Wouldn't foveated rendering leave the Pimax 8k x obsolete?


#41

For reference, the difference between the Vive and Vive Pro is 1.33 (plus a few hundredths due to slightly reduced fov (physically smaller display panels)). :7

I am still waiting to hear whether the simple scaler in the 8k will at the very least use content stuffed into the entire transport image, or if 20% will go to waste. Like this:

Don’t know what has changed since, but according to the ye olde Kickstarter page FAQ, the grey areas with the arrows are not visible through the lenses. As long as the scaler can use a different ratio per axis, that should not be a problem, and the transport image will be fully utilised all the way to the edges (a per the arrows), and then compressed horizontally by the scaler; But otherwise, if those black bars exist in the transport image, that’s pure bandwidth waste.

Then of course, with just the slightest increase in scaler complexity, and without even considering foveation, it should have been possible to do something like:

…where each darker band could be scaled to half the width (…and height for the top and bottom ones), just as a function of the lenses and anisotropy, without any discombobulating the image, stepping away from fixed discrete nubers, nor breaking rectangularity in any way – do any of those, though, and there’s more to save.

(EDIT: Major caveat: Those scaled down region boundaries would of course be subject to the pincushion distortion of the lenses, though.)


#42

I think it is, why do you not? If you are at the forefront then you have to do it first :slight_smile: The solution I would like to see is that Pimax 8K-X should have both dual DP and a Pimax implemented solution of F.R with a C++ SDK that can be ported to any engine. Tobil could drop a solution straight in right now.

Oculus have done a Fixed F.R. in their Go which is available already.
Even a fixed F.R can save up to 25% computation cost! Or lets put it another way, if this was on the Pimax 8K-X then that could allow the 1080Ti to run perfectly fine without waiting for the next Nvidia model to come out. It might even allow 1 x DP because of reduced bandwidth :wink:

Ref: https://developer.oculus.com/blog/optimizing-oculus-go-for-performance/

Tobil have a Vive dev kit which includes everything to try it out.

It just takes one known HMD manufacturer to implement it fully and Oculus are one step ahead already. Pimax could find themselves at the forefront of VR headsets when they launch the 8K so why stop the momentum? They need all possible solutions for the 8K-X and the recent NVidia announcement of a delayed GTX launch will not sit well with everybody.

Oculus determine a 4 to 5 x speed improvement using their own eye tracked FR system. Imagine running the Pimax 8K-X on a 3GB GTX 1060 Ti :sunny: (1080Ti being around 150% faster than the 1060 Ti 6GB)

I guess we will know more in September at Connect 5

Here is a connect 4 video from last year that talks about what they planned for Eye tracked F.R so I hope to see this mentioned at Connect 5.

I say take a leap :slight_smile: future proof yourself?


#43

[quote=“D3Pixel, post:42, topic:6297, full:true”]

I think it is, why do you not? If you are at the forefront then you have to do it first :slight_smile: The solution I would like to see is that Pimax 8K-X should have both dual DP and a Pimax implemented solution of F.R with a C++ SDK that can be ported to any engine. Tobil could drop a solution straight in right now.

Oculus have done a Fixed F.R. in their Go which is available already.

It just takes one known HMD manufacturer to implement it fully and Oculus are one step ahead already. [/quote]
The reason I think it is not ready is due, in large part, because I am a software engineer with experience in the graphics and gaming industries. (I’ve written games and drivers.) Just because someone has an SDK/API to do something doesn’t mean that it will actually get used. This technology is way more complicated than you seem to think and the first implementations will likely have issues.

VR is still a tiny market and Pimax is a tiny part of that. The reality is that most games will not use FR in the next few years. Sure, some will, but for example, Elite Dangerous probably won’t. They can’t even fix their HUD color, which players have been begging for, literally for 4 years now, and that affects ALL their customers, not just 0.01%.

Fixed FR is MUCH easier to do than full FR. The downside is that you can see the lower res areas, if you look in a direction without turning your head. The reason that people pledged for the 8KX is that they WANT to be able to read gauges and text that aren’t in the center of the screen. Fixed FR won’t allow that.

Imo, FR won’t gain wide support until the VR market matures, which is still years away.


#44

Fair enough. We have a different of opinion on it which is fine :slight_smile:

I too have developed software over the last 20 years, mostly in objective-c at a commercial level (including a year doing low level audio in C which was NOT fun lol) so I do have some reference to base my musings on. I now prefer animation and the technical aspects like scripting and plugins to pay the bills but I bounce back and forth between the two industries depending on the available contracts.

I haven’t done any driver development though. What did you do?

Neither will the 8K-X unless somebody solves the existing technical barriers. It needs innovation. Which is kinda what my point is :smiley:


#45

I’ve written drivers for: mice, printers, scanners, and worked on video drivers. Long ago, I was part of a team doing video poker consoles (for gambling) and I’ve written 2 video games:



#46

Ha, rather you than me. I dabbled in recompiling a printer driver once when it had an issue, what was it called? postscript / PCL or something, can’t remember the correct terms now, was in the 80’s and i’m late 40’s now so a long time ago.

