I have no affiliation with Engadget (in fact, I've spent over a decade working at competitors to Engadget). But I've found that it makes sense to evaluate news organizations' output on an article-by-article basis. Even the best news organizations can screw things up on occasion, and even excellent reporters can be assigned derivative stories by their editors.
The article-by-article approach is what http://recent.io/ uses in its pipeline, which is working very well in testing so far.
More to the point, this looks like a perfectly decent story and an engaging read. The writer even managed to avoid being snarky about Samsung trying (sigh) to second-guess John Carmack!
"And I had argued for it a long time. I wrote so many of these multi-page emails about how important it was. But once we had it so people could see and see that it really does work, then Samsung went and wrote a proper interface for it."
Wow, I really have to admire his patience. While reading this I was even thinking "They are arguing with John Carmack about what he needs to get the graphics right? Who do they think they are?!" and I am not him. I would never have the patience for this. That must be like my grandma arguing with me about programming.
I too admire his patience, but outside of the games and graphics industries and PC gaming community he is a relative obscurity. The other people involved are from other worlds with different idols.
His largest achievements were years ago, though. Doom was one, Quake another, and Quake 3 probably the last big step. Afterwards, he hasn't distinguished himself so much (I don't think Rage was very impressive, for example), so unless you keep pushing the limits every single time, your reputation can fade quickly.
I don't know if you could the Doom 3 tech impressive - I always found that at the time, it looked like a dirty hack rather than being genuinely impressive. Later games with soft shadows looked much better. And for Rage I'm aware of the megatextures, but that looked horrible for anything that was close to you - again a failed tech.
EDIT: interesting, getting downvoted because I wasn't impressed by Carmack's later productions?
Megatexture was simply a technology before it's time.
The problem is that there just isn't a way to ship the data set right now that you need to support it properly. The hardware is there, the infrastructure is not.
They really needed to ship several hundred gigs of textures to do it justice. Right now you can't realistically deliver that to players.
Now certainly basing a game off a technology that isn't ready for prime time is a mistake, but the technology itself is very impressive and the underlying concept will come back when the internet is fast enough that you can realistically expect someone to download a half terabyte game.
> Now certainly basing a game off a technology that isn't ready for prime time is a mistake, but the technology itself is very impressive and the underlying concept will come back when the internet is fast enough that you can realistically expect someone to download a half terabyte game.
Well I agree with your statement, but I'm not sure the tech is worth it in the end. The current tech is advancing faster than the Megatexture one, and by the time the Megatexture becomes a reality again, we won't need it anymore. Look at all games coming out in 1 to 2 years, they look fantastic while still relying on traditional texture concepts.
On a side note, it did not help to sell the technology that Rage was a notoriously bad game (just like Doom3, but hey, it's not a popular opinion apparently not to like having monsters jumping at you right when you open a door).
What do you mean by the current tech? I bet every streaming world engine programmer has look at megatextures and some of its concepts are in open world games available right now. Similarly the fast hard-edged shadows from Doom 3 were ahead of the curve.
And although I agree that as games Doom 3 and Rage were just OK, Carmack is known for his tech; I don't think he gets involved on the game design side.
He is also by his own admission still learning about programming:
On the software development side, you know there was an interesting thing at E3, one of the interviews I gave, I had mentioned something about how I’ve been learning a whole lot, and I’m a better programmer now than I was a year ago and the interviewer expressed a lot of surprise at that, you know after 20 years and going through all of this that you’d have it all figured out by now, but I actually have been learning quite a bit about software development, both on the personal craftsman level but also paying more attention by what it means on the team dynamics side of things.
This is one of John Carmack's great skills. Over the years he has steered the entire industry towards improving itself. He's probably sent thousands of emails to Intel, Microsoft, AMD, NVIDIA, OpenGL, etc. He's one of the greatest driving forces in computing advancement.
Millions of people upgraded their computers from 1996-2001 almost exclusively to play his games. He's probably responsible for more sold NVIDIA and Intel chips than any single person outside those companies.
> Gear VR's been in development for around 1.5 years now
Wow. That puts it not too long after the Oculus Rift kickstarter, which was in October 2012. I wonder how the Gear project came to start. Were there some graphics enthusiasts with moving power at Samsung that went "this. we need to get on this"?
