A curious historical document that provides various examples of using DVI Dual-Link, for example:

As you can see, Dual-Link already worked in driver 53.03. Although at the beginning of this document, the driver version is mentioned as " 61.76 (IBM)". It turns out that it was some specific version of the driver, specially modified by IBM employees? Where can I download it today (I would like to see how it differs from the reference version from nVidia)? But in the 77.72 driver, support for this monitor has already been officially announced (It is possible that nVidia further implemented IBM employees ' developments in 77.72 or filed it itself). But I still believe more that the key to solving this whole story can be found in this very small box (Between the video card and the monitor), or rather in its EDID. Obviously, it is in it that something clearly indicates to the old video card that it is necessary to switch to Dual-Link mode (After all, the image should appear even before loading any driver, in normal conditions, even at the BIOS stage). If you take this EDID and modify it in a special way (For the actual capabilities of a modern monitor), you may get something useful (or maybe not).

DoZator wrote on 2024-04-30, 20:23:

What happens if you try driver 77.72 (Or earlier) for Windows XP? Will 144Hz work? Or do you need exactly "93.71" and higher?

By the way, the 300 series drivers for Windows XP support on-the-fly EDID redefinition for nVidia Quadro video cards, which greatly simplifies testing (this feature does not work on GeForce cards).

Newer drivers than version 84.43 dont support the 7 series.

The earliest driver I can test with a dual link card is 71.84 from 02/24/2005 with the Quadro fx 3450 pcie (late 6 series quadro).

I tested under XP today again a 7600gt agp and a Quadro fx 3450 pcie.

The 7600gt agp uses the earliest driver version 84.43 from 4/5/2006. It works at 144Hz. I tested it playing UT99 patch 469 and it works with vsync at 144fps. Earlier drivers don’t support the card. Unfortunately only 60Hz under 98 using the patched 81.98 drivers.

The 7800gtx uses the earlier 81.85 in xp from 12/10/2005, but i dont have that one.

But the problem is the fx3450: Only 60 Hz under 98, XP and 10.

I looked at the xp and 98 driver versions and they dont support always the same models.
For example there is no official support under 98 for the fx3450 with the latest 81.98 driver.
But under XP already the 71.84 driver supports it.

Under 10 with driver version 309, I tried CRU and Nvidia patcher, but only 60Hz was possible to select. Not possible to add custom resolutions, nvidia driver gives an error. Tried also the EDID override tool.

Under xp I can add 144hz using driver versions 71.98 and 84.43 in the Nvidia control panel and I can also select it and it seems to work. But when I play UT99 with vsync it’s only 60fps.
I tried d3d dx9 control panel refresh rate override to 120Hz (max) but still 60fps, also in OpenGL, also with refresh locker app.
Monitor also reports 60Hz. But control panel shows 144Hz. Very strange.
Using custom resolution in powerstrip i get only black screen.

It’s actually not easy to verify the refresh rate. For example testufo.com doesn’t work under XP, no matter what I tried with Firefox and chrome, the 7600gt only shows 60Hz, no idea why, but works at 144Hz in UT99. I guess the on screen display of the monitor tells the truth when it states 60Hz, I thought that was a bug.

The FX3450 has two dual link ports, it should work at 144Hz, so why doesn’t it work even under XP and 10?
Maybe I should try a later driver under XP. Like the version 185 you mentioned.
But it makes no sense because the 7600gt works perfect at 144hz in xp, why not the fx3450?

The fx3450 is a 6 series card (nv42gl), released before the 6800gs, xe and xt in July 2005.
Dual link was not integrated on the 6 series but only using an external chip.
Maybe the implementation of the external chip (TMDS) on the 6 series was not done correctly for high refresh ?

On the 7 series 7600gt dual dvi is integrated on the chip.
G70, G72, G73 models 7300 to 7800 have one integrated dual link port. 7800gtx released in June 2005, 7600 GT AGP in July 2006.
G71 models 7900 have two integrated dual link ports. 7900 GTX released in March 2006, 7900 GS AGP April in 2007.

I find it weird that the same xp driver version 84.43 supports dual link dvi high refresh only on the 7 series not on the 6 series.
It has to be the difference between the integrated and external TDMS chip.

This arcticle from 2004 also explains TDMS: https://www.tomshardware.com/reviews/tft-conn … tion,931-4.html

The fact that dual link high refresh support for the 6 series quadros wasnt even working in xp, makes me even more sceptical that there is a way under 98.

The problem is that dual link was introduced at the same time that 98 support stopped and didnt even work properly for high refresh until the 7 series in xp.

