The spec uses USB Type-C cables to deliver power and DisplayPort video simultaneously.
Read the whole story
Read the whole story
VR rivals come together to develop a single-cable spec for VR headsets
- Thread starter JournalBot
- Start date
It's entirely intentional now to drive forced obsolescence.
Standards like VGA, RCA A/V, TOSLink, S/PDIF, 1/4" and 1/8" audio, MIDI, S-Video, and others have all stood the test of time and generally work. They never received "revision".
Aside from analog audio, every single one of those things has had revision and eventually been obsoleted because newer, better technology has come along.
VGA has bandwidth and resolution limits. It also huge problems with switchers or extension cables and can look horrible on flat panel displays under less than perfect conditions. That's why I went 100% digital for my video connections about 10 years ago.
RCA isn't an A/V standard; it's a connector type. The composite baseband video carried on RCA cables actually has different standards all over the world, and a TV made for US markets won't work with a VCR made for the UK. Also, it doesn't have the bandwidth for anything other than standard definition video, and quite frankly, composite video has always been a horrible mess.
TOSLink was never necessary; it was created by Sony to help sell Minidisc players and never should have been brought to market. SP-DIF was fine for the task, and both interfaces have several problems, both technical and legal, that make them less than ideal for general purpose use.
MIDI is dying; by the time you reach 32 notes of polyphony, the latency on a single MIDI connection gets noticeable at times. The physical layer is being replaced by USB and various network implementations.
S-Video is already dead; YPbPr component video superceded it, and you can't buy a new, mainstream television today with an S-Video port.
Even analog audio standards have been revised several times, with different connector types, reference voltages, and the introduction of balanced audio as a way to reduce noise. With all of the various audio standards and interconnection problems, it's probably the most confusing and difficult thing to work with, far more so than video, because it's so susceptible to noise and level mismatch issues.
Upvote
10
(10
/
0)
Should've jumped in a month or so ago when I did. I ended up grabbing a Lenovo Explorer for $200, and it works very well with my GTX 1050 I installed on a 3-year old computer. The WMR standard is really helping drive down the price to a more palatable & reasonable cost. It'll get more affordable as what is considered "high end" today becomes the norm tomorrow, and more manufacturers make WMR-standard devices.
Upvote
1
(1
/
0)
Yeah, having cables attached to you when you are effectively blind *and* trying to walk around a room doesn't sound like a super safe situation. VR headsets almost have to be wireless, unless you are sitting on the couch or standing in place while you use them.
Upvote
0
(2
/
-2)
There is already a wireless adapter for the Rift/Vive by TPCast. It isn't perfect but it's decent.
Upvote
0
(0
/
0)
You speak like someone who never tried VR. Compared to those implementations, it's on a whole new level. Plus, the price of a good WMR headset is about $200, which is about 1/3rd the price of an OR at launch, and the quality is near-parity with the currently-$400 OR. It still requires a fairly beefy PC to run well, but as we've seen even a 1050 can run it passably.
As for your precedents, Virtual Boy failed because the monochrome display was like taking Satan's dick in your eye socket, 3DS' 3D "failed" because it added nothing to the gameplay, 3D TVs never lowered in price enough for most people to consider them (plus there was little 3D content for them), and 3D movies still haven't found a way to make the 3D integral to the experience. VR, on the other hand, has already managed to get full-body tracking and provides a meaningful way to interact with a virtual world like the Virtual Boy/3DS never could.
Mind you, 2 years ago consumer VR was just barely there, nothing more than super-pricey goggles for enthusiasts. Today, between OR/Vive/WMR and a laundry list of smartphone-powered headsets, we're seeing massive technical improvements and far lower prices as well as more immersive titles and diverse genres. At the rate they're going, VR could become a dominant force within a decade.
Upvote
3
(5
/
-2)
No, Thunderbolt 3 is still limited to the same 4-lanes of high-speed interconnect as USB-C. That's only enough bandwidth to handle the DP Alternate Mode at 32Gbps.
