Esoteric Graphics Questions/Findings (another T221 thread)

[joevt]

I appreciate this, though something seems to have stopped working. I can verify Lilu and WhateverGreen both appear in the MBP's IOResources section in IORegistryExplorer and '-iofbon' appears in 'boot-args' when running 'nvram -p' but the MBP screen blacks out and the external display doesn't come on whenever the TH2G is plugged in. All I can seem to do to resolve this is unplug the display and hard reboot without the display plugged in to get the internal display back. Same thing happens if starting up with the TH2G already connected.

Some other boot-args:
-liludbgall output more stuff to log stream.
liludump=120 creates a log file in /var/log after two minutes from startup.
-v show verbose log during startup.
keepsyms=1 show functions in kernel panics.

If you had Lilu and WhateverGreen installed previously, then you'll need to rebuild the kext cache after installing my Lilu/WhateverGreen to make sure the correct version is loaded. You should be able to tell what version is loaded by looking at the /var/log/Lilu*.txt file.

I use my kextutil.sh script to load or install kexts.

source kextutil.sh
installkext Lilu.kext
installkext WhateverGreen.kext
rebuildkextcache

I'm pretty far out of my depth here but I did find the guide for installing OpenCore on the MacPro5,1 so I'm going to try following that to hopefully have a USB drive act as the OC EFI partition in case something goes wrong.

Ok, but since you're not running any unsupported versions of macOS, you don't have to add any special stuff like spoofing the MacPro5,1 to be a MacPro7,1 etc. and you don't need any of the extra kexts.

If you use OpenCore to load the kexts, then they should be removed from /S/L/E or /L/E and the kext cache should be rebuilt.

 

[jeremywork]

I tried to match IBM models to "clones" based on what EDIDs and configurations they can handle.

I'm compiling data from all the documentation now and this has lead to some additional questions 🤪

Iiyama's is the only manual which specifies polarity values at all, but it specifies both Positive and Negative for each of the H and V values. Taking the first from each pair lines up with the SRX dump for the 3840 x 2400 modes from 029 and 030, but neither of those two EDIDs list 960 x 2400 at all, and none of the EDIDs I was able to dump shows H+ and V+ for the 1920 x 2400 resolution.

960 x 2400 only shows up in 021 and 022, both with H+ and V+. I kind of wonder why EDID flash procedures are mentioned at all if the only valid choice is 029. It would seem Iiyama's 029 isn't IBM/IDTech's 029, and at minimum one additional mode must be available for there to be seven listed non-standard display modes (plus, the 1920 x 1200 standalone mode is listed with different timings than the 4x tile mode.)

IBM's and IDTech's manuals seem perfectly congruent on modes and timings (though IDTech's has the Stripe VGA mode listed as compatible with MD22282 Model B1, a typo I assume should read MD22292 Model B1)

ViewSonic's 'Supported display modes' page basically mirrors Iiyama's without polarity data but there are a few discrepancies between these and the IBM/IDTech 'Supported display modes' page:

640 x 400 is listed at 70 Hz on IBM/IDTech but 59.9 Hz on ViewSonic/Iiyama, with the same kHz and MHz totals as used by IBM/IDTech for 640 x 480 @ 59.9 Hz. Presumably this is a typo and not offset by blanking?

IBM/IDTech list three timings for UXGA-Wide: 40.9 Hz, 60.0 Hz, and 48.0 Hz. ViewSonic and Iiyama only list 40.9 Hz.

ViewSonic/Iiyama have QUXGA-Wide modes on the 'Supported display modes' page but IBM/IDTech only list these modes on the 'Extended modes' page. Iiyama doesn't have an 'Extended modes' page but ViewSonic does.

ViewSonic lists only QUXGA-Wide (1920 x 1200) x 4 at 40.9 Hz on the 'Supported display modes' page but both at 40.9 Hz and 48 Hz on the 'Extended modes' page.