I truly do not understand this. Why would you want to dilute presence? Why would you want to create an expensive phone accessory when the hardware changes constantly and the software is extremely fragmented? There are plenty of scenarios where I can see usinga Rift, but very very few where I am 'mobile' and want to experience VR. Of those cases I can't see any where a dedicated and occasionally updated set of hardware isn't a better development experience than trying to deal with VR hassles on top of all the standard android ones. Please someone point out what Carmack gets here that eludes me.
It is simple. The target Carmack is aiming for is mobile VR. Tethered experience is a necessary middle step. Oculus will release their version in the future and Gear allows him to develop the tech.
Wireless for starters is just better and required for VR to rock, with the exception of seated experiences (racing, flight sims, mech games, etc) where presence is kept and you can benefit from a PCs far less limited GPU/CPU capabilities.
Also they have a direct line with the phone manufacturer for the display technology and with this strategy of tying in with the mobile world, they can potentially get into a loop of rapidly updating hardware that's subsidized by phone contracts and the like.
The limitations of mobile graphics can and will be solved in hardware and software. It may turn out for instance that something focused on high FPS and resolution can be achieved via low complexity shaders and some novel custom display/GPU hardware.
It's less about the mobile as in using-while-walking-around than the ability to have with you wherever you're at and not tied down to the context of sitting in front of a PC. VR is for everywhere: the couch, the kitchen table, in class, at work, social gathering where you can easily share it, etc. The tether of the current dev kits actually takes away from presence.
While there are hassles of VR with Android that were only solved by making deep changes to Android, the hassles of PC are greater. And this is a phone accessory that is specifically not designed for changing hardware and software: it only works with the Note 4 running a custom optimized version of Android. But they do agree with you that a dedicated unit is what they want to do as well which is why this is not their own consumer unit.
Wireless video can have approximately one millisecond latency today, I'm not sure how bad that would be wrt to things like nausea but it seems pretty short to me.
Article sez Carmack's first big fight was convincing Samsung that a two frame delay was too long. WAY too long. And it was important enough that he wrote an "obscene workaround" to avoid that "pretty short" latency.
I've been casually following Carmack & Occulus for a while; recurring theme is "X ms delay is too long". IIRC (corrections please?), wasn't until latency got down to 20ms that the nausea etc subsided. That doesn't leave much room to both render a high-res complex image and display it by anything more than the shortest transmission time conceivable.
Reconsider: 2 frames at 60Hz was an intolerable 33ms too long, he needed it down to 16ms total. Your optimistic 1ms latency cuts into that maximum refresh time by 7%, time that will (or is already) in high demand by meaningful/viable applications.
You're also assuming a highly customized/optimized scenario for the wireless hop, nigh unto a straight analog signal. Digital processing will extend that, likely adding at least one full frame delay - again, something Carmack severely objected to and fought mightily against.
Even if it's 1ms, that means that you only have 19ms left for all of the processing / rendering, including to the I/O delay of getting motion data from the input.
How much more work can you do on a desktop CPU/GPU in 19ms than on the mobile hardware? If the desktop hardware is 10x faster, wouldn't it still be a win?
Also, consider that you might allow for a small window for when things don't go as planned. So, 15ms to do the computations, for example.
This is all assuming a hard 1ms transmission latency. Usually numbers like that are averaged and this seems like a use-case where 1ms would be an upper bound instead.
That's a really good point. The tech I'm referring to is used in high end (military simulation) gear but it is definitely todays technology and from what I can read it is an upper bound, not an average but that could be marketing, I haven't tested this stuff. But I've been following the 'wearable dual display' setup since a friend of mine built one in 1983 or so (using two very small monitors mounted to the side of a helmet and a bunch of mirrors (2 100%, 1 50%) to put the image in front of the eyes and that's the best latency for wireless that I've seen to date. Given that is a small fraction of the framerate and a latency rather than a throughput limit (so a constant offset rather than a limitation on quality) it imposes a small overhead (at 50 fps that's 20 ms/frame so 5% of the total time budget).
That seems a good trade-off versus being tethered, you could for instance make up for that by letting the rest of the pipeline generate an image that is slightly worse to achieve the same framerate in an untethered situation.
Of course any latency increase will always come at some price the question is whether or not that is an acceptable price for the improvement you get along another dimension all this assuming that's an upper bound.
(And maybe there is higher end gear that I'm not aware of)
Your eyes are only made for a certain frame rate. Any more and your eyes will ignore it, but your brain will need to piece it back together (ie nausea). Any less and it will seem choppy. It's a very very thin line at which it goes from nausea too choppy/laggy.
