Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Monday July 15 2019, @06:17PM   Printer-friendly
from the predicted-sales-of-6-or-7-mainframes-by-IBM dept.

Facebook will never break through with Oculus, says one of the VR company's co-founders

Five years after its $2 billion purchase of Oculus, Facebook is still pushing forward in its efforts to bring virtual reality to a mainstream audience. But one of the company's six co-founders now doubts Oculus will ever break through.

Jack McCauley told CNBC he doesn't think there's a real market for VR gaming. With Facebook positioning its Oculus devices primarily as gaming machines, McCauley doesn't believe there's much of a market for the device. "If we were gonna sell, we would've sold," McCauley said in a phone interview on Wednesday.

[...] The $199 Oculus Go has sold a little more than 2 million units since its release in May 2018, according to estimates provided by market research firm SuperData, a Nielsen company. The Oculus Quest, which was released this May, has sold nearly 1.1 million units while the Oculus Rift has sold 547,000 units since the start of 2018, according to SuperData.

[...] Since leaving in November 2015, McCauley has enjoyed a semi-retired life. He's an innovator in residence at Berkeley's Jacobs Institute of Design Innovation and he continues to build all sorts of devices, such as a gun capable of shooting down drones, at his own research and development facility.

The cheaper, standalone headsets are selling more units. Add foveated rendering and other enhancements at the lower price points (rather than $1,599 like the Vive Pro Eye), and the experience could become much better.

Related: Oculus Rift: Dead in the Water?
HTC: Death of VR Greatly Exaggerated
As Sales Slide, Virtual Reality Fans Look to a Bright, Untethered Future
Virtual Reality Feels Like a Dream Gathering Dust
VR Gets Reality Check with Significant Decline in Investment
Creepy Messages Will be Found in Facebook's Oculus Touch VR Controllers


Original Submission

 
This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
  • (Score: 0) by Anonymous Coward on Monday July 15 2019, @07:11PM (7 children)

    by Anonymous Coward on Monday July 15 2019, @07:11PM (#867290)

    Better hardware! No wires, easier set up, better ergonomics.

    It needs the equivalent of 1080p at least which is about 12k. For reference, the Pimax 8k costs $900 and is about as good as 720p.

    Cheaper

    A new console costs $500. A new TV costs $200-300. You'd need computer+headset+GPU costing about as much.

    Overall, it's multiple GPU and monitor node reductions away for over 5 years worth and then another 5 years for the prices to come down enough for actual market penetration. i.e. The market doesn't exist.

  • (Score: 3, Insightful) by takyon on Monday July 15 2019, @09:11PM (6 children)

    by takyon (881) <reversethis-{gro ... s} {ta} {noykat}> on Monday July 15 2019, @09:11PM (#867322) Journal

    Foveated rendering is the way forward. If you only have to render 5-10% of the screen in full quality at any given moment, maybe you can get away with having 1/10th of the GPU performance. It's a shortcut that will eventually allow 16K resolution in a standalone device using a smartphone SoC.

    --
    [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
    • (Score: 3, Interesting) by RamiK on Tuesday July 16 2019, @08:17AM (5 children)

      by RamiK (1813) on Tuesday July 16 2019, @08:17AM (#867463)

      Foveated rendering is the way forward.

      Assuming it works, only way actually: Keeping with OP's timeline, due to foveated rendering, in 5 years when screen nodes catch up you'll have an overnight revolution segmenting gaming and desktop graphics to silicon 10x cheaper and less power consuming. So, since embedded graphics will suddenly get good enough, all the new games engines and titles will use it.

      Of course, that's assuming it works and isn't an IP minefield. The right patent at the wrong hands can easily hold this back by half a decade.

      --
      compiling...
      • (Score: 2) by takyon on Tuesday July 16 2019, @09:57AM (4 children)

        by takyon (881) <reversethis-{gro ... s} {ta} {noykat}> on Tuesday July 16 2019, @09:57AM (#867479) Journal

        Oculus, HTC, and others are working on it. It seems like each company's headsets are advancing in features despite potential patent issues.

        Theoretically, monitors or laptops could support foveated rendering if they are capable of tracking the eyes as well as the distance of the head/eyes to the screen (so they would need to have sensitive front-facing camera(s) and a depth sensor). For fun, stare at the top right corner (or some other area) of your screen, from any distance. You can't read much text or notice fine details from other parts. The image on this article [theverge.com] also demonstrates it well. Even taking the 11.6" screen I'm using now and moving it several feet away from my face, portions outside of the view of the fovea centralis [wikipedia.org] are not well detailed and could be rendered at a lower resolution.

        5-10% may be a conservative estimate. If it's more like 1% of the field of view of a VR headset or screen being perceived by the fovea centralis, you could render 1% at 16K, 1% surrounding that (circle) at 8K, 1% at 4K, 2% at 1080p, 5% at 720p, and then 90% at 240p. Over a 16K display I assume for "ultimate" VR [soylentnews.org], that comes out to:

        Full 16K resolution = 15360*8640 = 132,710,400 pixels (132.7 megapixels)

        Foveated scenario = (0.01*15360*8640) + (0.01*7680*4320) + (0.01*3840*2160) + (0.02*1920*1080) + (0.05*1280*720) + (0.9*426*240) = 1,921,392 pixels (1.9 megapixels) = ~1.45% of 16K resolution, or ~92.7% of 1080p resolution

        Obviously, I have pulled numbers out of thin air here, and just because pixel count is that low doesn't necessarily mean that you can get away with having 1/70th of the GPU performance, but the example gives us an idea of what we're working with. Pushing a much lower pixel count could have implications for the memory bandwidth and internals of the device, or the data rate needed to transmit content wirelessly from a big GPU. It also allows you to get up to 240 Hz (or higher?) without breaking a sweat in terms of total pixels per second.

