Oculus "Rift" : An open-source HMD for Kickstarter

Post Reply
MaterialDefender
Binocular Vision CONFIRMED!
Posts: 262
Joined: Wed Aug 29, 2012 12:36 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by MaterialDefender »

What I can say out of experience with a DIY-Rift is that the FOV is absolutely big enough to give you a very compelling being-there-impression. One might compare it to looking through ski goggles: there is a visible border in the peripheral area of the image you see, but it's not distracting at all.
Endothermic
Binocular Vision CONFIRMED!
Posts: 284
Joined: Tue Jun 26, 2012 2:50 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Endothermic »

LikeMike wrote:Hello there, newbie here. I am really excited about the Rift - but I still have a couple of questions, perhaps some of the people here can answer:

1. As I understand it and see it in the videos, the screen is basically divided in two halfs showing basically the same thing - doesn`t that mean, that we have an about half as wide field of view than we would have on a monitor? And basically a larger vertical field of view than a horizontal? To me that seems kinda odd...
It's not just a matter of the screen being cut in half so the FOV is cut in half, you need to consider the horizontal FOV in relation the the vertical FOV, and it's that larger vertical FOV you mentioned which means the horizontal FOV isn't actually cut in half.

FOV in a game i just a number that is set, then depending on the shape of the display you'll see so much game world across and so much up. Cutting the screen in half doesn't do anything to the "FOV" sicne you can set that to whatever value you like no matter what shape the screen is, what cutting the screen in half does is affect the relationship between the horizontal FOV and vertical FOV.

If you're playing on a 16:9 monitor and you have the in game FOV set to 90 horizontally (you actually set the vertical FOV on widescren instead but we'll just say you set the horizontal one to keep things simple)then you will have about what... 50 or so vertically since the monitor is almost 1.8x as wide as it is high.

So if you simply cut the screen in half then yes you would only have half the horizontal FOV as it would only be 45 now, but the vertical FOV is still 50. That doesn't matter for how much FOV you see though as you can still change that 45 horizontal FOV to any (well technically not any) number you wish.

If with the screen cut in half tell it to display a horizontal FOV of 90 you will still see just as much horizontally as you would see on a widescreen monitor (just at a lower resolution at current) swith a horizontal FOV of 90, the difference is while on the monitor viewing a 90 FOV it is only able to display that 50 FOV vertically (because of it's physical shape, there simply isn't anythere "there" to display more vertical FOV if you keep things at a normal aspect) while on the RIFT with it only using half the screen there is physically more screen and pixels going vertically (in relation to horizontally) so you get the in this case 2x larger vertical FOV.

So rather then losing half the horizontaly FOV (since you tell it to display 90 like on the monitor and you stil lsee just as much across as you do on the monitor) you simply doubled the vertical FOV instead, the draw back being you lose resolution at the same time.
Endothermic
Binocular Vision CONFIRMED!
Posts: 284
Joined: Tue Jun 26, 2012 2:50 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Endothermic »

LikeMike wrote:Thanks everybody for the replies... that was quick. A quick follow up here:

With the standard resolutions standards - is it safe to say that the vertical fov is always larger than the horizontal (since the horizontal needs to be "cut in half" for each eye) in practically all hmd?
Only if the HMD is using a single screen then yes, unless the screen is =/> 2x as wide as it is high then vertical FOV will always be higher then horizontal
LikeMike wrote:Isn`t the extra vertical fov "wasted" - meaning that you can`t see it anyways since the natural fov seems to be for a lack of a better word widescreen?
Have a look when your watching tv, can you still see the ceiling and the floor infront of you? Or walk outside, can you still see the sky above and the ground bellow? Playing a game on a widescreen monitor you do not (of course you do when it's far away but not close like you do IRL)

What is in that extra vertical FOV will be blurry and non detailed since it is in your periphery but you will still notice it, the colours, movement etc, so it definately increases the immersion over not having it.
LikeMike wrote:That confuses me a little... people that tried it usually say something like: you can`t see any borders - or, only if you try really hard can you see a little black border. So it seems like almost all the eye sees is the screen. I know, that a persons fov is usually a lot bigger than 90, I though the fisheye lenses and stuff like that just kinda "expanded" that to make it seem, like there are no borders (I don`t know much about the technical side, so I may be dead wrong here). But if thats the case, it should "expand" the horizontal view just as much as the vertical view, shouldn`t it? So if the expanded 90 for horizontal is "almost" enough to give you a borderless view, the expanded 110 for the vertical view should be "too much" - since your vertical view isn`t as wide as the horziontal view anyways.

