Demostrating Pimax 8k has real RGB subpixels with an 4k panel resolution. And explaining the source of its image problems


#22

If you take the total number of pixels within the block making up the letter “R” for each panel you roughly get the following:

5K+ = 18 x 10.6 = 190.8
8K = 14 x 18 = 252

252 - 190.8 = 61.2 / 190.8 = .32 x 100 = 32% higher number of total pixels in the block area making up the letter “R” on the 8K.


#23

Exactly, just wanted to post the link to the original comment:

From where is this picture.

@antiflash If you look at the 8K arrangement and take the dominant green and red and form the pixels around them (even if not knowing where blue exactly is, it does not matter), you can only count as many pixels as @chiliwili69 at IL2 Sturmovik forum did, which means, if you get horizontal count right (as he did), you end up with only 14 pixels in vertical res.

The subpixel layout of 8K looks like it could be approximated by this arrangement:

G     B/R   G
B/R   G     B/R
G     B/R   G

If you agree that one RGB pixel has to include at least one G and one R (which are easily identified in the pic), you will end up with either

G   | B/R | G
    |-----|
B/R | G   | B/R
----|     |----
G   | B/R | G

or

G     B/R | G
---------------
B/R | G     B/R
---------------
G     B/R | G

The original poster chose the former, but either way, you end up with only half of resolution, in one direction.


#24

:sob::sob:16 * 1.5 = 27 --------- 14 (is the actual count)

14 maybe it could be 14.5 if we count better, We would say that the panel has pixels horizontally but in vertical they are half of what they should be.

3160 * 0.66 (2 subpixel instead of 3 subpixels) * 80% screen usage = 1668 more or les = 3160/2 = 1580. The numbers do not lie for me it is a panel not full RGB.

27 * 0.666666 * 0.8 = 14.39999999 pixels height. :sob::sob::sob::sob::sob::sob::sob::sob::sob::sob::sob:

The screen Horizontal has the same usage but vertical usage 5k+ is better near 20% better usage.

Green
Blue/Red PIXELs how say @risa2000 Vertical pixels are compresed to 2/3.

It also joins that the pattern is in diamond format, taking into account that humans tend to follow horizonal patterns and pour in many of their constructions, furniture, buildings. It is a bad decision to represent reality and add too much SDE.


#25

Personally I think the images, although good, are still not defined enough that I for one can determine anything for remotely sure either way. At the scale, and with the HMD lenses, it is hard to avoid smear, chromatic abberations, and overexposure.

I recall somebody posted a picture on one subreddit or other - this may have been before the kickstarter even started, which showed the rasters of several headsets in razor-sharp extreme closeup; Amongst them purportedly the 8k panel, which I vaguely recall indeed exhibiting diagonals, but also indeed being full RGB, and this without any differences in subpixel sizes/shapes for the primary colours.

I have tried to find that picture again, ever since, but have consistently come up short.

EDIT: Scratch that last – it was probably the one picture I found just now by trawling through my reddit history, and it was not a sharp image at all, and only showed the 8k(v2) and HTC Vive, to boot with a uniform cyan display, that doesn’t light up the red subpixel at all. Aah, memory, and where it leads one sometimes… :7

( https://www.reddit.com/r/virtualreality/comments/79buro/sde_comparison_pimax_8k_v2_vs_htc_vive_according/
)

( EDIT2: https://imgur.com/a/NCj5N )


#26

I made the same calculation as you twice. One time optimistic, one time pessimistic in terms of how big the area of the “R” letter actually is. With the optimistic apporach, 8K panel has about ~40% more pixel and subpixel than the 5K panel. Together with the 9% increased ppi of the 5K panel, this would fit the theory that this is indeed a “true” 4K panel.
With the pessimistic approach the 8K panel would have about the same amount of pixel and subpixel, which fits the theory of 1.5 subpixel per pixel and therefore a “fake” 4K panel.

I think we need better photos with a more standardized setup to use such an approach.