(1920 x 1200) x 4 @ 40.9 Hz is listed twice with different total timings, but one of them isn't actually listed as supported by a single ViewSonic EDID mode (line 25.)

IBM and IDTech also have two listed timings for this mode (found on EDID modes 021, 029, and 044 on my page.) Modes 021 and 029 should have the same 1920 x 1200 mode according to the manual, but actually mode 021 matches 044 in my dump.

Despite this, quite a lot of new data has made it in some usable state onto my page. Many thanks!

Thanks @joevt I'm trying the additional boot-args now and will look at the Lilu logfile. For some reason I get an error whenever I try to directly quote any of your last post, but I hadn't installed either Lilu or WhateverGreen to this laptop before installing yours. I'll try your kextutil.sh next time, or get OpenCore installed (appreciate your notes!) I had been installing using the instructions here: https://www.tonymacx86.com/threads/...an-sierra-high-sierra-mojave-catalina.268964/

For All MacOS versions :

Using Finder, create a folder on your desktop, copy all the kexts you want to install into it.

1. Open Terminal and change the active path to the folder you created in step 1 using the "CD" command.
2. Install the Kexts into /Library/Extensions using the following command :-

Code:

sudo cp -R *.kext /Library/Extensions




4. To repair permissions and rebuild the caches enter the following commands one at a time :-



Code:

sudo chown -v -R root:wheel /System/Library/Extensions
sudo touch /System/Library/Extensions
sudo chmod -v -R 755 /Library/Extensions
sudo chown -v -R root:wheel /Library/Extensions
sudo touch /Library/Extensions
sudo kextcache -i /



The above command set is a belt and braces approach (mimics what Hackintool and Kext Utility does) and should make sure that there are no problems with any file persimmons on the system kexts (in /S/L/E) and the 3rd party kexts (in /L/E).

Once finished reboot your system.

 

[jeremywork]

Deep into this page there is a reference to the ADTX models which perhaps explains the origin of EDID mode 030, since it's the only one in my dumps with different gamma and white balance.

【3840x2400】QUXGA-Wのモニタpart5【200ppi】

 

[jeremywork]

Deep into this page there is a reference to the ADTX models which perhaps explains the origin of EDID mode 030, since it's the only one in my dumps with different gamma and white balance.

This finding is corroborated in the IDTech engineering specification documents OEM I-MD22292 B0-03, OEM I-MD22292 B2-03, OEM I-MD22292 C0-03, and OEM I-MD22292 C2-03. Also in these documents is a timing chart which exactly matches the one used in the Iiyama and ViewSonic manuals. As there's a detailed page for each timing mode in these documents, I'll reconcile what I've compiled with these.

 

[jeremywork]

This finding is corroborated in the IDTech engineering specification documents OEM I-MD22292 B0-03, OEM I-MD22292 B2-03, OEM I-MD22292 C0-03, and OEM I-MD22292 C2-03. Also in these documents is a timing chart which exactly matches the one used in the Iiyama and ViewSonic manuals. As there's a detailed page for each timing mode in these documents, I'll reconcile what I've compiled with these.

To summarize the differences I found in those four documents, it seems variants B0 and C0 do not ship with a cable or a card, variant B1 ships with a card, and variants B2 and C2 ship with a cable.

Variants B0 and B2 (probably also B1 but I don't find its PDF) contain the gamma, color, and timing data found in EDID mode 029. Variants C0 and C2 contain the gamma, color, and timing data found in EDID mode 030 (including H+ / V+ for 3840 x 2400 @ 13.0 Hz.)

The B variants reference the ATI FireGL4, Windows 2000, and Linux.
The C variants reference the Radeon 8500 Mac Edition, OS 9, OSX, D50, and DG2.

IBM's declaration of service withdrawal on June 17 2008 mentions 9503-DG2 and 9503-DG4 among all the other variants.

If 9503-DG2 is a Mac variant of 9503-DG3, I wonder if 9503-DG4 is a Mac variant of 9503-DG5?
I'll try to flash those other EDID modes to see if any others return 48 Hz with 1.8 gamma...