Did I understand anything wrong, or are there more borders to be seen than I thought?
Take the hands to the side example given above, if you do that then you will notice you have less horizontal FOV since you can't see anything past your hands, but they are in your periphery so they are blurred and you don't see a "border" at the end of your FOV where you're hands are unless you actually turn you're eyes and look at it.

You will also notice after a while you're hands start to just kind of fade away and blend in (a kind of similar thing to when you put socks or underwear on and can feel and tell you have it on for a little but after a while you don't notice you're wearing them). This with the periphery basically helps to eliminate the border so that you notice you don't have as much a FOV but there is no distinct end to you're FOV unless you turn you're eyes and look at it.

The same thing is happening with the screen, the border is there but because it is in your periphery and all outside light is blocked out so that you can not see anything past that border it ends up just blending in with your vision, combined with it being blurred by you're periphery you simply end up not being able to distinguish it. You can tell that your FOV is smaller then IRL and you're not seeing everything you would be able to, but you are not seeing a distinct end to you're FOV unless you turn you're eyes to find it so after a while it seem's like there is no border.

Unless the HMD has a really small FOV in which case you do see the border since it is no longer in your periphery and closer to your foval area where things are getting in focus.
Mystify
Certif-Eyed!
Posts: 645
Joined: Fri Jan 11, 2013 5:10 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Mystify »

Endothermic wrote:
LikeMike wrote:Hello there, newbie here. I am really excited about the Rift - but I still have a couple of questions, perhaps some of the people here can answer:

1. As I understand it and see it in the videos, the screen is basically divided in two halfs showing basically the same thing - doesn`t that mean, that we have an about half as wide field of view than we would have on a monitor? And basically a larger vertical field of view than a horizontal? To me that seems kinda odd...
It's not just a matter of the screen being cut in half so the FOV is cut in half, you need to consider the horizontal FOV in relation the the vertical FOV, and it's that larger vertical FOV you mentioned which means the horizontal FOV isn't actually cut in half.

FOV in a game i just a number that is set, then depending on the shape of the display you'll see so much game world across and so much up. Cutting the screen in half doesn't do anything to the "FOV" sicne you can set that to whatever value you like no matter what shape the screen is, what cutting the screen in half does is affect the relationship between the horizontal FOV and vertical FOV.

If you're playing on a 16:9 monitor and you have the in game FOV set to 90 horizontally (you actually set the vertical FOV on widescren instead but we'll just say you set the horizontal one to keep things simple)then you will have about what... 50 or so vertically since the monitor is almost 1.8x as wide as it is high.

So if you simply cut the screen in half then yes you would only have half the horizontal FOV as it would only be 45 now, but the vertical FOV is still 50. That doesn't matter for how much FOV you see though as you can still change that 45 horizontal FOV to any (well technically not any) number you wish.

If with the screen cut in half tell it to display a horizontal FOV of 90 you will still see just as much horizontally as you would see on a widescreen monitor (just at a lower resolution at current) swith a horizontal FOV of 90, the difference is while on the monitor viewing a 90 FOV it is only able to display that 50 FOV vertically (because of it's physical shape, there simply isn't anythere "there" to display more vertical FOV if you keep things at a normal aspect) while on the RIFT with it only using half the screen there is physically more screen and pixels going vertically (in relation to horizontally) so you get the in this case 2x larger vertical FOV.

