madVR 10-bit Display Support Test

[copenhagenstreaming]

you don't need FSE anymore more on windows 10.
it'S not an option in mpc-hc.

Thank's huhn, could you help us troubleshoot why 10 bit doesn't seem to be working? At least 8 bit and 10 bit looks exactly the same.

[huhn]

that'a to be expected.

[Asmodian]

Haha, people always expect 10 bit to be significantly better but almost everyone needs to use very specific test patterns or bad settings to be able to notice a difference even while pixel peeping with their nose up against the screen. ;)

10 bit output from madVR is not important unless your GPU or screen behave badly with 8 bit HDR.

[nsnhd]

I just tested my 4k 32" monitor to see if I can see a difference with 10-bit via displayport output. Beside madVR I also watch videos in browsers.
One thing I noticed is that though I'm able to set 10-bit 4k@60Hz RGB Full in Nvidia CP and Windows reports 10-bit, but in madVR settings, devices, identification, bitdepths has 8-bit value. I'm not sure what that means, since when connected via HDMI it has 8, 10 and 12-bit values. Also I almost see no differences by playing test patterns in the first page.

[copenhagenstreaming]

hat'a to be expected.

... everyone needs to use very specific test patterns ...

I am using the test pattern from the first post where 8 bit should produce clear and obvious banding - and it does.

10 bit shows exactly the same banding, but as I understand the colours should blend much more smoothly in this mode.

Has things changed since #0 since . you say that it is expected to see no difference between 8 and 10 bit?

[nsnhd]

Finally, with the Test Patterns from this AVS thread, I can see clearly the difference of 10-bit to 8-bit on my 4k 32" 10-bit monitor. Played with madVR in PotPlayer and toggled 8/10-bit in NCP.
https://www.avsforum.com/forum/139-display-calibration/2269338-10-bit-gradient-test-patterns.html

[huhn]

this test file is flawed even the 10 bit has banding.
the file is not dithered from let's say 16 bit it is rounded to 8 bit on the bot and 10 bit on the top so both have banding.
beware that nvidia has an very very old bug in the driver where 10 bit WSF with nvidia set to 8 bit will produce a lot of banding.

Has things changed since #0 since . you say that it is expected to see no difference between 8 and 10 bit?
because 8 bit doesn't have banding what so ever you have to look for noise.

[mark0077]

Do you guys notice when using the gradient in the first post, when switching display modes in the NVidia Control Panel, madVR will get stuck in "D3D11 windowed (8 bit)" mode, until you reopen the player / redrag the png into the mpc window? Made testing very confusing until I noticed that.

[huhn]

the newest test build is for me ignoring the bit deep setting to avoid bad setup like 10 bit madVR and 8 bit nvidia.

[VBB]

Testing for 8-bit vs. 10-bit has always been flawed, but testing for banding can be done easily. The files from the AVS thread mentioned by nsnhd are perfect for that.

[nsnhd]

Less banding for 10-bit but now madVR can't auto switch refresh rate any more. It's fixed by 60Hz and lots of repeated frames, rendering times get higher. I'm not sure if that comes from DP instead of HDMI connection.

LE: I have fixed the stuck 60Hz issue by creating 23, 50, 59Hz custom mode. Now I have a full 10-bit path madVR-GPU-Display with less banding.

[huhn]

well looks like you found a serious issue with your setup then because 8 or 10 bit has nothing todo with banding.