It only tells the refresh in whole integers, but I have tested previously that .49 Hz rounds down and .50 Hz rounds up, so I should be able to get an accurate picture if I choose a total image size which places 360MHz below and 365MHz above this point.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Esoteric Graphics Questions/Findings (another T221 thread)
- Thread starter jeremywork
- Start date
- Sort by reaction score
The TripleHead2Go Digital SE uses IDT VMM1403NQG. I'll get the photos assembled a little later.
So it is actually rolling over, implying 365MHz is accurate. I also don't see any of the 6bpc banding (as I do with the HDMI adapter.)
Retesting 24 Hz and 30 Hz with the HDMI adapter also reveals the banding isn't present on these. It seems only to affect the faster modes, including 37 Hz @ 358.20 MHz.
Photos of this are now here: http://stormdecay.com:1984/rd/rd_matrox.html
Having looked at the IDTech Engineering spec documents it seems they have provisioned for vertical blanking as low as 1, 1, 1 for all modes except 4 x 960x2400, which requires 2, 1, 1. Testing again with the DVI-DL adapter reveals this works in 3840 x 2400 mode, so 2, 3, 2 can be disregarded as the tightest interval, which saves three lines. Using the 2012 MBP and Apple's mDP to DVI-DL adapter I was able to get new maximum data for >330MHz modes; they're on the custom/overclocked input chart on my page now.
As such I am now less confident about the 360MHz vs 365MHz finding from a couple hours ago.
I was paying good attention to the OSD during these tests and it revealed there are sometimes discrepancies. I'm not sure if the OSD is to blame or the computer/SRX is reporting false timing...
As such I am now less confident about the 360MHz vs 365MHz finding from a couple hours ago.
What's its DCC version? My DG3 has DCC version 3.5 and my DGP has DCC version 4.5. If any of your Berthas have other DCC versions, it may be worth trying the modes we've not yet gotten to work on them to see if more can be uncovered. My DG3 won't flash mode 021, which IBM lists as valid for DG3 but not supported by DCC 3.5 so there's definitely some variation within models.
Code:
001
004
007
---
011
014
015
016
017
018
019
---
023
024
025
026
028
---
032
038
039
---
042
043
047
048
049
---
050
051
052
053
054
055
056
057
058
059
---
060
061
062
063Modes 008, 009, 010, 012, 013, 020, 027 and 033 have the name on my DG3.
Thanks for the teardown. So it uses the same ViewXpand chipset family as the DualHead2Go ME and SE.
Interesting. I've found some more info in Japanese at [https://alfoday.blogspot.com/2014/08/toshiba-cdu2150a.html] and [https://ameblo.jp/ittekuruppa/entry-11739419628.html]. TN is kind of a letdown though, but that's not to say I wouldn't mind playing with one of these.
I've also been thinking if there's a spiritual successor to Bertha. Sure, she's been outdone in resolution etc., but none of these modern high-res monitors require more than one (or two) links to work at full resolution and/or refresh rate. So if the criterion is "requires at least three links", then the Canon DP-V3010 (30", IPS, 4096×2560) qualifies. It's definitely thick and heavy and expensive enough and, according to the manual, requires four DisplayPort links at 2048×1280@59.9Hz each.
All five of mine have DDC version 4.5 (though when I drive with two DH2Gs I see DDC version 0.0; other numbers persist and I get EDID mode 192.)
Sure enough! Selective blindness isn't a good thing for someone with this many pixels 😅
Gladly! So DP1.1a limits confirmed.
I did find this one for not too big a pricetag; looks like the Matrox Parhelia is also included. https://www.fromjapan.co.jp/japan/en/special/order/confirm/https://jp.mercari.com/item/m70295448946/13_1/lgk-link_top_searchInteresting. I've found some more info in Japanese at [https://alfoday.blogspot.com/2014/08/toshiba-cdu2150a.html] and [https://ameblo.jp/ittekuruppa/entry-11739419628.html]. TN is kind of a letdown though, but that's not to say I wouldn't mind playing with one of these.