There are also references to IDTech MD22292A1 which seems to be the equivalent to 9503-DG0 as this brief PDF shows the analog brightness slider in the photo.

https://www.phys.szote.u-szeged.hu/bbalazs/bb/lcd/MD22292A1.pdf

I think I'll cross "add comprehensive list of rebadged models" off my to-do list now 😴

 

[Amethyst1]

I'll definitely incorporate this information into the EDID data section as well as the model breakdown section.

Correction/addition for the 9503-DG1:

• It can also do 2× 1920×2400@20Hz (EDID 029). It's right there in the pdf I linked to, so sorry for missing that.
• EDID 021 ("DG1 compatible enhanced") for later models includes 3840×2400@12.1Hz. I'm not sure if this implies/prooves that the 9503-DG1 can do 3840×2400@12.1Hz though.

Out of curiosity did you try (after power cycling) dumping the EDID in SRX anyways?

Yes. It didn't change. Interestingly, I checked the current EDID before changing it and it showed up as "029x".

I did pop the shield off (easier than I expected) and it's using a Lontium LT6711A.

The datasheet says it can handle HDMI 2.0 and 8bpc RGB input, so it shouldn't™ cause banding.

ViewSonic's 'Supported display modes' page basically mirrors Iiyama's without polarity data but there are a few discrepancies between these and the IBM/IDTech 'Supported display modes' page:

I wish I'd kept my VP2290b despite the shattered panel ... or just gotten another one while the seller was still active (this was in 2019). Then, we'd be able to examine its EDIDs and add its modes to your page.

This finding is corroborated in the IDTech engineering specification documents OEM I-MD22292 B0-03, OEM I-MD22292 B2-03, OEM I-MD22292 C0-03, and OEM I-MD22292 C2-03. Also in these documents is a timing chart which exactly matches the one used in the Iiyama and ViewSonic manuals.

Interestingly, according to these documents, synchronisation (genlock) is only required for 4× 960×2400 and 4× 1920×1200 and merely recommended for 2× 1920×2400. So, maybe 9503-DG1/DG3 and equivalents only require genlock when using four links. Assuming genlock is required, would its absence result in sync issues/loss or just heavy tearing?

The C variants reference the Radeon 8500 Mac Edition, OS 9, OSX, D50, and DG2.

The Radeon 8500 Mac Edition has only one (single-link) DVI. Its manual explicitly mentions it can run 3840×2400 at 15 Hz (page 33).

IBM's declaration of service withdrawal on June 17 2008 mentions 9503-DG2 and 9503-DG4 among all the other variants.

Good find. I did find a link to an eBay auction for a 9503-DG4 from 2018 but it's gone (unsurprisingly).

 

[jeremywork]

It can also do 2× 1920×2400@20Hz (EDID 029). It's right there in the pdf I linked to, so sorry for missing that.

I missed it too! Got it on the page now.

EDID 021 ("DG1 compatible enhanced") for later models includes 3840×2400@12.1Hz. I'm not sure if this implies/prooves that the 9503-DG1 can do 3840×2400@12.1Hz though.

It's really hard to tell whether DG1 and DG3 are actually any different beyond their shipping bundles. Since 020 is said to be "DG3 only" on the DG3/DG5 manual and "DG1" on the EDID flash page (all others "DG1/DG3") it seems only to be a shipping default because the Matrox card would've been included.

It's separately perplexing that 12.1 Hz would be mixed with 20.1 and 40.9 instead of 12.7 like the other 41Hz schemes use, but that's definitely what 021 describes...
Related is another perplexing note in section 3.2.3 "Scan Conversion" on all the MD22292 Engineering spec PDFs that mentions "support 20.5 Hz to 41 Hz up conversion to reduce flicker effect under slow refresh mode" but then only lists 12.7, 13.0, 20.1, 24.1, 25.1, and 41Hz. Maybe for vintage 20fps film applications??

Yes. It didn't change. Interestingly, I checked the current EDID before changing it and it showed up as "029x".

