Official Oculus Rift Consumer version one announcement

Speaking about the screen and resolution, there is something no one has mentioned here and it is the most important thing of them all. There was quite a bit of backlash and general disappointment with the final resolution of the Rift, so Palmer Luckey threw in some very interesting comments, especially one in particular:

High fill and global update. This, is HUGE.

A 4K screen, the one everybody was asking for, manufactured with the same technology as Note 4's 1440p display for example, would have a horrible screen door effect, even if people seems to think the opposite. You would have smaller pixels, but the space between them would be bigger and therefore more screen door visible. High fill means that the space between pixels is reduced to as low as possible, so I doubt we will have screen door problems with the consumer products.

Global update, could be even more critical. Means the pixels of the screen update all at once, instead of the typical horizontal/vertical scan. This could eliminate the need of having to use Vsync at all, although it's not completely clear that this will be the case.

To put things into perspective, people who tried the Crescent Bay prototype and Samsung's GearVR back in September, have said that Crescent Bay looked like to have a higher resolution than GearVR, or at least the same 1440p. Some even said it was possible that the screen was 4K, so that gives you the idea that it really had to look better than GearVR. Well, being worse than CB, GearVR already looks much better than DK2, not just 25%, and Palmer Luckey said that CV1 will look even better than Crescent Bay. This automatically means that CV1 will look a lot better than GearVR, and GearVR has a 1440p screen. It's not all about resolution, the technology of the display and especially the optics play a much bigger part in the actual perceived resolution. Have no doubt the jump in perceived resolution between DK2 and CV1 will be much higher than between DK1 and DK2, even if the numbers aren't all that different.

Awesome times are coming boys, can't wait

 

"Global Update" sounds the same to me as "Lightboost", "Blur Reduction Mode", "ULMB" ("Ultra Low Motion Blur"), etc. I thought the Rift guys already called this "Low Persistance"? Are "Low Persistance" and "Global Update" actually two different things then?

BTW, the screen's pixels all refreshing at the same time mean that stutters and tearing become even more noticeable when your game's framerate doesn't perfectly match the screen's refresh rate. Therefore game-framerate-to-monitor-refresh-rate syncing (eg. VSync/FreeSync/GSync) is even more vital for a perfect image. This is the current case with "regular" monitors capable of "Global Update"/"Low Persistant" type of modes; maybe it's different with VR devices?...

 

Someone correct me if I'm wrong but it sounds like global update means that all the pixels get updated simultaneously so there is no screen tearing. But I also think that can't be if it's stated as running at 90hz fixed....if global update is what I think it means, that would require a g-sync like technology. Or perhaps they just have very low latency forced hardware vsync solution on the rift.

 

It is nothing like that. Global update means the frame will update as a whole when it is drawn. Tearing happens because the screen is showing half of one frame and half of the next. This is impossible to happen if all the pixels are updating simultaneously. Every time the display refreshes, it will take the most recent frame available drawn by the GPU and show it as a whole instantly.

Normal displays, if we take a normal 60Hz monitor, will spend 16.6ms doing a top/bottom scan. If you get a new frame from the GPU when the scan is halfway through, you will get half of one frame, and the other half will be the new frame, producing a tear just in the middle of the screen.

I can understand that you don't know how it works. What I don't understand is how you can think that a team of VR experts would choose that kind of display if it was worse in any way. Obviously it's because it is a thousand times better, tearing is impossible with global display. And no, Low persistence and Global update are two entirely different things

 

Cool. "Traditional" monitors that are capable of refreshing the entire screen's pixels at once rather than top to bottom scanning (I think the scanning is called "sample and hold") use a stroboscopic (strobe, flash) mode. It's why monitors with that feature have AMAZINGLY low motion blur because of how all the pixels are off/on. I'm guessing that this same type of feature is what OR refers to as "low persistence", because that's what it achieves.

However, unless some sort of sync is used (eg. GSync, VSync, FreeSync) to sync the frames with the monitor's refresh rate, tearing will still be easily noticeable. You need them perfectly matched-up to avoid tearing (and stutters as well). I'm guessing that the VR devs are doing something then to sync the game's frames to the screen's refresh rate, like VSync/ GSync/FreeSync, or else how can the frames be perfectly matched up with the screen's refresh rate? I'd love to know how they're doing this. Thanks.


P.S. No, I don't think the devs are choosing a worse display. I'm not sure where you draw such conclusions from; I'm simply trying to understand what they're doing and how it works. Simple as that.

 

I run at high fps, no type of vsync what so ever, never seen tearing in my dk2

( I cap in Iracing at 205fps but it's fine anything over 100fps )

 

source --> https://forums.oculus.com/viewtopic.php?f=26&t=10324&sid=8ab34eb0dc39d7a4610bf1f192ef58bb&start=20

 

Maybe the screens are 2160×1200 @ 90Hz, but they use something like nvidia drs to reach higher resolution effect. Just guessing here.

 

Close but small correction. UMLB/Lightboost (stroboscopic) monitors still operate by top to bottom scanning of the pixels but instead of keeping the backlight constantly illuminated, they strobe the backlight. By strobing the backlight, the user sees the image for only a short interval of the frame time before going black for the remainder of the frame thereby removing the sample and hold effect (which is a form of "pixel persistence") that is the predominant cause of display motion blur in games:

(left = desired perceived result , middle = strobing = no/low sample and hold effect = low persistence , right = non-strobing = sample and hold effect = high persistence)



Of note, this illustration is of a perfectly in-sync render and display. If however, someone used v-sync and the fps suddenly dropped below the display refresh rate for one frame, you will get two sequential frames with the same render image which will appear as a sudden stutter/pause in the flow of the render. This problem remains exactly the same for both strobed and non-strobed displays but I think having motion blur may actually help mask any stutters present which makes getting the render and display in sync all the more important for motion blur-less displays.

possibly but then i ask why cut it in two to produce two separate 1080x1200 displays.