        LG and Google recently showed off [roadtovr.com] an 18.4 megapixel display, with plans to get up to 86.4 megapixels, not far off from an "even" 16K 16:9 resolution (132.7 megapixels).

        As far as GPUs are concerned, I think there's reasons to be optimistic. Node shrinks help GPUs much more than CPUs since they are embarrassingly parallel. A new technology [soylentnews.org] could come onto the scene and "extend Moore's law by decades". The end result (over a decade from now) being that various GPUs and even smartphone SoCs *could* render 16K graphics without foveated rendering, but the foveated rendering would allow it to be done with ultra low power consumption, which is good since strapping a big battery to your face is a risk.

        --
        [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
        • (Score: 2) by FatPhil on Tuesday July 16 2019, @03:52PM (3 children)

          by FatPhil (863) <pc-soylentNO@SPAMasdf.fi> on Tuesday July 16 2019, @03:52PM (#867590) Homepage
          Wouldn't you need to render in hig rez the distance the eye might move in the time it takes you to render the next frame?
          You don't want to be repeatedly saccading onto fuzz.
          --
          Great minds discuss ideas; average minds discuss events; small minds discuss people; the smallest discuss themselves
          • (Score: 2) by takyon on Tuesday July 16 2019, @06:22PM (2 children)

            by takyon (881) <reversethis-{gro ... s} {ta} {noykat}> on Tuesday July 16 2019, @06:22PM (#867644) Journal

            http://vrguy.blogspot.com/2016/04/understanding-foveated-rendering.html [blogspot.com]

            If you have a peak saccade velocity of 900°/s or 0.9°/ms, the focus can't move very far in 1 frame.

            https://pdfs.semanticscholar.org/0eb5/b4e1c437bef4d8942fb327be190a74537124.pdf [semanticscholar.org]

            Since received eye tracking results can only be used for the frame after the results have been processed, the refresh rate of the camera should be higher than the framerate of the application to limit latency. Another reason the refresh rate has to be high is human saccade, which is the most common, rapid movement of the eyes to focus on a new location, which the camera will have to be able to keep up with and detect. Consumer cameras for computers, such as webcams, do not offer this kind of performance yet.

            [...] The performance requirements increase with the decision to utilise headmounted displays, however, since in order to prevent cybersickness, a form of motion sickness which is caused by a conflict between perception and the user’s expectation in an immersive environment, the frame rate needs to be higher than on general monitors and 95 frames per second is the established target as that should eliminate noticeable flickering of the screen [Val14].

            [...] The smallest layer also refreshes at 120Hz, while the bigger layers alternatingly refresh at 60Hz. This is to be able to better account for saccades and to make the updating of the foveal region unnoticeable. A big factor in this is the system latency, the time between the capture of the eye gaze position and the rendered image being displayed on the screen. If that system latency is too big, the system lags behind and the user will be able to see the update of the foveal region of the image. All the systems in the method work asynchronously, making the exact latency hard to predict. Guenter et al. carefully chose their system to minimise that latency and made an analysis of the best and worst case latencies, which are 23ms and 40ms, respectively.

            [...] The user study between three rendering algorithms (no foveated rendering, the method from Guenter et al. augmented with the proposed TAA and their own proposed method) showed that they can render more coarsely up to 30◦ closer to the centre of interest than Guenter et al. without introducing gaze-dependent aliasing or blur. Through metrics, they also showed that their final result is highly similar to their perceptual target image.

            That study used FOVE [wikipedia.org], which tracks eyes at 120 Hz but has a refresh rate of only 70 Hz.

            Today's headsets are in the 90-120 Hz range, and the target is 240 Hz [reddit.com] (or higher? [blurbusters.com]).

            240 Hz is a little over 4ms per frame. If eye tracking is at 240 Hz or greater and that peak saccade velocity mentioned is accurate, then the focus can't move much more than 3°. If you render a circular area at 16K, and outer rings at 8K, 4K, etc. decreasing quality, your eye won't reach any ultra-low resolution area fast enough before the next eye positions are calculated and the spot moves.

            That article also covers some more advanced topics like temporal artifacts, etc. But the bottom line is that this is doable. All of the numbers are subject to negotiation/testing. If you want to be conservative, you render a bigger slice at higher qualities, and might only reduce pixel count by 90%. If you can track the eyes faster, maybe you can lower pixel count by 98%. So even though high frame rates are a bigger burden for the hardware, you could get to decrease the total pixels rendered as tracking/rendering Hz increases.

            It seems likely that you want a standalone headset in order to reduce latency as much as possible. Standalone headsets have worse GPU performance but better latency potential. Good thing that foveated rendering greatly reduces the GPU performance that you need. Long live the smartphone SoC.

            --
            [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]