If you look at a gaming monitor from 2006, the viewsonic vx922 75Hz 1280x1024 2ms gaming monitor. It has only 75Hz. There were no monitors available that supported 144Hz at the time. So Nvidia couldn’t even test if the drivers work at full hd high refresh rates, because there were no monitors available. So I guess on many older cards like the 5 and 6 series the support for dvi dual link high refresh rates was just not there in the driver, because those external chips were only on a few Quadro cards and it wasn’t worth the work because nobody could use it anyway. Probably later drivers fix some of the problems for those older cards. I need to test that. But with the 7 series they integrated dual dvi on every card, so didn’t need to take care of external chips and I guess it was easy for them to implement it correctly for the whole series. At this point it started to matter, with better dvi tft displays becoming slowly available, and at the same time they dropped 98 support. So it seems that’s why dual link dvi high refresh was never possible under 98. But we haven’t tried everything yet, maybe it works on one of those other cards.

It’s a shame because it really makes a huge difference if you play UT99 at 60 or 144Hz. The fx3450 runs 1080p at over 300fps, the 7600gt even 600fps. So it’s really smooth at 144Hz. Of course we can play it under 10 or XP, but only under 98 we can have A3D, and I want both 😉

I wish I had a 2560x1440 dvi monitor so I could test those cards and drivers if they support only high resolutions but not high refresh.

You could just run Vsync locked benchmark, with control scores for both 60FPS results and 144FPS one.
Assuming GPU can do more than 144FPS, if you get higher number of the two control scores - 144Hz mode should work properly.

The problem has already been successfully resolved. More precisely, I was solving a slightly different problem, but this one was also solved along the way. Now it runs 1920x1080@144Hz without any complaints under Windows 98 with the ForceWare 77.72 driver on the G70. I checked it in games with Vsync enabled and framerate unlocked. Indeed, a full-fledged 144hz and FPS counter and smooth movement show this. What exactly was the problem-I did not find out (It is possible that initially there was not a compatible EDID, or the adjustments made somehow affected it). I wanted to make DOS games run in a different video mode by default via the DVI port. I downloaded the reference EDID from the DVI-DL port of this monitor on GitHub and slightly modified it for this purpose. I added all available "Established Timings", slightly changed the "Standard Timings" at my discretion, and also manually registered in "Detailed Data" -> "Preferred Timing Block" the standard data from the online EDID calculator for 1600x900@72hz mode (So that under DOS it would always be at least 72hz at any time game via DVI connection), the rest did not touch anything. After flashing the DVI-DL port, it became - under DOS, all games are now scaled to 1600x900@72hz (Before that it was the same, only up to 60hz), - and Windows now runs 1920x1080@(100\120\144hz), which confirms the presence of a functional Dual-Link. These modes are specified in the EDID, in the section "CEA Extension" -> "Detailed Timing Blocks". It turns out that the video bios or driver still handles this extended block. In Windows 98, I cleared the PowerStrip config (It made negative adjustments due to previous, no longer relevant settings), reinstalled the driver, added 100\120\144hz to the registry, and also enabled monitor scaling:

Jonsmith0815 wrote on 2024-05-21, 01:35:

But it makes no sense because the 7600gt works perfect at 144hz in xp, why not the fx3450?

If the video card doesn't support Dual-Link, it's likely that no driver can cure it. The video card can be tested in Windows XP with later drivers (185.20 and higher) that successfully handle even EDID curves (As was the case in my case). Also, under XP, it is possible to override the EDID on the fly, which allows you to check it without flashing it to the monitor (The function is available for Quadro video cards with drivers 257.21 or higher).

DoZator wrote on Yesterday, 09:26:

The problem has already been successfully resolved. More precisely, I was solving a slightly different problem, but this one was also solved along the way. Now it runs 1920x1080@144Hz without any complaints under Windows 98 with the ForceWare 77.72 driver on the G70. I checked it in games with Vsync enabled and framerate unlocked. Indeed, a full-fledged 144hz and FPS counter and smooth movement show this. What exactly was the problem-I did not find out (It is possible that initially there was not a compatible EDID, or the adjustments made somehow affected it). I wanted to make DOS games run in a different video mode by default via the DVI port. I downloaded the reference EDID from the DVI-DL port of this monitor on GitHub and slightly modified it for this purpose. I added all available "Established Timings", slightly changed the "Standard Timings" at my discretion, and also manually registered in "Detailed Data" -> "Preferred Timing Block" the standard data from the online EDID calculator for 1600x900@72hz mode (So that under DOS it would always be at least 72hz at any time game via DVI connection), the rest did not touch anything. After flashing the DVI-DL port, it became - under DOS, all games are now scaled to 1600x900@72hz (Before that it was the same, only up to 60hz), - and Windows now runs 1920x1080@(100\120\144hz), which confirms the presence of a functional Dual-Link. These modes are specified in the EDID, in the section "CEA Extension" -> "Detailed Timing Blocks". It turns out that the video bios or driver still handles this extended block. In Windows 98, I cleared the PowerStrip config (It made negative adjustments due to previous, no longer relevant settings), reinstalled the driver, added 100\120\144hz to the registry, and also enabled monitor scaling:

Awesome! That's so great that you could solve it!