Mine does this exact same thing. Any mode I set it to will change and (if it's valid) the EDID mode works, but displays with an 'x' at the end. I still just retry if it ever exceeds 63 or shows all zeros as the manual says and it seems okay. I've also noticed the display with the two DH2Gs will always show '192' whenever it's being driven that way. I need to plug in a more conventional source before it'll read its EDID mode normally.

The datasheet says it can handle HDMI 2.0 and 8bpc RGB input, so it shouldn't™ cause banding.

Interesting... I wonder if something in the Matrox devices is causing it to handshake differently. I assume since I see the same number of vertical bands all the way up joevt's procedurally generated test pattern that it is 6bpc at play here.

I wish I'd kept my VP2290b despite the shattered panel ... or just gotten another one while the seller was still active (this was in 2019). Then, we'd be able to examine its EDIDs and add its modes to your page.

One day another opportunity will present. Maybe by then some of the other WIP/todo tasks will be closer to done. I never imagined so much could be written on essentially a single IPS panel 😂

Interestingly, according to these documents, synchronisation (genlock) is only required for 4× 960×2400 and 4× 1920×1200 and merely recommended for 2× 1920×2400. So, maybe 9503-DG1/DG3 and equivalents only require genlock when using four links. Assuming genlock is required, would its absence result in sync issues/loss or just heavy tearing?

I figured the panel would exhibit intermittent blackouts as it does when the signal can't quite be directly driven and the retiming buffer cuts in and out. This happens very occasionally with the two DH2Gs and somewhat frequently at 48.772 Hz on the DGM/DGPs. It's possible this is also the cause of the blackouts at >55Hz with two DVI-DL 1920x2400 signals. If you see tearing right on the line between the two tiles/stripes but both independent signals look clean I'd say genlock isn't actually required. @Gordan does mention that ATI cards have a much worse time keeping non-genlocked signals aligned, surmising nVidia cards can use the same framebuffer for multiple signals even without genlock. Point 5): https://www.eevblog.com/forum/gener...HPSESSID=5quoakqme3ndkc3pfhemije030#msg265450

This was the case for me when using the flashed FireGL X3-256MB (X850 Mac ROM) with 2 x 1920 x 2400 @ 24 Hz. Scrolling a window across the center would be very noticeably delayed on one side. That card/ROM has a working dual-link port though so that's a better way to achieve 24 Hz...

If genlock isn't required even at 4x 960x2400 I really do wonder if the points about 41Hz being "fixed" have any merit either. On the DG5/DGM/DGP the retiming target is relatively "fixed" but the direct panel sync range definitely isn't.

The Radeon 8500 Mac Edition has only one (single-link) DVI. Its manual explicitly mentions it can run 3840×2400 at 15 Hz (page 33).

Interesting that they pushed it up to 15 Hz but not 17 Hz... The TiBook's Mobility 9000 (based on the 8500) has no problems at 17.206...

Good find. I did find a link to an eBay auction for a 9503-DG4 from 2018 but it's gone (unsurprisingly).

I saw that too. I wonder if it got picked up by any market tracking aggregator/etc...

 

[jeremywork]

@joevt I got closer to figuring this out.

The log reports Lilu version DBG-165-2023-04-02 and WhateverGreen DBG-161-2023-04-02.

I removed the SRX-installed override and this resolved the black screen condition. It seems with Lilu/WhateverGreen installed I'll get a black screen whenever I select any of the >360.00MHz resolutions. I ran AllRez dumps with the override removed (3840 x 1024 default for the TH2G) as well as with a new SRX 360 MHz mode selected and the existing 365 MHz modes showing 'Active' in SRX.

Hopefully this isn't a waste without a 365MHz mode actually running...

 

[Amethyst1]

Mine does this exact same thing. Any mode I set it to will change and (if it's valid) the EDID mode works, but displays with an 'x' at the end.