This post seems to indicate it must receive two links in order to display either 1920 x 2400 stripe; otherwise it's limited to 1920 x 1200 @ 48Hz. Seems more similar to the way the Apple 30" Cinema behaves.
Still, I do see mumblings of 33 Hz support on the description of this fan disabler thing, so it's likely a good candidate for use with the DH2G DP edition.
I've considered that in order to be even close to the same animal, the onboard framebuffer system would be a defining point. If it were possible to drive the Apple 30" at 2560x1600@30Hz over DVI-SL, for instance, it would be a more likely contender. I think DisplayPort's ability to transmit only update data separate from the panel refresh basically alleviates this need. That said, I wonder how the CDU2150A handles 24 Hz. I seem to recall LCD panels not performing very well when driven much lower than 40 Hz (hence all of Bertha's tricks.)
The DP-V3010 has always seemed like the next logical jump. Sadly, I don't think they were produced in nearly the numbers required to eventually pick up several for a scrap prices... That Bertha was produced at all was kind of a fluke, but in the quantities we find them seems even less likely.
...I'd like to be wrong though. It did take 20 years for Bertha
Last edited:
Maybe "04" is a typo and it's supposed to be 040?
Any insights on what the modes I uncovered on the DG3 are based on? We've seen new timings in 008, 010, 027 and 031 (what's the deal with 1280×1024@59.795 in 031?) but what about 009 and 031? 012 is four tiles and four stripes, 013 is four tiles, 020 (documented) is four stripes so they're basically variations of the four-input scheme.
I tried 040 on my DGP just now and it worked. I seem to remember it not working yesterday as I'd have dumped it otherwise. Perhaps it just needed a second try.
For me that would be 140 euros plus shipping. Hmmmmm...
That's a good point. 2560×1600 isn't high enough, resolution would have to be at least equal to Bertha's IMHO.
I had an Asus MG24UQ that accepted timings as low as 10 Hz. That's a modern LCD though, and I didn't test whether it was directly driven at that refresh rate or if buffering/retiming was involved.
But I'm infinitely glad she exists. The Huawei MateView (I use two with my main rig) has a "better" resolution (3840×2560) and aspect ratio (3:2) but it only appeared in 2021.
Last edited:
I've just made another pass at flying even closer to the sun with the 2015 MBP and 62603. It looks like even though vertical intervals 1, 1, 1 produce a slow crawl on the TH2G they work okay on a direct connection, at least when there's only one link being driven. These are now added to my chart.
I'm beginning to wonder if active DisplayPort adapters are obfuscating the actual driven intervals though. I seem to be having success with modes I'm sure the T221 rejected from the TiBook. (3840 x 2320 works at 48, 32, 56; 1, 1, 1 at 165.32 MHz, for instance.) If this turns out to be the case, my tested modes section will be split into active DP modes vs direct DVI modes and I'll get new numbers from the TiBook & 2008 MBP.
Now that I've had the second TH2G Digital SE handy I also rechecked this on the 2015. Seems I can actually use 16, 32, 72 for horizontal and remain stable at 2, 1, 1 or 1, 1, 1 vertical (1, 1, 1 still produces the undesirable crawl though.)
This brought about a new discovery. When I shortened the H from 48, 32, 56 to 16, 32, 72 I could no longer use 365MHz, now only 363.16 MHz. So I'm beginning to wonder if only the picture data must be fit into 360 MHz, and the blanking intervals are communicated to the active DP adapter when the mode is set? SRX wouldn't know this and would still total the blanking as DVI would require...
I noticed some of the EDID modes use very slightly different color calibration from all the 'official' ones, not just gamma or white balance. It's so slight it's basically margin of error, but there's definitely something going on there.