Love the cover for Icon. Reminds of Dragons Lair (Laser disc version) that used to be in the arcades.


#47

Looking at the entries of your games, I sort of even seem to remember the Ra game in terms of having read a review at the time - though not having cough organized cough a copy of it… and Sir-Tech even rang a louder bell - good ol‘ Jagged Alliance - man, I loved that one.
off-topic end


#48

We did run a few adds and even got reviews in PC Magazine “After Hours” column.

In particular, 7 Spirits was “shared” quite a bit and used for compatibility testing by video card manufacturers for quite a few years (according to rumors).


#49

are we certain the scaler can only take 1440p and upscale from that… Would be a pretty limited scaler. I would imagine. (maybe get some clarity from pimax) that it can take lower res and upscale to 3840x2160 as well. so you could run 1080 if you want to allow your lower end gpu to use the 8k headset


#50

The scaler chip is reportedly quite limited. There was some discussion about the possibility of 1920x1080 in addition to 2560x1440, but I have seen no confirmation of that from Pimax.

There is also a faint hope (in the forum) that it might be possible to bypass the scaler chip and output 3840x2160 at a lower framerate, using Brain Warp.


#51

Why would the frame rate be lower?


#52

Because 3840x2160 is more data than 2560x1440 and 1 cable cannot carry the higher res at the same framerate.

Consider that it’s the res of a 8KX, which requires 2 video cables to transmit the data at 90 Hz. With 1 cable you’d get 45 Hz (or maybe 90 Hz with Brain warp). Of course the 8KX could theoretically get 180 Hz with Brain warp, so the framerate would still be lower on the 8K.


#53

I think you’re mistaken. One display port can easily handle refresh rates of over 90hz at 4k as long as it’s only on one screen. The reason the 8k x requires 2 display ports is because it uses two screens and two bridge chips. Fps should max out with the refresh rate at 80. Maybe even 90 if we can bypass the hardware scaler. :slight_smile:


#54

I’m only concern about practical usage. The way Oculus is planning using Foveated rendering and how it should be used in current Vive or Pimax8K is bit different. My point is without support for Foveated rendering in games it is pointless. Also, we still don’t have a final solution for eye tracking on market in general so all this point that in practical usage FR will not be ready minimum in next 3 years. Till then we will get newer and better solutions for displays with much better PPI and PPD than today’s 4K displays. So in that 3 years or more Pimax8K will not be obsolete for next 5 years minimum.


#55

This new VR industry is moving on a large curve so predictions of product life cycles are nay on impossible as nothing has matured and lots of innovation is still needed. VR/AR are the next evolution of the humble 2D display (mobile, TV, monitors etc) so things will change constantly as battles to dominate the future unfold.

Personally I give anything VR/AR brought out now, about 2 to 3 years shelf life before it will be vastly out performed in either usability and functionality or as Pimax have gone for, full blown quality. Pimax still have the 8K-X to protect them and then whatever they bring out after that.

Saying that, 0.6% of people on steam still use a DK2 :smiley:


#56

Wait a minute. I just realized you’re wrong actually. The computer simply needs to pull the data from the eye tracker to figure out where you’re looking at. Then the computer would draw the foveated frame. This works because the eye tracker only has 5ms of latency. So in theory, the 8k should be able to do native 4k. It’s not as if we need to know what you’re actually looking at. Just where you’re looking.


#57

Bridge chip. No it can’t do native 4k. 8k-x yes.


#58

No hardware scaler in current design of 8k-X if it uses dual bridge chip (of bridge chip in 5k/8k).

Pimax has already stated cannot bypass scaler. The same is true 2x25601440 or 25601440/display in 8k. Yes a single dp can supply 4k, but not dual 4k.

Now if 8k-X is delayed to q2 2019 pimax might be able to use analogics new bridge chip for nicer bling like built in hardware scaler


#59

Even if the scaler wasn’t bypassed, bandwidth shouldn’t be an issue. This is because the resolution surrounding the center would be so much of a lower resolution, that it would end up being the same amount of, if not lower bandwidth. It should be able to go through the upscaler without running into any bandwidth issues.

The bridge chip wouldn’t run into bandwidth limitations either for the same reasons. It’s not as if we’re sending a full 4k image. We’re sending something closer to a few degrees of 4k image. It would work like this.

The computer takes data from the eye tracker to figure out where you’re looking at. Then the computer draws the foveated frame. The frame is then sent through a software upscaler that upscales the lower resolution area outside the native resolution center to fit the screen. The frame is then sent to the headset, with a much lower bandwidth than native 4k. It then goes through the hardware scaler (I don’t know what this would do, if anything at all) , then it would be sent through the bridge chip to the displays.


#60

Maybe I’m wrong though.