LG, Google Designed An 18-Megapixel, 120Hz Display For VR Headsets

Status
Not open for further replies.

araczynski

Distinguished
Jun 16, 2008
45
0
18,530
theoretically this should pretty much eliminate the screendoor and be as good as looking at a high def monitor at 2 feet away, nice.
 

Giroro

Splendid
"LG and Google used a two subpixel per pixel arrangement,"

No wonder they have to push it to impossible-to-render resolutions just to mask the horrid "chain link fence" effect that are inherent to penile displays. They may be reaching 1443 "PPI", but I bet it looks noticeably worse than a 962 PPI (2/3 the resolution) display that uses real, square, RGB pixels.
 

dabeargrowls

Honorable
Dec 28, 2012
29
0
10,530


I agree! you can't drive the current VR sets in the highest performance mode without issues. What makes anyone think the current video cards can drive this? maybe 5 years from now but its doubtful with Nvidia's release rate for cards. Remember the 980TI and 1080TI is maybe 20% difference. 1180...maybe another 20% and can maybe drive the Vive Pro? I keep hearing everyone so excited about pimax. It will work, sure....with 3 - 1080TI maybe? if they can make 90hz (stuck at 80).

 

lorfa

Distinguished
Mar 30, 2012
139
14
18,685
That 1443 PPI is actually pretty impressive. A 19" 5:4 display could have a resolution of 21410x17128. A 24" 16:9 display could have a resolution of 30182x16977.
 

The foveated rendering with eye-tracking is potentially a complete and total game changer if it's well implemented. That's a massive *if*, of course, but if ultra high res VR takes off in the medium term, it'll be on the back of foveated rendering and eyetracking.

Also you're massively underselling the 1080ti. A 1070 edges out the old 980ti. Once you take away the CPU bottleneck with 4K extreme settings the 1080ti is generally 60-80% faster than the 980ti (check out the Tom's Hardware launch review, the data is there).I'm not for a second suggesting that the next-gen cards will be able to run this display, but we are seeing steady progress and a well implemented FR + eye tracking system could be a complete game changer.
 
The fastest video cards in the world cant even come close to running this.
Maybe by the time we get to the ?1980 Ti? In like 20 years or something.
As long as you have at least 4 of them at just under $5,000 total... *UGH*
Price and graphical power are BOTH issues this VR system will face before it can be a reality (no pun intended).

However, this IS very forward thinking as Virtual Reality needs a huge boost in the visual department to create "Photo Realistic" graphics. This level of realism though would make VR far more amazing.
 

bit_user

Polypheme
Ambassador

I'm gonna say the DGX-2 might get you in the ballpark. Especially if you use one per eye.

https://www.nvidia.com/en-us/data-center/dgx-2/

The hardest part might actually be the connectivity to the HMD.

But the real answer is going to be foveated rendering. Hardware ray-tracing would make almost perfect FR almost trivial, assuming the eye-tracking is good enough.
 
Completely agree that foveated rendering will go a long way toward making the workload more realistic for the rendering devices.

Pascal behaves pretty much the same way that Maxwell 2 does. When you have parity of clock speeds and calculate for the difference in shader cores / render back ends, you pretty much get parity of performance. NVIDIA didn't give us more instructions per cycle with the Pascal release, or even a significant amount of increased rendering resources. They upped the clock speed, by a lot, which allows cards such as the 1070 to slightly outpace the 980 Ti part despite the fewer resources.
 

As the article mentioned (and Rhysiam already pointed out), the performance demands won't be anywhere near that high, since they are utilizing foveated rendering, which involves rendering the scene at a lower resolution, then using eye tracking to detect where the user is looking, and rendering that small portion of the screen at a higher resolution. The article actually points out that most of the scene is only being rendered at 1280x1922 pixels (I'm assuming per-eye), which might be comparable to a Vive Pro, while only a small area is getting rendered at the panel's full resolution. By doing that, they likely already have this running games on today's higher-end graphics hardware, at least in some capacity. I do suspect it may be some years before these panels actually appear in a consumer headset though.
 

Kelavarus

Distinguished
Sep 7, 2009
510
0
18,980
Everyone in the comments also seems to be operating under the assumption that every game has to take a ton of graphical horsepower to render. Some of the best VR experiences I've had are ones that have a simplistic art style with a focus on gameplay (SUPERHOT, House of the Dying Sun, the Vive demos, etc...). Personally I don't really even have that much of an issue with the screendoor effect on my Oculus, my biggest wish is better horizontal FOV. When I'm playing and actually enjoying what I'm doing, it's only the FOV that continuously bothers me.
 

maetrixss

Prominent
Sep 15, 2017
4
0
510
"but the team suggested that they would be ideal for standalone VR devices"

Yes, because the quickest path to the kind of graphics horsepower required to push all of those pixels is through mobile hardware and not a full desktop gaming PC.
 

Nintendork

Distinguished
Dec 22, 2008
464
0
18,780
There's cheating boys, it's a damn pentile display, same as the Samsung's 2560x1440 AMOLED that are no better than 1600x900.

For Samsung the only true RGB AMOLED is the Galaxy Tab S3 2048x1536 @9.7"
 

It may be a pentile display, just like the existing Rift and Vive headsets, but at this kind of resolution, it shouldn't really matter much. We're talking about having over 14 times the subpixel elements of these first-generation consumer headsets. A 4800x3840 pentile display still has almost 50% more subpixel elements than a 4K RGB display, and the human eye is better at perceiving detail in the green channel, which is fully represented.

Also, claiming that a 2560x1440 AMOLED display is no better than 1600x900 RGB is not accurate. 2560x1440x2 works out to 7372800 subpixel elements, while 1600x900x3 is only 4320000 elements. Even a 1920x1200 RGB display only has 6912000 subpixel elements. Now, certainly a pentile display won't provide as much detail as an RGB one of the same resolution, but the difference isn't that large. If you're at the limits of how many pixels can be packed within a certain area by a given manufacturing process, dedicating more of those pixels to the green channel can arguably be considered a reasonable tradeoff.
 

vmylo81

Reputable
May 23, 2018
10
0
4,510
The future will have some interesting stuff. Who knows how the world will be in 20 years from now. Everything goes so fast.
 
Status
Not open for further replies.