Not that a few milliseconds of latency really matter, but if you don't care about 2 milliseconds here and another 2 there, they start adding up. It's much better to try your best to have the lowest possible latency.
While it's true that a biological eye doesn't really have a defined "framerate" like video sensors do, the examples you're providing don't really show that. Sensitivity to very short stimuli doesn't prove temporal resolution in the sense that's applied here; what you want to be looking at is flicker fusion instead. As a cheeky example -- still pictures have approximately zero frames per second temporal resolution but would be able to pick up that 2.5ms plane
if their sensitivity was sufficiently high...
Measuring ERGs (electro-retinogrammes) indicates that our photoreceptors can't really resolve luminance fluctuations above, at the very limit, 70-80Hz. Even Drosophila's high speed cascade doesn't really modulate beyond 200-250Hz. I'm really doubtful about those claims.
Indeed, flicker fusion for humans pretty much tops out at 80Hz, but you're underestimating the effects of smooth pursuit eye movement. For example, if you animate a fast-moving object at 80Hz on a 160Hz display, and follow it with your eyes, you will see two copies of the object. There's really no upper limit for this effect.
There's no circumstance in which increasing the frame rate causes nausea. In fact, it tends to decrease latency and reduce nausea. In the limit, reality has an essentially-infinite frame rate and doesn't cause nausea.
When low frame rates seem choppy, that's normally because there's a mismatch between frame rate and display refresh. GSync/freesync fix that to make things smooth even when the frame rate drops. Choppiness can also be due to the impulse response; technologies with a bright and short apparent pulse (CRT, Plasma, some OLED) will show better 'motion' than those that display and hold (LCD, some OLED).
2014 and there are still websites that auto start videos? At the bottom of the page no less? I pretty much ctrl+q out of fear at this point. Autoplaying videos = virus/phishing website in my head.
I'm confused. I know Carmack joined Oculus, but what's this about Samsung? I take it they're working together now? It's still two different pieces of hardware, right? This article is leaving out some crucial context here.
It is called "the screen". Samsumg manufactures(mass produces) the best screens in the world. Their OLED screens are the best quality, simple, lightweight and extremely fast response.
Oculus needs Samsung screens. Samsumg needs oculus talent, so they are partners.
Sure, I'm not saying it didn't make sense, just that the article didn't bother to explain it at all. If you start off by saying someone works for one company, then start going on about his doing stuff for another company, you ought to explain what the two of the have to do with each other.
Yep. Gear VR is like a posh Google Cardboard with a Samsung Note 4 strapped to the front (a, probably, offensive oversimplification, it actually has tech in the headset, but it's just a face-holder for the phone), and the Oculus Developer Kit is the whole thing in one package that plugs in like a monitor to your computer.
Using the device camera together with Microsoft Photosynth technology would be viable. Maybe not in the time required to be confortable for the wearer of tbe VR set.
51 comments
[ 3.3 ms ] story [ 125 ms ] threadThe article-by-article approach is what http://recent.io/ uses in its pipeline, which is working very well in testing so far.
More to the point, this looks like a perfectly decent story and an engaging read. The writer even managed to avoid being snarky about Samsung trying (sigh) to second-guess John Carmack!
Wow, I really have to admire his patience. While reading this I was even thinking "They are arguing with John Carmack about what he needs to get the graphics right? Who do they think they are?!" and I am not him. I would never have the patience for this. That must be like my grandma arguing with me about programming.
He also developed megatextures for Rage. http://en.wikipedia.org/wiki/MegaTexture
EDIT: interesting, getting downvoted because I wasn't impressed by Carmack's later productions?
The problem is that there just isn't a way to ship the data set right now that you need to support it properly. The hardware is there, the infrastructure is not.
They really needed to ship several hundred gigs of textures to do it justice. Right now you can't realistically deliver that to players.
Now certainly basing a game off a technology that isn't ready for prime time is a mistake, but the technology itself is very impressive and the underlying concept will come back when the internet is fast enough that you can realistically expect someone to download a half terabyte game.
Well I agree with your statement, but I'm not sure the tech is worth it in the end. The current tech is advancing faster than the Megatexture one, and by the time the Megatexture becomes a reality again, we won't need it anymore. Look at all games coming out in 1 to 2 years, they look fantastic while still relying on traditional texture concepts.