SUCCESS! The EDID did change after all, but the Iris Pro and Delock 62603 were messing things up. Switching to a GeForce 8600M GT with a straight DVI connection did the trick, and I was able to examine all valid and reserved settings for the 9503-DG3 apart from 007, which resulted in the current setting being displayed.

Dumps from SwitchResX (030 is missing as you've already documented it) are attached, minus the hex and serial number part. New modes uncovered:

• 1280×1024@59.759Hz (031)
• 1920×2160@20.100Hz (008; 010)
• 3840×2400@15.000Hz (027)

 

[jeremywork]

SUCCESS! The EDID did change after all, but the Iris Pro and Delock 62603 were messing things up. Switching to a GeForce 8600M GT with a straight DVI connection did the trick, and I was able to examine all valid and reserved settings for the 9503-DG3 apart from 007, which resulted in the current setting being displayed.

Dumps from SwitchResX (030 is missing as you've already documented it) are attached, minus the hex and serial number part. New modes uncovered:

• 1280×1024@59.759Hz (031)
• 1920×2160@20.100Hz (008; 010)
• 3840×2400@15.000Hz (027)

This is exciting news!
I'll get these into the EDIDs section pronto.

So that's probably where the 15 Hz note from the Radeon 8500 manual comes from...

Thanks for saving me from the tedium 😁

 

[joevt]

@joevt I got closer to figuring this out.

The log reports Lilu version DBG-165-2023-04-02 and WhateverGreen DBG-161-2023-04-02.

I removed the SRX-installed override and this resolved the black screen condition. It seems with Lilu/WhateverGreen installed I'll get a black screen whenever I select any of the >360.00MHz resolutions. I ran AllRez dumps with the override removed (3840 x 1024 default for the TH2G) as well as with a new SRX 360 MHz mode selected and the existing 365 MHz modes showing 'Active' in SRX.

Hopefully this isn't a waste without a 365MHz mode actually running...

The point was to see why it can do > 360MHz so it's not going to be useful if it's not doing > 360 MHz.
At least we see that AllRez can show the attributes that are being set or get using the -iofbon patches.
I guess I should see if there's any default iofb patches that should be disabled so that default behavior will allow the > 360 MHz modes to output something as before.

While Lulu is installed, maybe change the boot-arg to -iofboff to see if that can get the > 360 MHz modes to work. Restart after changing the boot-arg.

 

[Amethyst1]

Thanks for saving me from the tedium 😁

It's been great fun Do you reckon it makes sense to try the remaining numbers from 001...063 that aren't documented at all (i.e. not even listed as "reserved")?

You have the modes for EDID 013 and 020 mixed up on your site. 013 is four tiles only, 020 is four stripes only.

If genlock isn't required even at 4x 960x2400 I really do wonder if the points about 41Hz being "fixed" have any merit either. On the DG5/DGM/DGP the retiming target is relatively "fixed" but the direct panel sync range definitely isn't.

I'll see that I can get some 240fps video of my 9503-DG3 at different refresh rates. I'll probably upload them somewhere for you to have a look at them, since you know what to look for.

I figured the panel would exhibit intermittent blackouts as it does when the signal can't quite be directly driven and the retiming buffer cuts in and out.

I don't remember seeing blackouts at 4× 960×2400@41Hz from the ATI Radeon HD 6870.

One day another opportunity will present.

Hopefully. I've only ever seen the IBM and ViewSonic variants. Have you come across the others? If I ever do, I'll see that I can get my hands on one.

Thinking about it, IIRC the seller I bought my Berthas from had DGP's (no DG5's/DGM's), DG3's, and VP2290b's. I don't remember whether I inquired about DG0's or DG1's. Part of me thinks I should have impulse-bought all Berthas he had, if only to have enough spares...

I never imagined so much could be written on essentially a single IPS panel 😂

Mighty Bertha is unlike any other monitor out there.

It's really hard to tell whether DG1 and DG3 are actually any different beyond their shipping bundles. Since 020 is said to be "DG3 only" on the DG3/DG5 manual and "DG1" on the EDID flash page (all others "DG1/DG3") it seems only to be a shipping default because the Matrox card would've been included.