If you haven't already started, I think I'll build out the EDID mode chart tomorrow to help visualize this better. My brain's too full of numbers right now 😜
I'm beginning to wonder if active DisplayPort adapters are obfuscating the actual driven intervals though. I seem to be having success with modes I'm sure the T221 rejected from the TiBook. (3840 x 2320 works at 48, 32, 56; 1, 1, 1 at 165.32 MHz, for instance.) If this turns out to be the case, my tested modes section will be split into active DP modes vs direct DVI modes and I'll get new numbers from the TiBook & 2008 MBP.
Now that I've had the second TH2G Digital SE handy I also rechecked this on the 2015. Seems I can actually use 16, 32, 72 for horizontal and remain stable at 2, 1, 1 or 1, 1, 1 vertical (1, 1, 1 still produces the undesirable crawl though.)
This brought about a new discovery. When I shortened the H from 48, 32, 56 to 16, 32, 72 I could no longer use 365MHz, now only 363.16 MHz. So I'm beginning to wonder if only the picture data must be fit into 360 MHz, and the blanking intervals are communicated to the active DP adapter when the mode is set? SRX wouldn't know this and would still total the blanking as DVI would require...
This, or perhaps it was meant for a ViewSonic display.
I'll have to sit with this and think. The 1280 x 1024 is likely for use with some legacy device which outputs its BIOS screen over a link which can't do 1920 x 1200, or something along these lines.
I noticed some of the EDID modes use very slightly different color calibration from all the 'official' ones, not just gamma or white balance. It's so slight it's basically margin of error, but there's definitely something going on there.
If you haven't already started, I think I'll build out the EDID mode chart tomorrow to help visualize this better. My brain's too full of numbers right now 😜
I'll slowly get to performing the entire sweep on all five of my displays. One down...
Go for itFor me that would be 140 euros plus shipping. Hmmmmm...
Maybe it's a conspiracy theory, but if Apple really wanted to bring 200ppi to market as 'their own' concept, creating a flagship display which could be driven over a single cable and did boast basically the largest area and number of pixels available in the consumer market as fast as possible would've been an excellent distraction. They could even feign ignorance for almost a decade about the value of >100ppi, relegating anything of the sort to the 'luxury' 17" models which they could just kill as soon as their version was ready anyhow. Might be supposing intention when really they had luck, but maybe they saw the writing on the wall for Bertha when HDCP was poised to stand in the way until DP was ready enough to handle what they wanted anyhow. The HDCP 'compatibility' they bodged in for the 30 is kinda sad. I think it drops to 1280x800+AA so the content can be passed without violating the rules.
I don't know that I would ever have understood 3/4ths of this tech without prodding at these. I had blindly followed "DVI-SL limited to 1920x1200 and DVI-DL limited to 2560x1600" logic unitl a few months ago.But I'm infinitely glad she exists. The Huawei MateView has a "better" resolution (3840×2560) and aspect ratio (3:2) but it only appeared in 2021. And it's boring: plug in one DP/HDMI/USB-C and be done
Last edited:
004 is "reserved" for ViewSonic, so I really wonder what's hiding behind that (and 005).
They're converting a DisplayPort signal to single- or dual-link TMDS, so who knows what else they might be doing?
That's what I thought since some TMDS transmitters are limited to 112 MHz pixel clock and 1280×1024@60 is below that (unless you use the variant that requires 135 MHz). However, don't all EDIDs already contain a 1280×1024@60 mode at <112 MHz?
I haven't started, so if you want to, go ahead
Great, looking forward to what you manage to uncover. Forget what I said about my DGP not taking 040: I tried it again just now and it worked.
For me, that logic started to break once I'd come across Bertha (no idea when that was, possibly during its active lifetime though)... and when I hooked up a 4K/UHD monitor to a PowerBook G4 via DVI-DL in 2018, I saw it with my own eyes.