So rather then losing half the horizontaly FOV (since you tell it to display 90 like on the monitor and you stil lsee just as much across as you do on the monitor) you simply doubled the vertical FOV instead, the draw back being you lose resolution at the same time.
The issue isn't in-game FOV, its the actual FOV of your vision the display takes up- for VR, ideally, this should be the same as the in-game FOV.
bobv5
Certif-Eyed!
Posts: 529
Joined: Tue Jan 19, 2010 6:38 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by bobv5 »

I imagine the Rift is like wearing a motorbike helmet, you lose FOV, but you lose the bit you don't really need anyway. I have never worn an army helmet, but apparenty it gives a similar FOV reduction to rift. The amount of money army's have, if it was a problem, they would redesign the helmets.
"If you have a diabolical mind, the first thing that probably came to mind is that it will make an excellent trap: how do you get off a functional omni-directional treadmill?"
Endothermic
Binocular Vision CONFIRMED!
Posts: 284
Joined: Tue Jun 26, 2012 2:50 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Endothermic »

Mystify wrote: The issue isn't in-game FOV, its the actual FOV of your vision the display takes up- for VR, ideally, this should be the same as the in-game FOV.
That wasn't what his "doesn't using half the screen mean you only get half the FOV you would have on a monitor" question was about though as it was about how much of the game world you see not how much of your vision it is actually covering.

Atleast that's how I read it, like someone was asking "if i'm playing on a monitor then I see 90 degrees of the game world so if you split the screen in half then wouldn't I only be seeing 45 degree of the game world?". How much of your actual vision that image cover's is irrelevant, it's purely about how much of the in-game world is viewable.

You could have the screen in the RIFT far enough away or use coresponding optics so that the actual H-FOV of your sight the RIFT covers is the same as the amount that is covered by looking at a monitor, and set the in-game FOV to match it, let's say 35 degrees.

In this case both the monitor and the RIFT are covering 35 degrees of you're view and both are showing you 35 degrees of H-FOV in-game. But the RIFT (assuming 100% overlap to keep things simple) would still only be using half the screen (which is the same widescreen format as the monitor) to display that though, so you end up with much more V-FOV on the RIFT then you do on the monitor while they are both showing 35 degree H-FOV and both take up 35 degree's of your actual vision. So using only half the screen has not halved the H-FOV like it seemed it would (because assuming cutting the screen in half horizontally means cutting the FOV in half horizontally) it instead results in additional V-FOV being viewable, but at the cost of lower resolution, if you still use the same value for the H-FOV.

And in-case someone jumps on my simplification of V-FOV being about 50 for a H-FOV of 90 before, because a widescreen is about 1.8x as wide as it is high, I know that's not right and quite a bit off from using the actual FOV equation, but it seemed good enough for the whole "half screen half FOV" discussion and simpler seemed better to get the point across.
User avatar
coresnake
Two Eyed Hopeful
Posts: 75
Joined: Fri Jun 22, 2012 5:32 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by coresnake »

Sorry if this has been addressed already, but will the consumer version absolutely have the same FOV or is this something they are looking to improve as with the resolution? Or do we need a curved OLED screen or such before that is possible?
Endothermic
Binocular Vision CONFIRMED!
Posts: 284
Joined: Tue Jun 26, 2012 2:50 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Endothermic »

It won't "absolutely" have the same FOV but it will likely be the same or very close.

They want a larger FOV, we want a larger FOV, but the main concern with a consumer model is positional tracking and the resolution since as it stands the FOV now is "good enough" and very immersive, but regular gamers will no doubt complain about the resolution if it isn't any better and to a lesser extent not being able to stick their head under a table etc, and a higher FOV means also needing a higher resolution to keep the same percieved resolution.

I believe the FOV is high enough that there is no point increasing the FOV (a meaningful amount like say to 120) if the resolution isn't increased, and not just increased to match the additional FOV, but increased in so the pixels are small as well. I don't think having a FOV 120+ (and resolution increased only so much that the pixels are the same size as they are currently) where you still can't read text that good and objects/people in the distance are just a few pixels in size, would really be much benefit over a 90 FOV with the same pixel size.

It will be a while before curved or flexible oleds are cheap enough at the resolutions needed, so we wont have any extra large FOV HMDs in the price range of the RIFT anytime soon, but I don't see there being any problem with the 90 FOV untill then, especially if resolution increases.
cerulianbaloo
Binocular Vision CONFIRMED!
Posts: 297
Joined: Mon Sep 03, 2012 3:36 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by cerulianbaloo »

Sorry this is somewhat off-topic, but your mention of using positional tracking to poke your head under a table reminded me of this quote from a dailymail article that popped up in my google alerts today for the Rift, "You’re inside the video game – lunging forward with your head to move forward, with phone-style accelerometers telling the system what you’re doing." That one gave me a chuckle. Now I could see that pose looking appropriate perhaps for a skiing simulation but not much else. There's a few other misinformed gems to be found in the article: http://www.dailymail.co.uk/home/moslive ... atrix.html
Endothermic
Binocular Vision CONFIRMED!
Posts: 284
Joined: Tue Jun 26, 2012 2:50 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Endothermic »

Well this thread is just about the RIFT so it's not really off topic :P

They do use the head/neck model in the demo's and doom 3 which takes into account that when you look down you don't just rotate down but move forward a little etc, you can see it in the unreal demo where people go over to the fruit tables and instead of just looking down their view actually looks over the table a little bit liek they are indeed peering over it.