What bogs me is what Machdisk said. If every pixel would be rectangular, than the screens would not have a physical 16:9 aspect ratio, but rather 32:9 or 16:18, which I don’t think is the case.


#27

but 4k panel must be much more 50% more pixels.

2560+1440= 3.686.400
3840*2160=8.294.400

8294400/3686400 = 2.25 8k must be 225% more pixeles than 5k+

Better usage of 5k+ panel can be maybe 8k has 200% more pixels than 5k+, but only 40% with diagonal patron its a joke.


#28

Pixel are diamond shaped but the result is the same if they were rectangular. Different quantities of rows and colums than standard squared pixels 4k) but 8.3M Pixels in total at the end. It could change the aspect ratio of you distribute the pixels one to one. But instead it resamples the image to the same aspect ratio causing the lost in clarity.


#29

Is there any advantage to that diagonal grid?


#30

Yeah, you are correct. I got the +50% from the fact that 8K would need to have 50% more pixel in x and y direction. However on areas +125% would be necessary. So this makes it quite clear to me that is indeed a “fake” 4K panel.


#31

Im not as knowledgeable as others here so i might well be wrong & im ready to eat my words, but however I define the pixels whether horizontal or vertical, as others have said, i only come up with around half the est pixels in either horizontal or diagonal.
this looks like some kind of trick marketing where you ask us to count red/blue as a 1 pixel & green as another pixel but borrow what’s missing from a neighbouring pixel without counting it as part of the pixel & hope the eyes are tricked into believing it’s true 4k because they are alternated From row to row?


#32

Correct.
I lost track of all the “math” in this tread. 1.5 the linear factor. comparing square pixels with square pixels: 3840/2560 and 2160/1440.
But once again the 8k panels have “diamond” pixels. The True “4k” resolution of the 8k is 2715x3055 (Not 3840x2160) That compared to the 5k+ 2560x1440 it means that it should have different factors per dimension.
H: 2715/2560 ~ 1
V: 3055/1440 ~ 2

And accounting for variation in panel utilization and focal distance that is approximately what we get in the famous “R” picture.


#33

You keep saying diamond shaped, but I do not see a diamond shape and you still have not shown a picture that shows how you think subpixel are arranged in a diamond shaped pixel on the 8K panel.

What I see are either rectangular or square pixels (depending on how one pixel is defined), where every second row (or column) is shifted.


#34

That was a hypothetical. What we have is a RB/G 4k instead of an RGB panel. The difference is that an RGB is a “true” 4k display while a RB/G basically uses a panel of half the vertical resolution to pretend to be one.

2x4 grid of pixels on RGB display:

RGB, RGB, RGB, RGB
RGB, RGB, RGB, RGB

2x4 grid of pixels on a RB/G display:

RB, G, RB, G
G, RB, G, RB

Now here is the fun one, 1 RGB line of half the resolution
RBG, RBG, RBG, RBG.

See how 2 lines of RB/G is the same subpixels as 1 line of half the resolution?

The only difference between the 8k “3840x2160” screen and a “3840x1080” screen is that you have split the subpixels into twice as many lines and then driven them like they were different entire pixels.

Now you can either do that directly and that would mean you might miss entire pixels worth of red or blue on a “green sub pixel” so what the driver will do is resample some of the colour information to adjacent pixels causing the blur effect we all know and love. Alternatively you can drive them directly as individual pixels but at best you get the clarity of a 3840x1080 image in any one colour but with gaps where the other colours lie which would lead to worse perceivable SDE on scenes with no green or no red and blue. However this choice will be baked into the panel interface so I doubt there is anything Pimax can do about it. Because of this I doubt the 8kx will look much different at all.

Basically it’s a really dodgy trick to claim twice the vertical resolution with the same number of pixels. Don’t blame pimax entirely though because I have no doubt the device datasheet did not lay out that this is what they were doing. It would have absolutely claimed to be a 4k screen, might even have boasted about not being pentile and any reference to this compromise would have been deeply hidden. Because that’s what LCD datasheets are like for some reason (Given me a headache or three in the past).