 

Exactly Dr1, stutters are even more noticeable when you have extremely low motion blur, having a monitor with terrible motion blur (which even a "1ms" monitor does when it's not in low persistance mode like for eg. ULMB) does have a positive knock-on effect of slightly making such anomalies less obvious.

Ya, thanks for clarifying. Technically speaking it still does sample and hold, but only invisible "in the background" as the end-result is an entire image appearing at once.

Regarding the out of sync issues. Yes, if you can't maintain the VSyned framerates then of course you will get some stutters because, well, VSync isn't really doing anything since your framerate isn't high enough, non-VSync though is much, much, much more prone to stutters, not to mention tearing. I don't think you can compare them to eachother in terms of fluidity, stutters, and tearing.

Not that I'm defending VSync for racing sims as I want ultimate lowest latency/input lag as possible.

 

Yes, your right. I should have said "strobing makes the pixel sample and hold effect imperceptible".

Perhaps. It's been a long time since i've used my home computer. I know that when i was playing rf2 however i always used adaptive sync and light boost mode, probably because as you say, it's worse when the fps is above the native refresh rate.

 

Sorry guys but, rFactor2 never run on a Oculus virtual reality system. Isi lacks the level of programming required. besides having a game in DX9. something that will become incompatible with Oculus.

 

I can't say I know the exact numbers involved, but I know the mechanism behind low persistence and the benefits a global update display would bring.

Let's put, for example, a 90Hz display, which shows a frame every 11.11ms. I don't know exactly how long it takes for a normal display to do the full scan from top to bottom, but let's say it takes 1ms since the first pixels from the top start updating, until the last ones from the bottom finish. What low persistence does, is showing you the frame for the first 2ms, and then turns the display to black so that it doesn't smear against your retina (so that it doesn't "persist", hence the name). It goes on again after 9.1ms to show you the next frame for another 2ms, and goes off again, and so on. What is the problem? You need vsync with a normal display, because otherwise, if you get a new frame from the GPU 0.5ms after the screen had already started scanning the last one, you will get half of the old and half of the new (tearing). A global update display doesn't have that 1ms window or whatever for this kind of thing to happen, it just updates simultaneously when it is time to refresh, and it will show the most recent frame at that exact time. You cannot get a new frame in the middle of an update, because the update happens all at once.

And obviously, if you drop frames will be bad in both, but it is worse if you have Vsync on

I don't know how to explain it better, sorry...

 

I thought the time taken to perform a top to bottom scan of a display was almost the same as the time interval of one refresh cycle (e.g. 8.3ms for 120hz and 16.7ms for 60hz)?

For example, my monitor is 120hz (=8.3ms) vertical and 140khz horizontal. At 140khz horizontal with 1080 rows, that works out to be 130hz (=7.7ms) net horizontal, which is a little faster than the vertical rate which it has to be to fit within the vertical.

 

Maybe that's the case for some monitors, but it can't be the case of DK2 display for example. Otherwise low persistence wouldn't be possible because it would spend the whole time drawing the frame, so it wouldn't be able to be turned off after 3ms. This is why they say low persistence is only possible with OLED tech, because the pixels can change very quickly. It will still do a top/bottom scan without global update, but it will be much quicker

 

Low persistence is also possible to do with TN panels - ULMB, Lightboost, Blur Reduction Mode, Turbo240, etc. You can even adjust the persistence timing (or whatever it's called) with some monitors.

[HR][/HR]

As expected, I'm glad that Nvidia has just confirmed that they're doing what AMD announced with regards to SLI, rendering each eye separately, cutting down latency even more, etc. Nvidia are calling it "GameWorks VR".

Awesome

 

gosh, never thought I go back to a multi gpu setup, this will tempt me but only after user reviews
( hated multi gpu setups that I've had in the past )

if it works well I'll be after another Titan !

 

The only thing I didn't read with Nvidia that I did read with AMD is multi-GPUs reducing latency even more than a single GPU. Hmmm...

 

yes that's what I'm interested in, I'm quite sensitive to latency / lag ( one of the reasons I like the dk2 so much )

any further reductions through any hardware config & I'll be trying that route

I run all my graphic settings etc about at half what I could to help reduce PC work load etc where ever possible

am now getting a totally fantastic simulator experience, run at cap 205fps, super smooth 1:1 (feel) tracking etc

I'd always hoped sim racing would get to the stage it actually is at now immersion & simulation wise

 

Those monitors aren't actually turning off though. Not completely. The backlight is strobing on and off (but the LCD panel itself remains on). That backlight strobing is all we really need for the brain to get full motion resolution out of LCD. That's why they're so great at improving motion resolution.

Such a simple solution is problematic with VR. Because VR smearing comes from turning your head (even if the picture itself is clear). If you turn your head and the image is still displaying your previous location on screen, the brain turns into into a smeared mess.

OLED's have a response time of .01ms. For all practical purposes, they can draw their image instantaneously. They're also self-illuminating and don't need a backlight. So when they turn off in low persistence, the screen goes completely off.

Turning off completely is the key for VR. Because the screen is mostly black, and only active for a short period when the frame is actually drawn, you don't have smearing when you turn your head in VR. Because when you turn your head, unless it's when the frame has been updated, the screen is black. So you're not viewing a frame that is out of date (and no longer in the position your head has moved to).