On a side note, it did not help to sell the technology that Rage was a notoriously bad game (just like Doom3, but hey, it's not a popular opinion apparently not to like having monsters jumping at you right when you open a door).
And although I agree that as games Doom 3 and Rage were just OK, Carmack is known for his tech; I don't think he gets involved on the game design side.
On the software development side, you know there was an interesting thing at E3, one of the interviews I gave, I had mentioned something about how I’ve been learning a whole lot, and I’m a better programmer now than I was a year ago and the interviewer expressed a lot of surprise at that, you know after 20 years and going through all of this that you’d have it all figured out by now, but I actually have been learning quite a bit about software development, both on the personal craftsman level but also paying more attention by what it means on the team dynamics side of things.
https://blogs.uw.edu/ajko/2012/08/22/john-carmack-discusses-...
Millions of people upgraded their computers from 1996-2001 almost exclusively to play his games. He's probably responsible for more sold NVIDIA and Intel chips than any single person outside those companies.
Wow. That puts it not too long after the Oculus Rift kickstarter, which was in October 2012. I wonder how the Gear project came to start. Were there some graphics enthusiasts with moving power at Samsung that went "this. we need to get on this"?
http://blogs.valvesoftware.com/abrash/why-you-wont-see-hard-...
Also they have a direct line with the phone manufacturer for the display technology and with this strategy of tying in with the mobile world, they can potentially get into a loop of rapidly updating hardware that's subsidized by phone contracts and the like.
The limitations of mobile graphics can and will be solved in hardware and software. It may turn out for instance that something focused on high FPS and resolution can be achieved via low complexity shaders and some novel custom display/GPU hardware.
While there are hassles of VR with Android that were only solved by making deep changes to Android, the hassles of PC are greater. And this is a phone accessory that is specifically not designed for changing hardware and software: it only works with the Note 4 running a custom optimized version of Android. But they do agree with you that a dedicated unit is what they want to do as well which is why this is not their own consumer unit.
I've been casually following Carmack & Occulus for a while; recurring theme is "X ms delay is too long". IIRC (corrections please?), wasn't until latency got down to 20ms that the nausea etc subsided. That doesn't leave much room to both render a high-res complex image and display it by anything more than the shortest transmission time conceivable.
You're also assuming a highly customized/optimized scenario for the wireless hop, nigh unto a straight analog signal. Digital processing will extend that, likely adding at least one full frame delay - again, something Carmack severely objected to and fought mightily against.
[1] http://www.roadtovr.com/samsung-gear-vr-official-specificati...
> so 1ms wouldn't be a big deal
Even if it's 1ms, that means that you only have 19ms left for all of the processing / rendering, including to the I/O delay of getting motion data from the input.
Also, consider that you might allow for a small window for when things don't go as planned. So, 15ms to do the computations, for example.
That seems a good trade-off versus being tethered, you could for instance make up for that by letting the rest of the pipeline generate an image that is slightly worse to achieve the same framerate in an untethered situation.
Of course any latency increase will always come at some price the question is whether or not that is an acceptable price for the improvement you get along another dimension all this assuming that's an upper bound.
(And maybe there is higher end gear that I'm not aware of)
Not that a few milliseconds of latency really matter, but if you don't care about 2 milliseconds here and another 2 there, they start adding up. It's much better to try your best to have the lowest possible latency.
Noticing a flash of light can go into the 1000 fps territory.
http://www.100fps.com/how_many_frames_can_humans_see.htm
Measuring ERGs (electro-retinogrammes) indicates that our photoreceptors can't really resolve luminance fluctuations above, at the very limit, 70-80Hz. Even Drosophila's high speed cascade doesn't really modulate beyond 200-250Hz. I'm really doubtful about those claims.
There's no circumstance in which increasing the frame rate causes nausea. In fact, it tends to decrease latency and reduce nausea. In the limit, reality has an essentially-infinite frame rate and doesn't cause nausea.
When low frame rates seem choppy, that's normally because there's a mismatch between frame rate and display refresh. GSync/freesync fix that to make things smooth even when the frame rate drops. Choppiness can also be due to the impulse response; technologies with a bright and short apparent pulse (CRT, Plasma, some OLED) will show better 'motion' than those that display and hold (LCD, some OLED).
It is called "the screen". Samsumg manufactures(mass produces) the best screens in the world. Their OLED screens are the best quality, simple, lightweight and extremely fast response.
Oculus needs Samsung screens. Samsumg needs oculus talent, so they are partners.