The original USB-C standard says that you can SPLIT the DP Alternate Mode into 2 channels and use 2 channels of USB 3.1 10 Gbps. OR you can use all four DP Alternate Mode, and be forced to use USB2.0 for all peripherals (there are 4-dedicated pins for USB2)
It's a new standard released in USB-C version 1.3, which allows captive connectors to re-negotiate the functionality of the four USB2 pins AFTER the cable handshake. That allows you to add TWO MORE high-speed channels, increasing your aggregate bandwidth to 32Gbps DP plus 10Gbps for USB.
This is just the first use of the new official standard by anyone. They still need to add the USB-C version 1.3 connectors to video cards and laptops to make use of this, so this is at least another five years from becoming reality. But it's still an official standard cable connector, it's just a corner case of that standard.
See here for the new version 1.3 cable spec released last year:
https://cambrionix.com/wp-content/uploa ... se-1.3.pdf
Upvote
5
(7
/
-2)
Hey, remember the Apple Newton?
How about the Psion range?
Yeah I totally completely sure this time hand-held computers are the future! For sureeee!
:-D
Upvote
4
(5
/
-1)
https://en.wikipedia.org/wiki/DisplayPo ... ompression
Display Stream Compression breaks each frame into multiple slices which are independently compressed. It adds for less than 1 frame of latency. Framed a different way, DSC can deliver 2160p120 compressed with fewer ms of latency than 2160p60 uncompressed. Oddly I can't seem to find any actual publicly available numbers for this.
It would be lovely to live in a world where VR latency is so low that DSC would be a substantial contributor towards is, but alas, we do not live in that world mid 2018.
Upvote
5
(5
/
0)
A new protocol over the existing Type-C port.
Type-C already handles multiple protocols: Thunderbolt, USB (1, 2, 3, 3.1) , DisplayPort, PCIe, whatever, and 100W of power, etc.
Upvote
4
(5
/
-1)
Well, there were other problems with 3D for gaming and home entertainment.
First, the net output to each eye was <50% less than otherwise, since each eye is only getting half the light output per frame (plus overhead). Second for any pixel-fill limited game, doing two views meant each got just half the processing, which led to a significant loss in visual fidelity. 2x 640x720 just didn't look as good as 1x 1280x720.
Having available glasses and getting everyone to wear them at once was a pain.
Good VR headsets with two separate screens helps a lot, but there are still issues with.
Dominant? I think VR is a fun use case, but it's not appropriate to lots of content and environments. I wouldn't want to ride a bus or subway not being able to see or hear my environment! Dollar-for-dollar, TVs provide a much better and less fatiguing viewing experience.
And human stereoscopic vision only really works out to 20-30 feet, so we don't get much "3D" in lots of scenes, and trying to compress the stereoscopic field means making your virtual eyes farther apart, and makes everything appear proportionately closer to you. Which is why VFX often look LESS realistic in 3D. If the content has a space battle happening in stereo, your eyes will think everything is 10-20 feet away, and thus those massive spacecraft are model sized. Applied to sports, this can look like GI Joe figures playing in lunar gravity.
And prerendered content really only has 2 DOF - left/right and up/down. A full 6 DOF requires a CGI rendered environment, or CGI rendering of video and textures. Doing a photoreal 6 DOF will require several more turns of Moore's Law. And accurate proprioception with movement is an aspiration in the home, not something that actually exists.
Headsets suck for multitasking. No easy way to look down at your phone or over at your clock or notice who's walking towards your desk.
Personally, I'd guess we are a decade away from a lean-back entertainment experience via VR from even matching the fidelity and low-fatigue of a <$10K home theater experience.
Gaming or telepresence via VR are much better suited for the technology, and I'd expect that the first experiences that a million people are willing to sustain for a minimum of an hour will come from that. Today, it's extremely hard to find many people outside of the industry who have enjoyed more than eight total lifetime hours of VR headset experiences.
...all my personal opinions, but after filing a half-dozen patents about VR.
Upvote
6
(6
/
0)
As far as I can tell, everything this does is possible with thunderbolt 3. I'd rather they just all agree to use the existing thunderbolt standard rather than create a new one to do the same thing.
Upvote
-1
(1
/
-2)
I pretty much agree with the gist of your commentary, but the snippet above stuck in my mind.