I think the problem might be that some devices use the priority #1 mode without verifying compatibility. The description of this old listing for LFH-60/DVI-DL+EDID override cable seems to support this. I hope one day I can grab a pair of these somehow before I cut some cables up...
IBM T221専用 長さ&機能オーダー形式 LFH60-DualLinkDVIケーブル「中古」の落札価格｜22インチ～｜Yahoo!オークション(旧ヤフオク!) 落札相場- オークファン（aucfan.com）
IBM T221専用 長さ&機能オーダー形式 LFH60-DualLinkDVIケーブル(Yahoo!オークション(旧ヤフオク!) )は2件の入札を集めて、2015/10/18 22:56に落札されました。
Will post it when I get it done.I haven't started, so if you want to, go ahead
That's nice to hear, consistency is a relief
I had learned about them too back when they were still in production, but I just figured they must need some highly custom DVI interface, not literally just a PowerBook. All those spec sheets with "max" resolution at 1920x1200/2560x1600 didn't help the assumption.
The specs implicitly assume a 60Hz refresh rate and some form of standardised timings. When single-link DVI was introducd in 1999, everyone said it would do 1600×1200@60Hz (161 MHz GTF/CVT). Then, reduced blanking timings (CVT-RB) were standardised in 2002 or so and it "suddenly" did 1920×1200@60 (154 MHz).
I presume this was also done to avoid confusing people who might know the resolution of their brand-new LCD, but have never heard about refresh rate. A spec sheet that says e.g. "single-link DVI: up to 1920×1200@60Hz, 2560×1600@30Hz or 3840×2400@15Hz" might have provided more questions than answers in that case ("it says 2560×1600@30Hz, so will it run my 30" ACD?")
It all makes sense in context, and given that Berthas carried five figure pricetags and 30” Cinemas can unfortunately not run at 30Hz over DVI-SL it wouldn’t have helped much to explain it on every spec sheet.The specs implicitly assume a 60Hz refresh rate and some form of standardised timings. When single-link DVI was introducd in 1999, everyone said it would do 1600×1200@60Hz (161 MHz GTF/CVT). Then, reduced blanking timings (CVT-RB) were standardised in 2002 or so and it "suddenly" did 1920×1200@60 (154 MHz).
I presume this was also done to avoid confusing people who might know the resolution of their brand-new LCD, but have never heard about refresh rate. A spec sheet that says e.g. "single-link DVI: up to 1920×1200@60Hz, 2560×1600@30Hz or 3840×2400@15Hz" might have provided more questions than answers in that case ("it says 2560×1600@30Hz, so will it run my 30" ACD?")
I caught up on the thread where you originally tested 4K with the Atlona. FWIW I also tried the T221 with a VTBook and got only 1600x1200, but now I know I should’ve checked that I was using an EDID mode with 1920x1200 in the standard timings section. I got it to work anyways by starting it connected to a 23” Cinema (might’ve been the SX2462W?) and then hotplugging the T221. I don’t think DisplayConfigX could even attempt to send custom modes to it though. I was going to try the mode 006 scheme on the inhibited Radeon 9700 DLSD since the framebuffer issue wouldn’t matter.
Last edited:
Does SwitchResX 3.8.7/4.3.6 not work for you on PPC? The VTBook is basically a PCI (Trident XP2) graphics card.
Quick test: 2× 3840×840@48Hz CVT-RB works on the DGP. Tighter timings and/or >165 MHz eek out some extra height. I was doing this with the Iris Pro 5200 and two Delock 62603s, but SwitchResX still saw my DG3's EDID! So, something’s up with this particular setup.
Can you do some testing to reveal at what pixel clock the banding starts? As it starts below 360 MHz, the adapter isn’t very useful in our cases. Delock have some adapters using the Lontium LT6711A but they’re a bit pricey.
Last edited:
I was doing things backwards- I found SRX when I moved forward in time after DCX didn't work on my newer machines. I'll go back and test eventually.