Maybe when he saw that bit of shift happening when he looked down he just assumed that his head going forward was actually moving his character forward or something.... it's the only time i've ever heard anyone say that lunging forward makes you actually move forward (which would be just stupid anyway, who wants to move forward by having to push their head forward :|) so he is obviously getting that impression from something since it definately isn't happening.

"you can explore video games such as Doom 3 and Unreal Tournament, where lunging forwards makes you move in the game"
"lunging forward with your head to move forward,"
"But whether buses to work will soon be full of people lunging forward like pecking chickens remains to be seen."

Wow... really... what is his obsession with "lunging" especially considering it's something you can't do, nor something you would even wan't to do either :|
User avatar
Randomoneh
Binocular Vision CONFIRMED!
Posts: 227
Joined: Wed Oct 17, 2012 12:42 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Randomoneh »

Endothermic wrote:If you're playing on a 16:9 monitor and you have the in game FOV set to 90 horizontally then you will have about what... 50 or so vertically since the monitor is almost 1.8x as wide as it is high.
on the [16:9] monitor viewing a 90 FOV it is only able to display that 50 FOV vertically (because of it's physical shape, there simply isn't anythere "there" to display more vertical FOV if you keep things at a normal aspect)
90 horizontally on 16:9 display will give you almost 60 vertically (58.7).
Endothermic wrote:So if you simply cut the screen in half then yes you would only have half the horizontal FOV as it would only be 45 now, but the vertical FOV is still 50.
If you would cut the screen in half you would end up with 53.13 horizontally and 58.7 vertically.
Physical aspect ratio doesn't matter. You can have (and see, perceive) horizontal FOV of 179 and vertical FOV of 178 on a 16:9 display.
bobv5 wrote:I imagine the Rift is like wearing a motorbike helmet, you lose FOV, but you lose the bit you don't really need anyway. I have never worn an army helmet, but apparenty it gives a similar FOV reduction to rift. The amount of money army's have, if it was a problem, they would redesign the helmets.
No. Peripheral vision is very important for a sense of your surroundings. With peripheral vision you can sense someone walking in the room, someone trying to sneak from one of your sides, someone sitting or standing next to you. Without it, we don't really feel safe. However, coming from 40 degrees, it's an achievement and one huge step towards full (~270) horizontal FOV. No need to worry, it won't be completely natural but I think it'll be more than you expect from a display.
This member owns things.
User avatar
drifter
Binocular Vision CONFIRMED!
Posts: 242
Joined: Tue Nov 27, 2012 9:39 am
Location: Little Britain

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by drifter »

Endothermic wrote:Wow... really... what is his obsession with "lunging" especially considering it's something you can't do, nor something you would even wan't to do either :|
ho well, generalist newspapers and new techs...
LikeMike wrote:2. How easy would it be to get UDK or Unity creations into the rift? Is it basically just checking a box ("add support for Rift") - or do I have to really code it in there somehow?
That should be easy in Unity : - import package (DLL and script) - add script to your camera.
Theoretically no extra code needed.
3. Positional tracking seems to be the biggest thing missing from the developer kit - is that something that could be added later with a software update, or is the necessary hardware missing from the devkit?
No hardware now but it *may* be added to the dev kit later (more logically before the consumer version launch).
5. Any idea when the consumer version comes out?
They keep saying "in months not in years". I hope for the end of the year... A bit before Christmas would be a cool spot.
Work to do : launch the devkit, get the devs feedbacks and take it into account, find a better res display (without exploding the budget) and put it in the Rift, find a good positional tracking solution (again without exploding the budget) and put it in the Rift, update the SDK for all of that.
edit : Am I good ?
Last edited by drifter on Sun Jan 20, 2013 1:11 pm, edited 2 times in total.
User avatar
MrGreen
Diamond Eyed Freakazoid!
Posts: 741
Joined: Mon Sep 03, 2012 1:36 pm
Location: QC, Canada