I don't think we'll ever be at that "lean-back entertainment experience" with VR as we envision it today. It's one to have a "canned" VR experience (non-interative) but once you start expecting an interactive experience, it's going to be fairly high-fatigue...maybe when we get to the point of electrode implanted in our skulls...
Upvote
0
(0
/
0)
If Google Cardboard has not make VR popular by now... eh, it will only be used because it will end as just as another function on Cellphones at this rate...
In fact considering using a cellphone is the cheapest option and doesn't need people to buy anything but get some cardboard, is amazing Google Cardboard is not more popular.
Upvote
0
(0
/
0)
While Intel has promised that they will eventually open up the Thunderbolt spec under a royalty free license, they haven't actually done that yet. That means their competitors, like AMD and nVidia, can't just agree to use it. They might be able to start new designs some time in the future, when Intel finally lives up to their promise, and deliver a new product implementing it several years later, but USB-C 1.3, which was modified explicitly for this use case, was released nearly a year ago.
They could have working silicon of VirtuaLink on their next generation product launch.
Upvote
1
(1
/
0)
Thunderbolt is in no way straight PCIe lanes.
A 2-port Titan Ridge Thunderbolt 3 controller has:
• 2x DisplayPort 1.4 sinks (8 lanes, up to HBR3)
• PCIe 3.0 x4 connection to an integrated PCIe switch
• integrated Intel USB 3.1 xHCI (host controller)
• side video port providing 1x DisplayPort 1.4 source (4 lanes, up to HBR3) or 1x HDMI 2.0a source via integrated level shifter/protocol converter
• HDCP 2.2 pass through
• integrated protocol adapters which can convert both DP and PCIe packets to the Thunderbolt meta-protocol for transport via Thunderbolt signaling
• integrated Thunderbolt crossbar switch
• 2x USB 3.1 Type-C ports capable of Gen 2 10 Gbps signaling with support for USB Power Delivery 3.0 (15 W minimum), DisplayPort Alternate Mode, and Thunderbolt Alternate Mode
When a Thunderbolt 3 port is operating in Thunderbolt Alternate Mode, it is creating a 40 Gbit/s full-duplex serial link using the 4 differential signaling pairs of a Type-C connector / cable normally used for SuperSpeed USB. It can use that insane bandwidth to transfer both PCIe and DisplayPort packets. It is not a PCIe link, it is not a DisplayPort link, it is not a USB bus, it is Thunderbolt.
If you're willing to shell out $9.65 to put a JHL7440 2-port Thunderbolt 3 controller in a VR headset, you would be able to use the side video port to output either DisplayPort 1.4 or HDMI 2.0, and you could connect to the second Thunderbolt port internally in USB 3.1 Gen 2 signaling mode. This is already more flexible than VirtualLink, suitable hosts and cables already exist, and we haven't even touched the PCIe lanes yet.
If you need more display bandwidth, you could connect to the second Thunderbolt Port in DisplayPort Alternate Mode. This is even more flexible because Thunderbolt 3 controllers support 4-lane DisplayPort main links (with or without USB 2.0), or a 2-lane DisplayPort main link plus USB 3.1 Gen 2.
Of course you still have 4x PCIe 3.0 lanes to play with, which are considerably more versatile and lower latency than a USB connection. If all you need is more USB bandwidth, you can just add another USB host controller or two, seeing as how they all connect to the host via PCIe anyway.
Depending on the host and the cable being used, power delivery can go all the way up to 100 W.
And finally, the JHL7440 allegedly allows for compatibility on the upstream port with standard (non-Thunderbolt) USB Type-C hosts.
Bottom line is a single Thunderbolt link allows for the transport of up to 40 Gbit/s of display data, and anything short of that can be used for up to 31 Gbit/s of PCIe data, which can include up to 10 Gbit/s from an integrated USB 3.1 host controller. VirtualLink tops out at 25.92 Gbit/s of display data, doesn't support HDMI (TMDS) signaling, and only offers up to 10 Gbit/s of USB data with no built in support for USB 2.0.
edit: fixed minor errata
Upvote
5
(6
/
-1)
Nice to see them using DisplayPort instead of HDMI.