Edit for clarity. It’s not all bad. For white on black or vice Versa you do see 4ks worth of lit pixels. However you are actually seeing two slightly offset overlapping 3840*1080 images in different colours hence the slight colour fringe effect. Also the exact pixel resampling technique used puts a hard limit on clarity.

Effectively you get a screen that sometimes looks like a blurred 4k one and sometimes looks like a 3840*1080 one and which you see depends on the exact colours present in the image. Much of the time it will veer between the two. On average with a native source and an optimum resampling technique it “might” achieve an average clarity roughly equivalent to the 5k but I really don’t think it ever will when upscaled.


MRTV Rant: Pimax Communcation Is Terrible Regarding 8K / 5k+
MRTV Rant: Pimax Communcation Is Terrible Regarding 8K / 5k+
#35

Actually, it was taken from this orignal video, where it also compares it to Pimax 4K panel, and the difference is evident (jump to 10m7s if it does not automatically):


8K, or 5K+, that is the question - Discussion
#36

That the devices are different is indeed very evident, but contrary to how I remembered it (EDIT2: …assuming this was indeed the picture I remembered :P), the 8k picture is unfortunately not clear enough that one can pick out the subpixels - just brighter cyan-ish dots, in a cross-hatch layout, on a darker cyan-ish background. :7

Tried giving it some saturation and contrast boost, but that didn’t help. :7

(EDIT: Also worth keeping in mind that we don’t know the respective zoom levels, in the different pictures, so we can’t exactly compare dot pitch… )


#37

What you say is exactly how I think it is. No diamond bullshit, no non-square pixel. Instead on average 1.5 subpixel per pixel.
I don’t know if Pimax was aware of this, and actually I don’t really care. However, I think this panel is a waste on computational resources and therefore totally unfit for virtual reality. I also think that the 8K X will not look much better than the 8K and maybe match the 5K at much higher computational cost.

I really do hope that Pimax will not use those panels for the 8K X.


MRTV Rant: Pimax Communcation Is Terrible Regarding 8K / 5k+
#38

Oh wow, yeah you can see it clearly there. That is a blue background so you would expect the green sub pixels to be off leaving gaps between lit blue pixels on the 8k and sure enough on the zoomed in bit:

So same as we have now. The 8k shows half the resolution of the 4k on a blue/red or green screen. I bet you anything if you changed that scene to green the pixel pattern/SDE would all move one pixel down in the 8k one but look identical in the 4k.


#39

Yeah me too mate, even pentile oled 4k is better than this if not by much. Still at least we know now (barring new evidence that upends the theory but it looks pretty clear to me).

8k has a effective colour resolution of 3840x1080 for scenes that are solid blue/red or green. In more mixed scenes your SDE effect will seem to improve to some extent as the subpixels will light up giving you SDE similar to a true RGB 4k screen but it will probably always have some blur due to pixel resampling and lower effective vertical resolution than 5k+. The upscaling also contributes blur but this can be very nearly eliminated by upscaling while the screen based issues are probably pretty much baked in. So 8kx wont improve much over the 8k (as pimax did warn us in fairness though I must admit I thought that was hyperbole as I couldn’t understand how that could be the case… doh).

In summary, I’ll get the 5k+. Probably keep it and sell my 8kx unless pimax improve that before it ships.

Having thought on it some more, white on black and black on white will be the best possible use cases and might see some sharpness benefits over 5k+ on 8kx depending on what exactly happens regarding pixel resampling. Anything in red/blue/green however, no chance.


#40

I am suspect that how much advantage if pimax have a 4k native 60Hz mode for 8k, ever have someone asked for this and they tell that they will check the possibility.


#41

Won’t make any difference to the above issue. You can drive it at native 4k and it would still look similar.