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by MrGreen »

This thread needs more Palmer damnit! :)
User avatar
drifter
Binocular Vision CONFIRMED!
Posts: 242
Joined: Tue Nov 27, 2012 9:39 am
Location: Little Britain

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by drifter »

MrGreen wrote:This thread needs more Palmer damnit! :)
yeah, a bit too much of calculations and conditional around here :)
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by PasticheDonkey »

there is a limit to the FOV the current lens system can give. at the moment a higher FOV would make the resolution too low tho. but in the future we want increased resolution and FOV with new displays and optics. but it's hard to make a one size fit's all solution when it comes to higher FOVs.
LikeMike
One Eyed Hopeful
Posts: 25
Joined: Sat Jan 19, 2013 10:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by LikeMike »

Alright, other question: apparently you need a stable 60 fps for convincing VR - now I've read several times, because of the stereo 3D you actually need 120 basically, since the image has to be rendered twice. Why is that? The screen is basically cut in half, so there are 2 images created that together nets you the resolution 1280x800 - that should need about the same resources as 1 full image, since the same number of pixels are rendered, shouldn't it?

Also I've heard Palmer say, that they are thinking about adding something that lets you adjust the dioptren for near sighted people, while normal or far sighted are fine... Why are far sighted people fine? I am far sighted, and while I can see alright without my glasses, I'd definetly prefer to have them on while... So is there a way to help us far sighted people as well, or do we really don't need any help on this?
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by cybereality »

@LikeMike: Its not just about the number of pixels. When you do a full stereo render you also have to transform all the polygons to a different angle, and this is basically re-rendering the scene (as you can see things from a different angle, occluded objects, etc,). Of course the main game loop, and things like AI, physics, etc. do not need to be computed twice. So if a developer is doing their own native stereo render, then performance could potentially be better than 50% of 2D performance. But usually with 3rd party drivers that's the performance you can expect.
LikeMike
One Eyed Hopeful
Posts: 25
Joined: Sat Jan 19, 2013 10:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by LikeMike »

Thanks for the reply... So how does this work - is the image first rendered normally and then again for the stereo 3D - or is there something like a stereo 3D filter layered over the image before it is rendered (and that filter being very, very hardware demanding). Because if it is the first, then I'd imagine a game specifically designed for the Rift wouldn't have any slowdowns, since the game wouldn't have to be rendered 2 times, but could be rendered at once with the stereo 3d effect already calculated in...

I always thought the slow down for monitors with shutter glasses was, because the image has to be rendered twice, once for each eye - a problem the Rift should not have, since you could view the 2 small images as one large image.

Sorry if I sound dumb or naive - I really don't know much about the technical stuff, but I find it very interesting!
MaterialDefender
Binocular Vision CONFIRMED!
Posts: 262
Joined: Wed Aug 29, 2012 12:36 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by MaterialDefender »

There are two possible solutions. The first one does what Cybereality described, it renders everything twice from a slightly different angle. This is slower than just rendering one view, regardless whether it's done by the game itself or not. Unfortunately it doesn't help whether a game is designed for the Rift or not.

The second solution creates the stereo images out of one rendered image and additional information provided by the game. This is done as a kind of 'filter' and is much faster.

Both methods have their strengths and weaknesses.
Endothermic
Binocular Vision CONFIRMED!
Posts: 284
Joined: Tue Jun 26, 2012 2:50 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Endothermic »

The game still has to be rendered twice or you only have one perspective so it isn't 3d. You can off set that image to create a 3d effect but it isn't as good.

I could be wrong but I believe when people make reference to 120fps they are referring to needing to be able to run the game in 2d at the 1280x800 resolution at 120 fps in order to be able to have the game running at 640x800 at 60 fps in 3d.

The difference being with shutter glasses playing in 3d the game needs to run at 120fps in order to get the 60 fps for each eye in 3d since it only shows the frame for one eye at a time. So the frame rate on screen has to be 120 fps.