I had some issues with SRX dumping the wrong EDID when I was using it on my 2008 in Mavericks. I settled on flashing the T221, disconnecting the DVI, quitting SRX, then powering the T221 back on and reconnecting it before relaunching SRX and then it always pulled the correct EDID.
I'll see if I can find the limit on this scheme. After all this testing I basically need to re-write my tested machines section. Those limits weren't the limits!
New tests with the Lontium HDMI -> TH2G Digital SE:
I can use 16, 32, 72 instead of 48, 32, 56 (2, 1, 1) but only up to 419.24 MHz for 3840 x 2400 @ 44.039 Hz. At 419.26 MHz or above, the TH2G just flashes its sync light (could be the built-in cable on the Lontium isn't well shielded; I seem to be able to change the frequency of those blackouts when I 'Mr. Bean' the cable like an antenna.) For now, fullres @ 47.314 Hz cannot be improved upon.
The 6bpc banding seems tied to a limit precisely between 338.54 MHz (no banding) and 338.56 MHz (banding.)
This applies to 48, 32, 56; 2, 1, 1 @ 35.418 / 35.421 Hz and 16, 32, 72; 2, 1, 1 @ 35.562 / 35.564 Hz. Seems like maybe an HDMI 1.3 related limit?
Last edited:
DCX and SRX are supposed to do the exact same thing: create and install EDID overrides. But it’s possible SRX works where DCX doesn’t. SRX has some idiosyncrasies itself.
The 2013 MBP (running Mavericks), Delock 62603s and DGP were power-cycled before pulling the EDID. And I hadn’t tried flashing a different one before. So it must have somehow remembered the DG3’s EDID. Maybe it’s a bug caused by both models sharing the same vendor and product ID.
That’s what I’m thinking as well. The adapter seems fishy.
Last edited:
I have noticed SRX treats the TH2G as the same display whether the Lontium adapter is present or not, but the second TH2G was recognized as a new display (also regardless of if the Lontium was on the chain.)
It's generally kind of a PITA, but I'm glad I found it nonetheless. The other two 4K HDMI > DP adapters I have don't seem to be able to engage the TH2G beyond DP1.1 limits, probably because they stick to 8bpc. It's been a while since I tried these but I know nothing in the 400MHz range worked at all.
Just overhauled the TH2G portion of the maximum modes section on my page.
Okay, I've finished all of the "tightest" timings for the 62603 and for >165MHz modes the T221 seems to limit based on total horizontal timing. Edging .02 MHz past each of the following sends the T221 into a green/yellow fritz; .04 MHz past and it's out of range.
3840 x 2400 - 48, 32, 64; 2, 3, 2 @ 17.274 Hz / 41.579 kHz / 165.65 MHz
3840 x 2400 - 48, 32, 56; 1, 1, 1 @ 17.303 Hz / 41.579 kHz / 165.32 MHz
1920 x 2400 - 48, 32, 56; 2, 1, 1 @ 34.522 Hz / 82.990 kHz / 170.63 MHz
1920 x 2400 - 16, 32, 72; 1, 1, 1 @ 34.536 Hz / 82.990 kHz / 169.30 MHz
1280 x 2400 - 48, 32, 62; 2, 1, 1 @ 51.672 Hz / 124.219 kHz / 176.64 MHz
1280 x 2400 - 48, 32, 56; 2, 1, 1 @ 51.676 Hz / 124.230 kHz / 175.91 MHz
1280 x 2400 - 16, 32, 72; 1, 1, 1 @ 51.697 Hz / 124.229 kHz / 173.92 MHz
I'm going to make a final pass with the A1306 / DVI-DL and then go back to the 2008 MBP and TiBook.
Since the 62603 and A1306 will always be able to exceed 165 and 330 I'll leave them in the slots for overclocked timings, and then just take back the 165.00 and 330.00 slots for direct DVI driven modes, supposing they have stricter interval limits.