Yeah, 020 includes 960×2400@41Hz only, presumably because each of the GPUs on the Matrox card only has 8 MB RAM, which is insufficient to run 1920×2400 or 3840×2400 at high colour depth. This evaluation of a 9503-DG1 mentions it being used with an ATI FireGL 4 and Radeon 8500 after flashing the appropriate EDIDs, so it's possible that the only differences between the DG1 and DG3 are the video card, cables and default EDID.

I do have a Matrox G200 MMS BTW but could only get the analog outputs working. DVI monitors are black.

This is exciting news!

The 1920×2160 modes made me smile: Mighty Bertha can even do Ultra HD! Joking aside, those may have been included because apparently some 3Dlabs Wildcat GPUs can only run 3840×2160 (pdf: page 29).

 

[jeremywork]

The point was to see why it can do > 360MHz so it's not going to be useful if it's not doing > 360 MHz.

Got it. Sorry about that.

While Lulu is installed, maybe change the boot-arg to -iofboff to see if that can get the > 360 MHz modes to work. Restart after changing the boot-arg.

Unfortunately I still get the black screen after restarting with -iofboff. It seems like the whole machine freezes when the display tries to activate. The onboard goes black but backlight stays on, and the T221 begins syncing and also displays a backlit black screen. Even trying to turn off the keyboard backlight doesn't work.

It's been great fun Do you reckon it makes sense to try the remaining numbers from 001...063 that aren't documented at all (i.e. not even listed as "reserved")?

You have the modes for EDID 013 and 020 mixed up on your site. 013 is four tiles only, 020 is four stripes only.

I'm also curious about the missing numbers, though that's uncharted territory and I don't want to advise you do that to a working panel. I'll probably try on my damaged one (plus, with the plastics removed it's much easier to hit the hidden reset switch if things go haywire.)

Thanks for noticing that error, it's corrected now. I also realized I shouldn't have ordered the modes based on size; they're all in the order they appear on the original EDIDs now.

Have you come across the others?

I honestly hadn't been looking for them for over a decade- it just hit me last year that suddenly they're not thousands of dollars anymore.

I do have a Matrox G200 MMS BTW but could only get the analog outputs working. DVI monitors are black.

There does seem to be some mention of "custom" G200MMS. I didn't know if this was in regards to the card itself or the bundled cabling (LFH-60 straight cables seem rarer than the displays!)

The 1920×2160 modes made me smile: Mighty Bertha can even do Ultra HD! Joking aside, those may have been included because apparently some 3Dlabs Wildcat GPUs can only run 3840×2160 (pdf: page 29).

I had wondered how that 2160 limit was accounted for... also most of the EDIDs you supplied have 1920 x 1080 on the list of standard timings. It's not on any of the ones I had previously.
It seems silly that EDID mode 008 bothers to define vertical refresh limits of 40-56 Hz and then only supply resolutions using 12.7 and 20.1.

 

[Amethyst1]

I'm also curious about the missing numbers, though that's uncharted territory and I don't want to advise you do that to a working panel. I'll probably try on my damaged one (plus, with the plastics removed it's much easier to hit the hidden reset switch if things go haywire.)

OK, I'll let you explore the uncharted territory then. It's probably "better" to use a DG5/DGM/DGP anyway as it may have additional undocumented EDIDs compared to the DG1/DG3.

There does seem to be some mention of "custom" G200MMS. I didn't know if this was in regards to the card itself or the bundled cabling (LFH-60 straight cables seem rarer than the displays!)

My G200MMS came with LFH-60-to-2×VGA cables. I bought LFH-60-to-2×DVI cables. The card has four TMDS transmitters so unless something else is missing or I've got the wrong cables, DVI should™ work I think.

I had wondered how that 2160 limit was accounted for... also most of the EDIDs you supplied have 1920 x 1080 on the list of standard timings. It's not on any of the ones I had previously.

I wonder why they didn't document 008 and/or 010 for these cards.

 

[joevt]

I did some testing with ATI Radeon HD 4870 (Mac upgrade GPU), Snow Leopard 10.6.8, MacPro3,1. This is a DisplayPort 1.1 card so the max link rate is HBR x4 (360 MHz for 8bpc).

I was able to get 4K > 360MHz modes with Acer XV273K set to DP 1.2. However, the refresh rate from the detailed timing does not match the refresh rate of the display. It appears to be some kind of bug with the 10.6.8 drivers allowing these modes without verifying the max link rate of HBR x4.

41Hz -> 41Hz (pixel clock is less than 360MHz so this mode works as expected)
42Hz -> 40Hz (refresh rate drops the further away from the max pixel clock you get)
43Hz -> 35Hz
44Hz -> 29Hz
45Hz -> 24Hz
46Hz -> ??? (black screen since display doesn't support < 20Hz)

Since the refresh rate doesn't match, I didn't bother trying to check attributes with Lilu/WhateverGreen.

Same occurs in 10.7.5 Lion.

 

[Amethyst1]

However, the refresh rate from the detailed timing does not match the refresh rate of the display.

Thanks for getting to the bottom of this. This explains it. So as the refresh rate drops, the pixel clock never actually exceeds 360 MHz, right?

 

[Amethyst1]

@jeremywork: I couldn't help going through all the remaining totally undocumented settings from 001≈063 on my 9503-DG3 and 9503-DGP and have uncovered EDIDs 005, 033, 041 and 046. Dumps are attached. Only 033 works on DG3.

New modes (all in 005):

• 1920×1200@47.999Hz
• 1920×2400@48.000Hz (dual-link)
• 3840×2400@24.000Hz (dual-link)

I'm also considering making a table to list all known modes and EDIDs that have them. My idea was to have one row per mode and the EDIDs in the columns. Do you reckon it's better to have one column for all EDIDs that have a given mode, listing them as comma-separated values, or to have one column for each EDID, and a yes/no dichotomy?

Just for kicks, I also tried documented EDIDs listed as DG5/DGM/DGP-only on the DG3. None were accepted.

 

[joevt]

Thanks for getting to the bottom of this. This explains it. So as the refresh rate drops, the pixel clock never actually exceeds 360 MHz, right?

As far as I can tell, the pixel clock drops as I try to increase it beyond 360 MHz. The display's onscreen menu doesn't show MHz. It only shows kHz and Hz. Same for the info reported by AllRez in the Timing Report and the VCP parsed info.

The 3840x2160 timings I used were CVT-RB2 so you can calculate the expected MHz. Maybe there's a way to calculate the relationship between the expected MHz and the obtained MHz (calculated from the result Hz using the total pixels of the timing).

So I guess the question is, did you guys get a correct Hz match between the timing and the display for the > 360Mhz modes? Or did you get a mismatched refresh rate? If the former then there's still a mystery. If the latter then it's what I experienced.

 

[joevt]

10.8.5 and later won't let the display work as DP 1.2 so I have to set it as DP 1.1. Then macOS won't valid > 360 MHz modes.

 

[Amethyst1]

So I guess the question is, did you guys get a correct Hz match between the timing and the display for the > 360Mhz modes? Or did you get a mismatched refresh rate? If the former then there's still a mystery. If the latter then it's what I experienced.

Neither the Huawei MateView nor LG UltraFine 4K (my only displays that can do >360 MHz) show refresh rate.

 

[jeremywork]

42Hz -> 40Hz (refresh rate drops the further away from the max pixel clock you get)
43Hz -> 35Hz
44Hz -> 29Hz
45Hz -> 24Hz

This is interesting. Perhaps then my MBP isn't actually exceeding 360MHz (and maybe Lilu/WG tries harder and fails as a result?)

I couldn't help going through all the remaining totally undocumented settings from 001≈063 on my 9503-DG3 and 9503-DGP and have uncovered EDIDs 005, 033, 041 and 046. Dumps are attached. Only 033 works on DG3.

I don't blame you 😄 Just didn't want that on my conscience if something were to brick. I'll add these to the list.

I'm also considering making a table to list all known modes and EDIDs that have them. My idea was to have one row per mode and the EDIDs in the columns. Do you reckon it's better to have one column for all EDIDs that have a given mode, listing them as comma-separated values, or to have one column for each EDID, and a yes/no dichotomy?

Thinking about this, the way it is on the 'Extended modes' section of the manuals makes the most sense, so that for each time a mode appears on an EDID, the priority can be expressed as P1-P4. I'd probably just add all the unmentioned EDID modes as new columns, the specific blanking intervals as additional header columns, and also the color and timing ranges as header rows. Then the secondary chart below can still separate DG1/DG3, DG5/DGM/DGP, and all variants for the different EDIDs.

Just for kicks, I also tried documented EDIDs listed as DG5/DGM/DGP-only on the DG3. None were accepted.

This makes sense but it's good to know!

So I guess the question is, did you guys get a correct Hz match between the timing and the display for the > 360Mhz modes? Or did you get a mismatched refresh rate? If the former then there's still a mystery. If the latter then it's what I experienced.

I'll test to confirm this later today. It looks like none of the modes I had used in the AllRez dump would've rolled over on the OSD between 360 and 365 MHz.

Mighty Bertha is unlike any other monitor out there.

Strange footnote: looks like Toshiba produced an imitation of Bertha (CDU2150A) to fulfill part of a contract they won.

New poorman's 4K - あかわにの巣

（1920x2400x2でちゃんと4K表示できるのは確認したぞ）今各所でT...

a-mate.cocolog-nifty.com

Seems it'll only sync on 2x DVI-SL@1920x2400@24Hz (or 34Hz) or 1x DVI-SL at 1920x1200@48Hz, and uses a TN panel...

 

[jeremywork]

uncovered EDIDs 005, 033, 041 and 046

Interesting finds.

It looks like:

005 is logically similar to 002 but with precise 24 and 48 Hz modes instead of NTSC-friendly ones
033 is a copy of 030 with the typical (colder) white balance and missing 1920x2400
041 is a copy of 030 with a 25 Hz 1920x2400 instead of 20.1 Hz
046 is a copy of 045 with a warmer white balance

There was lots of debate in forums regarding a warmer appearance on the DGM models than the DGP versions, but many attributed it to the power-on hours (apparently DGPs weren't produced until the DGM had already been in circulation for a few years.) It would be great to unearth some of the EDID specs for DGM and DGP to see if they were simply running the warmer EDIDs.

Also, post #6 on this page instructs to set EDID to 04. There was nothing for that mode?

【3840x2400】QUXGA-Wのモニタpart5【200ppi】

 

[Amethyst1]

Also, post #6 on this page instructs to set EDID to 04. There was nothing for that mode?

Nothing on the DG3, nothing the DGP. Maybe the DGM is different or the DCC version matters?

 

[joevt]

This is interesting. Perhaps then my MBP isn't actually exceeding 360MHz (and maybe Lilu/WG tries harder and fails as a result?)

Maybe. Does the display tell you what the refresh rate is? Does the refresh rate shown by the display match the expected refresh rate?

 

[jeremywork]

Nothing on the DG3, nothing the DGP. Maybe the DGM is different or the DCC version matters?

I just did a full sweep on one of my DGMs (week 36, 2004.)

It was only able to run 002, 003, 005, 006, 021, 022, 029, 030, 033, 036, 037, 040(new!), 041, 044, 045, and 046.

Mode 033 is the only one which uses a descriptor for Monitor name; your DG3 reports 'IBM9503-T221' and my DGM reports '9503-DG5.'

Mode 040 is the same as 041 but 23.999 Hz instead of 25.001 Hz for 1920 x 2400.

I finally received that other TH2G Digital SE from UPS, so I'll take some photos and test >360MHz now...