With the RIFT it is showing the frame for both eyes at the same time instead of one after the other, so for each eye to see 60 fps ithe frame rate on screen only has to be 60 fps not 120 fps.

But it still has to render a seperate frame for each eye ( so it is still rendering 120 frames per second just that it is showing 2 of them at a time so on screen the frame rate is only 60 fps not 120 fps) so in order for it to be running at 60 fps you need to be capable of running the game in 2d at almost double that ( since you rendering twice as many frames ) even though each frame only has half the horizontal resolution, because of the reasons cyber outlined above.

Think of it as if you were playing on a monitor with shutter glasses then it has to render 120 frames each second as well as display those 120 frames each second, while on the RIFT it still has to render 120 frames per second but it only displays 60 frames per second ( since it combines two of those frames into one to display).
German
Binocular Vision CONFIRMED!
Posts: 209
Joined: Tue Nov 23, 2010 5:18 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by German »

The fps don't change for the way things need to be rendered for the Rift, it's always 60fps. The amount of geometry that needs to be rendered is approximately doubled. Subtle distinction but that's what it is.
LikeMike
One Eyed Hopeful
Posts: 25
Joined: Sat Jan 19, 2013 10:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by LikeMike »

Endothermic wrote:
Think of it as if you were playing on a monitor with shutter glasses then it has to render 120 frames each second as well as display those 120 frames each second, while on the RIFT it still has to render 120 frames per second but it only displays 60 frames per second ( since it combines two of those frames into one to display).
Actually that's exactly what I'm thinking about, on a monitor you'd have to have 120fps@1280x800 - on the Rift you need 120fps@640x800 or 60fps@1280x800 - so theoretically rendering on the Rift should be a lot faster. But if course there is also this other thing, the fish eye correction that they have to do - but as they stated themselves, that doesn't take up a lot of performance.

Basically what I am asking - since the effective resolution is 640x800, shouldn't the game run much faster on a rift, than on a monitor with an effective resolution of 1280x800 in 3D? And if that is the case, wouldn't that mean that you don't need 120fps on a 2D 1280x800 monitor to be able to run it with 60fps on a Rift?
MSat
Golden Eyed Wiseman! (or woman!)
Posts: 1329
Joined: Fri Jun 08, 2012 8:18 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by MSat »

As mentioned earlier, for typical stereoscopy, the geometry needs to be set up and rendered twice to make one "Rift frame".So while a 640x800 mono frame may be faster to render than a 1280x800 mono frame, the fact that you have to set up and render two different frames is where the performance hit comes from.
LikeMike
One Eyed Hopeful
Posts: 25
Joined: Sat Jan 19, 2013 10:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by LikeMike »

MSat wrote:As mentioned earlier, for typical stereoscopy, the geometry needs to be set up and rendered twice to make one "Rift frame".So while a 640x800 mono frame may be faster to render than a 1280x800 mono frame, the fact that you have to set up and render two different frames is where the performance hit comes from.
Yeah, I understand that - rendering two different frames takes about a 50% performance hit, thus getting a 3D image of the same resolution basically cuts your fps in half. What I am arguing or not understanding correctly is: you have to treat the Rift as a 640x800 3D display in this calculation if I am correct. Because it renders only one image for the left half and one image for the right half and not, like monitors, two images that both take up all of the resolution.

So when you want to calculate the performance loss it should be 50% of what your computer can run the game in 640x800, or you could also say: you need a computer that us able to get 120fps while the game is running at 640x800 - not one that gets you 120fps while the game is running at 1280x800 in 2D.
MSat
Golden Eyed Wiseman! (or woman!)
Posts: 1329
Joined: Fri Jun 08, 2012 8:18 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by MSat »

LikeMike wrote:
MSat wrote:As mentioned earlier, for typical stereoscopy, the geometry needs to be set up and rendered twice to make one "Rift frame".So while a 640x800 mono frame may be faster to render than a 1280x800 mono frame, the fact that you have to set up and render two different frames is where the performance hit comes from.
Yeah, I understand that - rendering two different frames takes about a 50% performance hit, thus getting a 3D image of the same resolution basically cuts your fps in half. What I am arguing or not understanding correctly is: you have to treat the Rift as a 640x800 3D display in this calculation if I am correct. Because it renders only one image for the left half and one image for the right half and not, like monitors, two images that both take up all of the resolution.