For software interoperability, there is OpenXR effort (strange you didn't mention it): https://www.khronos.org/openxr
Valve, MS and Oculus are participants.
For software interoperability, there is OpenXR effort (strange you didn't mention it): https://www.khronos.org/openxr
Valve, MS and Oculus are participants.
Upvote
1
(1
/
0)
When OEMs include Thunderbolt controllers in PCs or on motherboards, they're required to feed at least one DP stream per port to the controller. Macs (and I'd imagine PCs with embedded dGPUs) tend to use the dGPU for this whenever it's an option. Obviously PC motherboards are a dicier proposition, but they can either connect to the iGPU (assuming that there is one) or provide a DP-in port somewhere on the board. All Thunderbolt add-in cards have DP inputs. You just run a DP cable from your GPU to the DP-in on the Thunderbolt card.
Upvote
0
(0
/
0)
So I work in the mining industry and we already have externals working in the VR space to allow for a variety of issues. A lot of it will probably fail to excite simply because current real life versions are acceptable/better. But with centralisation of engineering teams etc, this allows them to examine something physical "remotely", which is a big sell for a corporate.
10 years is likely too short a time period, and dominant will always be questionable. But certainly increased market share is going to happen, which of course will make it cheaper and more accessible, which will further increase the market etc etc etc...
I agree though that I don't believe VR TV/Movies will ever truly be a thing. TV/Movies are by their nature "passive" viewing. VR is really an Active solution. manipulating the space you are looking at etc.
Well except for pr0n...... ... . .. .....
Upvote
0
(0
/
0)
repoman,
Clearly you know all the technical details here as well as anyone
Aside from any technical reasons, I don't think AMD (Intel Rival), NVIDIA (Intel Rival) and Microsoft (no shipping TB3 systems) would sign on to TB3 as a standard for VR. Especially because as of right now, the only place you can source a TB3 controller is from Intel, as much as it's an "open standard."
Also, making the HMDs rely on TB3 to function properly means non-TB3 systems can't use it at all. Normally with Titan Ridge you could gracefully fall back to USB + DPAM, but if that were possible then we wouldn't need this new standard in the first place, and hence wouldn't need TB3 either. I could envision a pretty inexpensive VirtualLink "breakout box" that's just a Type-C demux that outputs DP, USB-A, and has a power brick, whereas doing that with TB3 could be much challenging, but I'm not sure.
I think it's more likely that the video card manufacturers move towards Type-C than motherboard (or video card) manufacturers adopt TB3 en-masse. VESA actually certified HBR3 over Type-C before they created the DP8K certification program earlier this year, and we're already seeing monitors ship with Type-C.
I'm not trying to seem down on TB3, I've been an eGPU person for a while (thank you b-plus) and I love TB3.
Upvote
1
(1
/
0)
Upvote
-3
(1
/
-4)
I think you're exactly right in that the plan isn't necessarily for this alternate mode to see widespread adoption on PCs. Microsoft, NVIDIA, and perhaps a few others can include this port on their devices, but for everyone else, there will be a breakout box / dongle. As you said, it's straightforward to make a device with an upstream facing USB port, DisplayPort input port, and a power brick. Then you can use off-the-shelf signal muxes and USB Type-C / PD controllers to support the VirtualLink alternate mode on a downstream facing Type-C port. This also somewhat solves the problem of the fixed length tethered cable on the headset, because you can locate the box wherever you like and run commodity USB and DP cables of the required length back to the PC. Thunderbolt needs to be tightly integrated from the outset, and for the majority of systems, cannot be added after the fact. Plus it may be a non-starter for various political reasons, as you mentioned.
repoman,
Clearly you know all the technical details here as well as anyone
Aside from any technical reasons, I don't think AMD (Intel Rival), NVIDIA (Intel Rival) and Microsoft (no shipping TB3 systems) would sign on to TB3 as a standard for VR. Especially because as of right now, the only place you can source a TB3 controller is from Intel, as much as it's an "open standard."