Can you describe how you flashed the EDID of the XL2411?

I would like to flash my XL2411 as well 😉

Jonsmith0815 wrote on Yesterday, 12:04:

Awesome! That's so great that you could solve it! […]
Show full quote

DoZator wrote on Yesterday, 09:26:

The problem has already been successfully resolved. More precisely, I was solving a slightly different problem, but this one was also solved along the way. Now it runs 1920x1080@144Hz without any complaints under Windows 98 with the ForceWare 77.72 driver on the G70. I checked it in games with Vsync enabled and framerate unlocked. Indeed, a full-fledged 144hz and FPS counter and smooth movement show this. What exactly was the problem-I did not find out (It is possible that initially there was not a compatible EDID, or the adjustments made somehow affected it). I wanted to make DOS games run in a different video mode by default via the DVI port. I downloaded the reference EDID from the DVI-DL port of this monitor on GitHub and slightly modified it for this purpose. I added all available "Established Timings", slightly changed the "Standard Timings" at my discretion, and also manually registered in "Detailed Data" -> "Preferred Timing Block" the standard data from the online EDID calculator for 1600x900@72hz mode (So that under DOS it would always be at least 72hz at any time game via DVI connection), the rest did not touch anything. After flashing the DVI-DL port, it became - under DOS, all games are now scaled to 1600x900@72hz (Before that it was the same, only up to 60hz), - and Windows now runs 1920x1080@(100\120\144hz), which confirms the presence of a functional Dual-Link. These modes are specified in the EDID, in the section "CEA Extension" -> "Detailed Timing Blocks". It turns out that the video bios or driver still handles this extended block. In Windows 98, I cleared the PowerStrip config (It made negative adjustments due to previous, no longer relevant settings), reinstalled the driver, added 100\120\144hz to the registry, and also enabled monitor scaling:

Awesome! That's so great that you could solve it!

Can you describe how you flashed the EDID of the XL2411?

I would like to flash my XL2411 as well 😉

As for flashing, taking into account all the risks, I remember how at the beginning of this (Or somewhere in the next) topic they suggested using the "EDID-EMULATOR"-a small programmable device connected between the monitor and the video card, which allows you to easily and simply switch several pre - prepared EDID overlap profiles. Something like this:

https://www.startech.com/en-us/audio-video-pr … ducts/vsediddvi

The main thing is to first check that Dual-Link is supported and all modes are transmitted correctly. I haven't tried anything like this yet, but even, for example, a simple KVM-Switch passed only a Single-Link through itself, so you need to be more careful when choosing such things (The fact that the link above is only for an example of how it may look in principle, but in no case is there a guarantee that it is at least like- then it will work and even more so support Dual-Link).

Nevertheless, in case of confirmation of full functionality, one of these pieces seems to be the most interesting and preferable solution, since it would allow you to easily and quickly change any pre-prepared modes, which is useful, especially in MS-DOS.

Regarding the firmware of the monitor itself, I used EDID-RW, following the instructions from GitHub (a pretty good step-by-step guide). In addition, a nuance surfaced - at the last step (Before entering the firmware command), you need to turn off the monitor. After that, just press "ENTER" and wait a little while until the firmware process is completely completed there and only then turn it on. In the monitor's service menu, the "Burn-in" option was previously enabled (It seems that it is also needed).

Previously, I tried flashing through the old version of PowerStrip under Win9x, but I didn't succeed then (Although at that time I still didn't know about the nuance with the need to turn off the monitor, maybe that's why). In the future, I didn't try using PowerStrip (I need the old unlocked version there, but it also turned out to be so buggy, besides. Moreover, there was already a ready-made and proven firmware tool at hand at that time).

darry wrote on 2022-10-25, 18:15:

That 2560x1600@60 limitation over dual-link DVI is often mentioned in DOCs [1] and ads, but detailed exact timings are not menti […]
Show full quote

The Serpent Rider wrote on 2022-10-25, 17:06:

In the first message, I indicated that when a modern video card is connected, DVI Dual-Link works in full. That is 1920*1080@144hz.

AFAIK Nvidia DVI connector of that era can handle 2048x1536@85Hz or anything that can fit into that bandwidth range (2560x1600@60 as mentioned above), but no more than that.

That 2560x1600@60 limitation over dual-link DVI is often mentioned in DOCs [1] and ads, but detailed exact timings are not mentioned (at least I haven't seen them), but I'm guessing at CVT-RB, so maybe 268MHz or so as calculated by [2] .