So when you want to calculate the performance loss it should be 50% of what your computer can run the game in 640x800, or you could also say: you need a computer that us able to get 120fps while the game is running at 640x800 - not one that gets you 120fps while the game is running at 1280x800 in 2D.
Just because the resolution is cut in half, doesn't mean everything else in the scene is. If a scene has X number of polygons, that's true regardless of the resolution - the only thing that's reduced is the number to texels. So no matter the resolution, the system has to transform and cull the same number of vertices. Try this - run a game where you can display the frame rate. Select a resolution and note the FPS, then switch to a resolution that is roughly half of that and again note the FPS. I'll be surprised if your frame rate doubles.

And yes, it's safe to say you roughly need to get 120fps @ 640x800, and not 120fps @ 1280x800.. Yet, it's still not as clear cut as that. From my understanding, if you want to get maximum detail from the rift, one way to do it would be to render the scene above 640x800 since the Rift's optics are going to squeeze the pixels towards the center. Otherwise, you either don't preserve the correct dimensions of the objects, or you slightly upscale them in the warp shader and lose the opportunity to take advantage of potentially additional detail.
Owen
Cross Eyed!
Posts: 182
Joined: Mon Aug 13, 2012 5:21 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Owen »

Still not quite that simple. Things like shadows and environment maps still only need to be computed once per frame, not once per eye. Not to mention animation, culling, sorting, GI, and any non-graphical CPU tasks.

Though I wouldn't expect Unity to take advantage of that at least. Their rendering pipeline is really inflexible. Basically when you draw the scene, it doesn't return scripting control until the whole rendering process has cleaned itself up and thrown away everything but the resulting image (even by the time it draws transparent stuff in the same pass, its already discarded any shadow maps), so reuse of anything is basically impossible unless they make drastic changes. So with Unity it is probably about the same as 120hz, but for a custom engine it generally won't be unless naively implemented.

For Unreal who knows, but since they have had split screen in shipping titles I would guess that they can do it pretty efficiently, since consoles don't leave you a lot of headroom to half-ass it.
phort99
One Eyed Hopeful
Posts: 27
Joined: Tue Jan 01, 2013 2:25 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by phort99 »

Owen wrote:Still not quite that simple. Things like shadows and environment maps still only need to be computed once per frame, not once per eye. Not to mention animation, culling, sorting, GI, and any non-graphical CPU tasks.
The shadow map is rendered once per frame. Rendering the actual shadows is also pretty expensive and that happens once for each eye. Baked shadows are not rendered once per frame, they're typically precomputed or computed much less often, but they can be applied cheaply.
Paladia
Cross Eyed!
Posts: 154
Joined: Mon Jan 21, 2013 5:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Paladia »

Will the developers version or the consumer version be running at 120 hz? As 120 hz is so much more comfortable and life-like to look at.
pewpewk
One Eyed Hopeful
Posts: 18
Joined: Fri Aug 03, 2012 4:42 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by pewpewk »

Paladia wrote:Will the developers version or the consumer version be running at 120 hz? As 120 hz is so much more comfortable and life-like to look at.
Developer? No. Consumer? Nothing confirmed. It's possible, but not sure if it's likely. Only time will tell and I doubt even Oculus knows the answer yet.
Paladia
Cross Eyed!
Posts: 154
Joined: Mon Jan 21, 2013 5:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Paladia »

pewpewk wrote:
Paladia wrote:Will the developers version or the consumer version be running at 120 hz? As 120 hz is so much more comfortable and life-like to look at.
Developer? No. Consumer? Nothing confirmed. It's possible, but not sure if it's likely. Only time will tell and I doubt even Oculus knows the answer yet.
Thanks for your reply. I really hope it is 120 hz as that is the minimum requirement for anything but casual gaming, in my opinion. Even my four year old 2233RZ is running at 120 hz.