Also, making the HMDs rely on TB3 to function properly means non-TB3 systems can't use it at all. Normally with Titan Ridge you could gracefully fall back to USB + DPAM, but if that were possible then we wouldn't need this new standard in the first place, and hence wouldn't need TB3 either. I could envision a pretty inexpensive VirtualLink "breakout box" that's just a Type-C demux that outputs DP, USB-A, and has a power brick, whereas doing that with TB3 could be much challenging, but I'm not sure.
I think it's more likely that the video card manufacturers move towards Type-C than motherboard (or video card) manufacturers adopt TB3 en-masse. VESA actually certified HBR3 over Type-C before they created the DP8K certification program earlier this year, and we're already seeing monitors ship with Type-C.
I'm not trying to seem down on TB3, I've been an eGPU person for a while (thank you b-plus) and I love TB3.
I believe what the JHL7440 falls back to for compatibility is just DisplayPort Alternate Mode. But apparently that can't provide enough display bandwidth for VR applications when operating in the 2-lane main link + SuperSpeed USB mode, hence the need for VirtualLink. The loss of the USB 2.0 bus makes VirtualLink feel like a rather inelegant solution though. And I still find this to be more of a way to side-step an industry problem, rather than the best way to address the issue from a consumer standpoint.
Upvote
0
(0
/
0)
Ah, but for all those whose love for thumbscrews springs eternal, fear not, the USB-IF has not forsaken you. Behold, from the Universal Serial Bus Type-C Locking Connector Specification:
Not even kidding. It's actually a thing.
Upvote
10
(10
/
0)
HTC didn't participate in this consortium?
That seems like a problem. They are the pilot manufacturer for the SteamVR stuff. Still a significant player in the market with the most advanced product.
Hmmm... Hopefully they still try to use it.
That seems like a problem. They are the pilot manufacturer for the SteamVR stuff. Still a significant player in the market with the most advanced product.
Hmmm... Hopefully they still try to use it.
Upvote
0
(0
/
0)
HTC was always only manufacturer for Valve. This was the reason why there is no substantional
improvement to lenses, FOV, etc, in Vive Pro compared to original Vive.
Upvote
0
(0
/
0)
Yeah, you tell them!
Remember that time many insisted in building a flying contraption? All those imbeciles insisting humans could fly only to look ridiculous and crash down every time... The future they said.
Oh, and lets not forget those computers. I mean, outside of a corporation or government, who's going to want one of those? They will never fit in a home!
How wrong were all those people to keep developing their inventions, even more so all the inventors that came after and built upon earlier work.
The fools, all of them. Amirite bro? Bro?
/S
Upvote
0
(0
/
0)
This should have happened sooner. And the fact that its happening now suggests unpleasant things about the practicality of wireless solutions.
Wired connections give you a much more reliable method of transmitting data in real time, which is important for high frame-rate, high-resolution, interactive media. The bandwidth you need is huge for uncompressed video (for example, the 2x 2560x1440 screens at 120fps scenario is more than 21gbps for 8-bit video data). Once you add compression for wireless transmission you add latency. Without significant advances in consumer-grade wireless transmission we won't be seeing top level VR wirelessly connecting to its base-station any time soon in the consumer space.
Wouldn't foveated rendering dramatically reduce the bandwidth required?
It would. But even at a 10x reduction you are still looking at an uninterruptable 2gbps for 8-bit video. Although much more manageable it's still not a solved problem. And it introduces other problems to do with eye-tracking, which also require a lot of work.
Upvote
1
(1
/
0)
Yes, kind of like Pixvana's Diamond Plane Encoding, which doesn't render what is behind the user in a 360 video, allowing higher resolution at similar bandwidths. However, until eyes are tracked the only way to know where the user's point of focus is to assume that the eyes are straightforward and to use the headsets orientation, but people do move their eyes around without moving their headset. Otherwise, this is great.
Upvote
-1
(0
/
-1)
"The new headset spec combines four lanes of HBR3 ("high bitrate 3") DisplayPort video (for a total of 32.4 gigabits per second of video data)"
Am I missing something here? 4 lanes would mean four times 32.4 if they used all four lanes. So the total bandwidth would be 129.6 gbits.