Maybe the TMDS and accompanying logic in the the 7950GT (and other similar GPUs) are either limited to 268-ish MHz and/or aren't flexible enough to handle what would have been, at the time, oddball timings like 1920x1080@144Hz .

OP could try setting up custom resolutions/timings with reduced blanking to test (trial and error) how far things can be taken. This is assuming the Nvidia drivers of the time allow this or that a third party tool like Powerstrip is used .

EDIT : Maybe the old Nvidia driver is trying to push wacky timings with huge blanking intervals that exceed even 330MHz by default .

[1]
https://www.evga.com/products/specs/gpu.aspx? … 6D-E9455529E242

[2]
https://tomverbeure.github.io/video_timings_calculator

I looked at EDID and for 1920x1080@100hz and 1920x1080@120hz it uses the classic CVT-RB, which fully corresponds to the values that the EDID calculator calculates, but for 1920x1080@144hz the values are very close to CVT-RB, but still slightly different from everything that the calculator has. Nevertheless, the nVidia driver for Windows 98 normally now accepts both the usual CVT-RB and even this custom mode from the monitor manufacturer. In the dual channel mode of operation under 9x it still remains in the range of 330MHz:

Thank you for the detailed explanation!

I will try edid-rw as well.

Could you copy and paste the contents of the EDID file you flashed?

I might not flash the same file, but use it only for reference when editing my EDID file.

@DoZator

Kudos for you persistence in researching/testing, documenting and sharing a solution.

Jonsmith0815 wrote on Today, 11:28:

Thank you for the detailed explanation! […]
Show full quote

Thank you for the detailed explanation!

I will try edid-rw as well.

Could you copy and paste the contents of the EDID file you flashed?

I might not flash the same file, but use it only for reference when editing my EDID file.

Yes, of course, but I'm still experimenting with refining the EDID and have recently made and already switched to a newer version of "EDIDM5.bin".

All changes (relative to the reference version) in "EDIDM4.bin":
* In the "Established Timings" section
- All existing modes are added

* In the "Standard Timings" section
- Replaced:
1920x16:9@60Hz -> 1600x16:9@100Hz
800x4:3@120Hz -> 800x4:3@100Hz
1024x4:3@120Hz -> 1024x4:3@100Hz
1280x5:4@60Hz -> 960x16:9@100Hz
1280x5:4@120Hz -> 1280x16:9@100Hz
1440x16:10@120Hz -> 1440x16:10@100Hz
640x4:3@120Hz -> 640x4:3@100Hz

- Added:
720x5:4@100Hz

* For DOS (Section "Detailed Data" - > "Preferred Timing Block")
- Replaced:
1920x1080@60hz [CEA-861\DMT] -> 1600x900@72hz [CVT] ***

* For Windows (Section "CEA Extension" - > "Detailed Timing Blocks")
- No changes

What has changed (compared to "EDIDM4. bin") in "EDIDM5. bin":

* In the section" Established Timings "
- Removed some exotic modes:
-- 640x480@67Hz [Apple, Mac II]
-- 832x624@75Hz [Apple, Mac II]
-- 1024x768@87Hz [IBM - Interlaced]
-- 1152x870@75Hz [Apple, Mac II]

* In the "Standard Timings" section
- Changed the "Refresh Rate" for all existing modes to standard 60Hz

* For DOS (Section "Detailed Data" - > "Preferred Timing Block")
- No changes (?)*

* For Windows (Section "CEA Extension" - > "Detailed Timing Blocks")
- Added new modes (For native support)**
-- 960x540@60Hz [CVT-RB]
-- 960x540@100Hz [CVT-RB]
-- 1600x900@100Hz [CVT-RB]

*** For more information, visit https://tomverbeure.github.io/video_timings_calculator

** Now in Windows, three added new modes work natively (As is), without forced scaling by the video adapter to 1920x1080 (As was the case in the previous version of "EDIDM4").

* Major changes were made in the previous version (Described above). This section has not yet been discussed in more detail. It is planned to test a few more ideas regarding the use of multiple video modes under DOS in the future. I came across several EDIDs from other monitors that use not one, but several blocks at once in this section, with different values in each of them. If everything works out, it may be possible to change modes in DOS (If the video bios understands this, although, in theory, it should, because the Dual-Link monitors from Apple, in which I saw this, were released just around the same time). In addition, I am interested in the question of why the "Established Timings" and "Standard Timings"sections are ignored under DOS at all. Instead, go straight to "Detailed Data" -> "Preferred Timing Block" and take the mode from there (Including the refresh rate). And then any set video mode is converted to the one that was taken earlier from "Detailed Data" - > "Preferred Timing Block". It's not very clear yet.

What a cherry on the cake: