about gaming, there are fields where absolute minimum latency is not that important.

I would have to disagree. The roadmaps and designs for major phone companies are often locked for years before they launch, it takes a long time to design, test, and manufactures millions of phones. Unless smartphone manufacturers have already decided that their next gen hardware will have well calibrated tracker systems capable of positional tracking and low latency video streaming hardware, there is just no way for it to happen before our consumer launch. If our consumer launch was 2 years out you might be somewhat right, but it is not.

VR is not only about gaming, but minimal latency is critical for ALL VR, not just gaming. Even slow paced applications like architecture visualization and virtual movie theaters require high frame rates and low latency to avoid making people feel ill. It is hard enough to get the pipeline short enough with cables, adding the latency of network based video streaming won't be realistic until at least several years after our consumer product launches.

[Attreyu]

Even slow paced applications like architecture visualization and virtual movie theaters require high frame rates and low latency to avoid making people feel ill. It is hard enough to get the pipeline short enough with cables, adding the latency of network based video streaming won't be realistic until at least several years after our consumer product launches.

Apart from the first paragraph, which you obviously know better than all in here, the thing that really troubles me is this "latency making people ill".

I have yet to test the Rift, my only experiences so far have been with DIY designs, obviously quite rough and everything, but one of the apps I have tried, there was no computer involved, only the app on the tablet. It was made by someone using the Dive SDK, it's about a rollercoaster.

Since the app was completely integrated, in SBS, there was no latency whatsoever and the sensors worked perfectly. Still, after about 1 minute I've begun to fill sick in the stomach, as if my brain was trying to accomodate with not being out there in the carusel, but still experiencing the broken equilibrium in my ears. Extremely weird and realistic sensation.

I've mentioned it because it's one of the cases where sickness is induced, preserved and wanted. I've yet to try a game using a helmet, in order to see what kind of sensations I experience due to the latency factor being introduced.

About the network based streaming, do you think that it would take years for hi-transfer mediums to appear ? We already have USB 3.0 on both ends.

[cybereality]

Latency, and also performance, has a huge impact on cyber-sickness.

I know from first hand experience, if a demo is running really slow (lets say sub-30fps) or the latency is large, it will quickly make me feel uneasy. Even if the game is running at the full 60fps there could be other latency involved (for example if an injection driver uses mouse emulation, or the game adds mouse smoothing) and this will also add to the perceived sickness. There is no question here, it's a huge issue.

[PalmerTech]

Apart from the first paragraph, which you obviously know better than all in here, the thing that really troubles me is this "latency making people ill".

Since the app was completely integrated, in SBS, there was no latency whatsoever and the sensors worked perfectly. Still, after about 1 minute I've begun to fill sick in the stomach

About the network based streaming, do you think that it would take years for hi-transfer mediums to appear ? We already have USB 3.0 on both ends.

Latency does make people ill. It is not an opinion, there is a huge body of research on simulator sickness to back it up.

It does not matter that the app was "completely integrated", and it is just not true to say that there is no latency whatsoever. Running natively removes the wireless streaming lag, but it does not account for the slow/inaccurate/heavily filtered trackers in phones, the time it takes to render the scene, and the multiple frames of buffering that Android introduces. The fact that you felt ill quickly falls perfectly in line with that, I have not seen a mobile demo outside of Oculus that gets sub-50ms latency! The threshold for simulator sickness is well below your ability to consciously detect lag, something can look fine at first glance and still have enough latency to make people sick. That is one of the main reasons we created our latency tester hardware.

Dedicated wireless video hardware like WiHD or WHDI is the only reasonable way to stream video these days, but it is power hungry and expensive. I don't expect things to change significantly for at least a few years.

[Attreyu]

Ok, you've made a believer out of me

[the_wretched]

Latency, and also performance, has a huge impact on cyber-sickness.

I know from first hand experience, if a demo is running really slow (lets say sub-30fps) or the latency is large, it will quickly make me feel uneasy. Even if the game is running at the full 60fps there could be other latency involved (for example if an injection driver uses mouse emulation, or the game adds mouse smoothing) and this will also add to the perceived sickness. There is no question here, it's a huge issue.

Overall I've been pretty lucky and am not one to suffer from lots of nausea when using the Rift. But over the months I have developed a sense for feeling something "is wrong" (latency/framerate issues) and I tend to stop those experiences/demos/games almost instantly because they not onyl make me feel weird, they tend to ruin my mood for VR - sometimes for days.

That's also why I argued against the whole "phone/tablet as a VR device" repeatedly. Sure those devices are capable of some sort of VR, but the experience is most likely going to be average at best. I'm actually a little afraid that many people are going to try VR on such a device, have a bad experience and then say "nah I got sick really fast, this sucks even worse than the whole 3D movie fad"...