3840 x 2400 - 48, 32, 64; 2, 3, 2 @ 17.274 Hz / 41.579 kHz / 165.65 MHz
3840 x 2400 - 48, 32, 56; 1, 1, 1 @ 17.303 Hz / 41.579 kHz / 165.32 MHz
1920 x 2400 - 48, 32, 56; 2, 1, 1 @ 34.522 Hz / 82.990 kHz / 170.63 MHz
1920 x 2400 - 16, 32, 72; 1, 1, 1 @ 34.536 Hz / 82.990 kHz / 169.30 MHz
1280 x 2400 - 48, 32, 62; 2, 1, 1 @ 51.672 Hz / 124.219 kHz / 176.64 MHz
1280 x 2400 - 48, 32, 56; 2, 1, 1 @ 51.676 Hz / 124.230 kHz / 175.91 MHz
1280 x 2400 - 16, 32, 72; 1, 1, 1 @ 51.697 Hz / 124.229 kHz / 173.92 MHz
I'm going to make a final pass with the A1306 / DVI-DL and then go back to the 2008 MBP and TiBook.
Since the 62603 and A1306 will always be able to exceed 165 and 330 I'll leave them in the slots for overclocked timings, and then just take back the 165.00 and 330.00 slots for direct DVI driven modes, supposing they have stricter interval limits.
Last edited:
1, 1, 1 also works in dual-link modes with the A1306, but on fullres requires 48, 32, 64 instead of 48, 32, 56.
The tested DL/SL limits are effectively the same:
3840 x 2400 - 48, 32, 64; 1, 1, 1 @ 34.537 Hz / 82.992 kHz / 330.64 MHz (DVI-DL)
1920 x 2400 - 16, 32, 72; 1, 1, 1 @ 34.536 Hz / 82.990 kHz / 169.30 MHz (DVI-SL)
1920 x 2400 - 16, 32, 72; 1, 1, 1 @ 68.795 Hz / 165.314 kHz / 337.24 MHz (DVI-DL)
960 x 2400 - 8, 32, 56; 1, 1, 1 @ 68.790 Hz / 165.303 kHz / 174.56 MHz (DVI-SL)
I presume if the override/output driver could be controlled more granularly the actual T221 limit is precisely identical.
The tested DL/SL limits are effectively the same:
3840 x 2400 - 48, 32, 64; 1, 1, 1 @ 34.537 Hz / 82.992 kHz / 330.64 MHz (DVI-DL)
1920 x 2400 - 16, 32, 72; 1, 1, 1 @ 34.536 Hz / 82.990 kHz / 169.30 MHz (DVI-SL)
1920 x 2400 - 16, 32, 72; 1, 1, 1 @ 68.795 Hz / 165.314 kHz / 337.24 MHz (DVI-DL)
960 x 2400 - 8, 32, 56; 1, 1, 1 @ 68.790 Hz / 165.303 kHz / 174.56 MHz (DVI-SL)
I presume if the override/output driver could be controlled more granularly the actual T221 limit is precisely identical.
Last edited:
Reflecting more on this - SRX reads both the 62603-attached Bertha and the A1306-attached Bertha as the same display, except when I switch between them I get not only the modes I last used with the other adapter, but also all the modes I deleted while using the other adapter. It's like it knows they're different, but only sort-of. They're not the same T221 either- one DGM (SL) and one DGP (DL.)
If a display doesn't change it's vendor or product ID, then it's possible SwitchResX may be using an EDID that it read from a different source.
For example, the XV273K uses the same vendor/product ID for both its DisplayPort inputs and HDMI inputs. If I create custom timings for the display while connected to HDMI, then the EDID that the custom timings get added to will be from the HDMI port. The custom EDID is stored in the overrides file. If I then connect the display using DisplayPort, then the EDID that gets used will be that of the HDMI port since that is what is being used in the override file.
In this case, it is necessary to click the Restore Factory Settings button to remove the old overrides. Another option is to rename the overrides file and reconnect the display so a new overrides file is created.
SwitchResX may be using the display's manufacture week of the year for the overrides file name instead of product ID. The same problem applies for this naming convention. If the week of manufacture doesn't change between two different connections, then they'll share the same EDID override.
Windows is different than macOS. EDID overrides are stored in the registry instead of a file (use CRU to list and edit the overrides). Each EDID override is per vendor/product/GPU port. In that case, you have to copy an EDID override to multiple GPU ports if you want the same override to exist when the display is connected to a different GPU port. It is useful to include the GPU port name in the EDID override so you can tell the difference.
For SwitchResX, only some timings are stored in a custom EDID. Lower pixel clock timings (< 655.35 MHz) can be stored separately from the EDID in the overrides file in a
dspc property. AllRez shows these in its output.I've finished the EDID chart which includes more granular resolution data and reserved EDID modes.
@Amethyst1, were you able to get 030 on both the DG3 and the DGP?
That's the only assumption I made ...I think
Let me know if you spot any errors.
It's apparent there's a mix of GTF based timings, CVT-RB based timings, and some which were probably based more directly on what the framebuffer could handle (still with plenty of margin.)
Edit: reuploaded; I forgot to add the standard timings
@Amethyst1, were you able to get 030 on both the DG3 and the DGP?
That's the only assumption I made ...I think
Let me know if you spot any errors.
It's apparent there's a mix of GTF based timings, CVT-RB based timings, and some which were probably based more directly on what the framebuffer could handle (still with plenty of margin.)
Edit: reuploaded; I forgot to add the standard timings
Last edited:
Wow. Great work.
Yes.
What adapters are they/what chipset do they use? I have a Wacom Link Plus (DisplayPort 1.2 or HDMI 1.4 —> USB-C) but since it’s 1.4 it’s no good for >340 MHz via HDMI I presume (I’ve never tested it; no need to when there’s DP 1.2).
Last edited:
Thanks for your help gathering the data!
I made a final cosmetic change which should help readability (will update the file I posted above.) Beyond this I made a second version which removes the three duplicate resolutions by shifting polarity onto the priority fields. I'm not convinced this helps make it easier to read though.
Awesome. Here's the final (for now) then. Both versions are separate sheets in the xlsx.
I'll check on them and let you know.
Forgot to say the Iris 5100 is also affected. Maybe all Haswell iGPUs are? Sandy Bridge, Ivy Bridge and Broadwell iGPUs are not affected. Broadwell iGPUs are limited to a height of 2400 regardless of width.
I did some testing. Initially it seemed easy with one 62603 crapping out after 308.00 MHz and the other doing 334.56 MHz right away, but some switching back and forth between modes resulted in the bad one suddenly doing 336.00 MHz somewhat reliably and 337.00 MHz less reliably, i.e. if I got a picture in the first place, there were no dropouts during the few minutes I observed it (longer tests to come). 338.00 MHz or more was a no-go.
Last edited:
I've assembled all of the source documents and mirrored them on my page here:
I found the commonly available T221 manual was a bit mangled where text and images fit together, so I made some attempt to fix that. Most of it was redundant to the MD22292B version anyways, but now both are legible.
Also in this process I discovered the User Guide supplements for both the T220 and T221, the former of which mentions the Matrox card being different than retail (but the details aren't specified; maybe different firmware?) and the latter mentions unlocking additional settings on the OSD.
I found the commonly available T221 manual was a bit mangled where text and images fit together, so I made some attempt to fix that. Most of it was redundant to the MD22292B version anyways, but now both are legible.
Also in this process I discovered the User Guide supplements for both the T220 and T221, the former of which mentions the Matrox card being different than retail (but the details aren't specified; maybe different firmware?) and the latter mentions unlocking additional settings on the OSD.