Am I missing something here? 4 lanes would mean four times 32.4 if they used all four lanes. So the total bandwidth would be 129.6 gbits.
Upvote
0
(0
/
0)
You are going the wrong way. It's 8.1 gbits per lane so the 32.4 is the total for all 4 already. That is actually the link bandwidth and after encoding it's only 25.92 Gbit/s for actual video data.
https://en.wikipedia.org/wiki/DisplayPo ... ifications
Upvote
2
(2
/
0)
Not that anyone else cares, but I see that PlayStation isn't playing nice with others again. I figure they don't need USB-C when they don't really need to interoperate and already utilize HDMI, but still.. find it odd that a major VR platform is left out here. And.. does this news make it more or less likely that VR comes to Xbox.
Upvote
0
(0
/
0)
That's very underwhelming. Just do HDMI 2.1 instead, it's already higher bandwidth than that.
Upvote
0
(0
/
0)
The DisplayPort 1.3 spec that introduced HBR3 came out in September 2014. DP 1.3/1.4 sources already exist.
The HDMI 2.1 spec was only finalized in November 2017, and it's based on an entirely new architecture. So maybe next year when the first silicon that supports it begins to trickle out, it will finally offer more bandwidth than DisplayPort. It's also encumbered by licensing, which may preclude its use in derivative standards such as VirtualLink.
Upvote
0
(0
/
0)
There'll need to also be some panning/fade/etc protocol too (basic elements of a video codec) in the brand new display protocol. Since having varying framerates for different parts of vision was tested to to be bad.
They've come up with workarounds like that for that. And it can be done hand-in-hand with reprojection (Asynchronous Space Warp / Time Warp) which does 90fps scrolling with 45fps 3D object rate.
So we spin that into a cable optimization (and also do bigger frame rrate amplification ratios, e.g. 60fps->240fps amplification instead of 45fps-90fps)
The foveated rendering could simply be a more spatially flexible layering of that sort of thing, with even fewer artifacts. All this is done at the GPU level, but this could be transferred responsibility into the video cable.
Long-term, a "simple GPU" (phone grade) may be needed in the headset, as co-GPU, to do some of the reprojection tasks.
So a lower framerate can be sent from the main PC, to result in a higher framerate within the headset. (e.g. 8K 240Hz with with the bandwidth of 4K 60Hz). All done laglessly, of course, thanks to high-rate geometry updates (1000Hz and headtracking).
Geometry updates are low bandwidth, extremely high Hz and lagless, and can be incorporated into creating a high framerate more laglessly from a low framerate.
Basically, think of Oculus style timewarp -- but greatly improved (fewer artifacts) -- and you get lower cable bandwidth -- by having the frame rate amplification being done in the headset.
Theoretically, the co-GPU idea is standardizable (e.g. like an "Metal Remote API" where two GPUs intentionally communicate to each other over a cable -- one big desktop GPU and one high-end-smartphone-league GPU -- to co-operate on things like:
-- frame rate amplification responsibilities
-- optimizations for foveated renders
-- input lag reduction
-- develop even more cable bandwidth optimizations
-- preemptive headset-side scrolling during head turn (even before new frame arrives from main GPU)
-- etc.
Obviously, this might be a multiyear tech progress journey, but would work around the cable bandwidth brick wall quite nicely towards Retina headsets.
Upvote
1
(1
/
0)
This should have happened sooner. And the fact that its happening now suggests unpleasant things about the practicality of wireless solutions.
It's still impractical to run 1080p60fps over a wireless VGA/HDMI setup, let alone anything of a higher resolution or refresh rate. Not only that, but it has to be done for a device that will supposedly be battery-powered. Truly wireless VR setups are a long ways off unless you're happy with Google Daydream.
But that's not true at all. TPCast is transmitting 1200p 90fps with less than one frame of latency and has been selling a product for over a year. They have a model for both the Oculus Rift and HTC Vive. I've used it and it's effectively flawless, at least on the image transmission and gameplay side of things(it lacked mic support for a majority of it's existence, though).
In addition, DisplayLink and Intel have a wireless solution for Vive that's been announced and demo'd at CES and other events, and is coming out in the next few months.
Upvote
0
(0
/
0)