So I hope for either 120 hz by default or a 120 hz alternative.
2EyeGuy
Certif-Eyable!
Posts: 1139
Joined: Tue Sep 18, 2012 10:32 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by 2EyeGuy »

My laptop is 120Hz. But it keeps overheating and switching itself off. 120Hz would be ideal for head tracking, but I don't know if computers can perform that fast to make it worthwhile or even if HDMI can support that rate.
Paladia
Cross Eyed!
Posts: 154
Joined: Mon Jan 21, 2013 5:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Paladia »

2EyeGuy wrote:My laptop is 120Hz. But it keeps overheating and switching itself off. 120Hz would be ideal for head tracking, but I don't know if computers can perform that fast to make it worthwhile or even if HDMI can support that rate.
It depends on the resolution and the HDMI version. HDMI 1.4b and onwards supports 120 hz at 1080p.

Regardless, Oculus Rift uses DVI, which supports 120 hz at 1080p (dual link).
Owen
Cross Eyed!
Posts: 182
Joined: Mon Aug 13, 2012 5:21 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Owen »

Rendering out shadows from a shadow map is just multitexturing, its not that expensive. Also its a fixed per pixel cost, so 1280x800 vs 2x640x800 makes no difference.
Kazioo
Two Eyed Hopeful
Posts: 81
Joined: Thu Aug 09, 2012 10:17 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Kazioo »

I'm 99% sure first consumer version of Rift will be 60hz. 120hz would be a total overkill, like 1600p resolution.
There are no 120hz mass produced screens available and even devkit requires a beast computer for modern AAA games.
120 x 2 = 240 FPS. Only old games like Half-Life 2 or some simple tech demos would be able to take a real advantage of 120hz screen.
Paladia
Cross Eyed!
Posts: 154
Joined: Mon Jan 21, 2013 5:26 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Paladia »

Kazioo wrote:I'm 99% sure first consumer version of Rift will be 60hz. 120hz would be a total overkill, like 1600p resolution.
There are no 120hz mass produced screens available and even devkit requires a beast computer for modern AAA games.
120 x 2 = 240 FPS. Only old games like Half-Life 2 or some simple tech demos would be able to take a real advantage of 120hz screen.
Even if it is just running 50 fps a 120 hz screen is superior. It is more comfortable to look at and will always be able to draw every new frame, where as a 60 hz screen will have to randomly add double frames, making it more jerky. Not to mention that a 120 hz screen has less input lag as the screen is drawn more often.
German
Binocular Vision CONFIRMED!
Posts: 209
Joined: Tue Nov 23, 2010 5:18 pm

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by German »

Paladia wrote:Even if it is just running 50 fps a 120 hz screen is superior. It is more comfortable to look at and will always be able to draw every new frame, where as a 60 hz screen will have to randomly add double frames, making it more jerky.
120 hertz = 120 times per second = 120 frames per second. 50 frames per second doesn't evenly divide into 120 or 60. Both would have to re-draw(not insert) frames. The only difference is the amount of gimmicks they can cram into 120hz displays to handle the 70 frames of missing content.
Laserschwert
Cross Eyed!
Posts: 108
Joined: Mon Aug 06, 2012 11:28 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Laserschwert »

Owen wrote:Rendering out shadows from a shadow map is just multitexturing, its not that expensive. Also its a fixed per pixel cost, so 1280x800 vs 2x640x800 makes no difference.
But don't forget stuff like Screen Space Ambient Occlusion (SSAO), which - as the name suggests - renders ambient occlusion shadowing in screen space, thus it would be necessary to render it for each eye separately.
User avatar
PasticheDonkey
Sharp Eyed Eagle!
Posts: 450
Joined: Sun Jan 06, 2013 4:54 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by PasticheDonkey »

what you need is virtual screen space which covers the areas seen by both views.
Kazioo
Two Eyed Hopeful
Posts: 81
Joined: Thu Aug 09, 2012 10:17 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Kazioo »

Paladia wrote: Even if it is just running 50 fps a 120 hz screen is superior.
But is it worth using more expensive screen? Furthermore, Oculus claims 60fps with vsync is a must.
Laserschwert
Cross Eyed!
Posts: 108
Joined: Mon Aug 06, 2012 11:28 am

Re: Oculus "Rift" : An open-source HMD for Kickstarter

Post by Laserschwert »

PasticheDonkey wrote:what you need is virtual screen space which covers the areas seen by both views.
I don't think that would work... if I'm not mistaken screen space effects are purely 2D. Maybe creating the screen space effects for the second eye via the depth buffer (like what MaterialDefender does with his VorpX driver) might be an option?
Post Reply

Return to “Oculus VR”