Is there any noticeable difference in performance with reduced clock timings?
Lower pixel clock means lower pixels per second, which means lower resolution or lower refresh rate.
However, 4K60 using 594MHz HDMI timing vs using 522MHz CVT-RB2 timing will not show a difference in performance. HDMI draws pixels slightly faster but a higher fraction of the time is spent in the horizontal or vertical blanking.
With only 8.64Gbps of effective bandwidth, is that still enough to fit 4K@60Hz without DSC even with custom timings?
4K60 is 497.664 MHz without any blanking period. Having zero blanking period is not possible though. This just serves as an ideal minimum bandwidth number for comparing to actual usable timings such as this:
4K60 522.62 MHz using CVT-RB timing. I suppose you might be able to reduce the blanking further, but you won't gain much.
Lower bandwidth reduces the maximum bits per pixel allowed for a timing. With 8.64Gbps (HBR2 x2 or HBR x4), 4K60 can be done with 16bpp using CVT-RB2. To achieve 16bpp, you have these options:
- DSC which is 12bpp (though there might be a setting in macOS to change that).
- 8bpc 4:2:2 (16 bpp)
- 10bpc 4:2:0 (15 bpp)
- 8bpc 4:2:0 (12 bpp)
DisplayPort 1.2 does not ordinarily support 4:2:0 but maybe some GPUs allow it (I don't remember which if any).
DisplayPort 1.2 does not support DSC.
I'm not sure when macOS chooses 4:2:2.
4:2:2 reduces chroma (color) resolution by half horizontally.
4:2:0 reduces chroma (color) resolution by half horizontally and vertically.
In both cases, luminance (grayscale) resolution is unmodified.
The hack is a Lenovo 15iKB with Alpine Ridge JHL6240 TB3 connected to older Intel iGPU, hence no DP 1.4/HBR3 support. All dumps with Intel HD 630 came from the hack.
Maybe Alpine Ridge Thunderbolt changes 2201h to HBR2 only in peripheral Thunderbolt devices (the Thunderbolt dock case). A host Thunderbolt controller does not do tunnelling of DisplayPort from GPU to a DisplayPort device connected directly to the host. A host Thunderbolt controller does not normally have a Thunderbolt DisplayPort Out Adapter to do the tunnelling. An exception is a Thunderbolt 1 iMac that supports Thunderbolt Target Display Mode - it uses a Thunderbolt DisplayPort Out Adapter of the host controller when doing Thunderbolt Target Display Mode, since DisplayPort is being tunnelled from a different Mac.
But wait, you did use a peripheral Thunderbolt device with the hack and it did not change 2201h to HBR2. Maybe because it's a hack and the host Thunderbolt controller has different firmware from a Mac?
The Lenovo 40AY USB-C Dock only outputs 4K@120Hz RGB (bpc not determined, BetterDisplay says 10bpc) most likely with DP1.4/HBR3 x2 (maybe with DSC) on Apple Silicon. Sadly, I wasn't able to get the same display information from Allrez on M1 platform.
AllRez can get a list of timings and modes on Apple Silicon. Each timing includes a list of possible color modes but it doesn't show which color mode is currently being used. Each color mode may have a score which might determine the preferred color mode.
#502
The same dock outputs 4K@60Hz RGB 10bpc with DP1.2/HBR2 x2 on 2017 Intel Mac with Radeon Pro
Which Intel Mac and which Radeon Pro and which OS?
and outputs 4K@30Hz RGB 10bpc with DP1.2/HBR2 x2 on Lenovo 15iKB Hack.
Different display? Not HDMI timing? Or custom timing?
I noticed that Cable Matters DP VMM7100 also has this variable. I was never able to get VRR exposed in macOS Settings even after updating EDID.
The AllRez outputs that you've included so far did not include an EDID with VRR.
Sadly, I don't have Cable Matters USB-C VMM6100 around anymore to test but EDID was able to detect VRR advertised by TV when it was plugged in. However, macOS never showed VRR option in Settings.
My Cable Matters USB-C VMM6100 does not have MSA_TIMING_PAR_IGNORED. I don't get any VRR option connecting that to a Dr HDMI 8K with EDID set to "17 - 4K120-444 FRL-5 10b VRR HDR 7.1". I do get VRR with DisplayPort connected to Acer XV273K. Both connected to a Radeon Pro W5700.
