Why 8k is more Gpu demanding than 5k?

This is confusing for me, too. Does Pimax “compress” more information into the image for the 8k that the upscaler can “uncompress” again? Otherwise I see no correlation between supersampling or not and upscaling or not. Intuitively supersampling should improve image quality on both, upscaling (being a hardware feature) is obviously only available for the 8k.
Or does Pimax want to convince shop visitors that 8k is a higher fidelity product when A/B testing with 5k by hardcoding the supersampling at a lower or respectively higher level? That would be a very odd move…
Will be interesting how this turns out…

3 Likes

From the nature of VR rendering there is always some supersampling necessary in order to not lose precision, so I would not expect that the image for 5K+ is rendered at its native resolution. I know @deletedpimaxrep1 wrote it somewhere, but I do not believe it. Let’s wait for what testers will report.

By my reckoning its all to do with super sampling for instance ,the 5k+ no matter if you sample @ 5k or 8k you can only render @5k because thats all the pixals you have although an 8k sample may fill in a few jagged edges a 5k sample could not pick up
Now throw the 5K sample at the 8k screen and theirs alot of blank spaces to be filled hence the upscaler but the catch is which i think was common knowledge from the start that you would not get an 8k image ?

heres my working out its different to what i come up with earlier, but ive forgotten how i come to that figure but i think it amounts to the same

5k carrier signal sampling 8k @ 90fps is 37.5% greater now with a 11.1% drop due to 80fps gives us a 26.4% increase due to sampling @80hz ,so a base 5k signal sampling 8k image can only convert into 6.320k image but with more precision than the 5k due to the 8k screen

so sampling with the 8k is expensive gpu wise due to the down scaling as lots of losses

i do hope some one will correct me if this is wrong with the idiots guide to vr as im just trying to figure out the max potentials of it all given that i think 8k sampling is going to be varied with performance

1 Like

Does the Pimax Bridge chip use latest transport compression?

e.g. VESA’s Display Stream Compression (DSC) standard which I think is up to version 1.2a from CES 2018.
https://vesa.org/featured-articles/rise-of-beyond-4k-resolution-and-hdr-displays-growth-of-arvr-driving-adoption-of-
displayport-and-related-vesa-standards/

1 Like

there is no mention of DSC in the ANX7530 producr brief pdf
only mentioned for ANX7533 and ANX7580
also i think the “missing” DSC was mentioned in the discussion about ANX7530 last year
beside DSC not present, even if available it would have to proof to be usable for vr content, there might be quality issues depending on the content and also any millisecond counts for vr, any latency might be critical, there where even discussions about >2ms for wireless vr

1 Like

btw. the fact that 8k renders higher was already documented by pimax on 1st of august here
http://community.openmr.ai/t/updated0814-pimax-8k-progress-update/6021/12

“…
Please explain why the 5K would have better performance than the 8K with GTX1070?
A: rendering at 6K or 8K, then downscale to 5K, and finally 5K upscale to 8K.
B: rendering at 5K, then 5K upscale to 8K.
The visual quality with the rendering path A looks much better than B.
that’s why the supersampling works well & with 1070 graphics cards, 5K performs better than 8K.
…”

so its no question about “if”, the only question might be how bad it looks with 5k render on a 8k headset, i guess it will be possible to spot and if that looks noticeably bad compared to the 5k headset then the 8k headset makes no sense with that setting
maybe that was already covert from the testers when testing the M1 so we might hear about that after the 16th of september

There is no mention of HBR3 either. Note that it supports “HBR2.5”, which no standard defines. There is only HBR2 (DP1.2) and HBR3 (DP1.3). ANX7530 is “DP1.4 Ready” the same way 1280x720 TV was “HD ready”. If you search this forum, this chip has been already discussed many times. Then there are other limitations on MIPI interface (also discussed already here). It may work with DP1.4 but not the way you would expect.

I think it does, but current nVidia cards don’t support that type of compression. I have no idea as to whether or not RTX 2080 supports it.

No, it does not. (20 chars)

1 Like

OK, thanks for the clarification.

1 Like

Does the upscaler also supersample again before sending to the display? And should it?

1 Like

No and no. (20 chars)

1 Like