I'm interested in true 4k (3840×2160), is 2560x1440 your LCD native and only trying get higher refresh? Your LCD dosn't have DP input?

RayeR wrote on 2023-06-13, 11:24:

I'm interested in true 4k (3840×2160), is 2560x1440 your LCD native and only trying get higher refresh? Your LCD dosn't have DP input?

I have tried DP but it is not working now. It worked at first and I entered 165hz and it blacked out and I can't get it to work again. I even changed the drivers. I have ordered some more cables to try different combinations so I can see if the cable quit working. I don't think it will go much higher since that article stopped at 50hz at 4k without hex editing. 75 might be as high as it will go.

I have currently access only to FHD 1920x1200/60Hz monitor with DP input and it works normally under XP on GTX 970, I use drivers version 355.98. I hope I'll have a chance to test 4k LCD in future...

^Source : LINK

^Source : LINK

I would be OK with 3840x2160 33/41Hz if it is really approved to work. Still wouldn't possible to unlock HBR2 mode? One idea - let Linux initialize it via newer driver and then do some kind of soft-reboot/chainload to WinXP?

quote=agent_x007 post_id=1170679 time=1686724858 user_id=32631]
NV Unsupported WinXP.png
^Source : LINK
DP WinXP NV.png
^Source : LINK
[/quote]

So 2560x1440 should be able to do 90 fps then. I'll have to keep testing.

I'm not sure how ATI/AMD cards perform but I recall seeing reports about those video cards (if supported DP1.2) having working HBR2 on XP thus can show higher resolutions. It was said that ATI/AMD used a different approach than nVidia to achieve that.

For nVidia, however, you're limited to DP1.1 resolutions, which is the maximum output supported by Fermi (the last nVidia generation with a sane VGA BIOS).

Note that for some nVidia cards and drivers, the default 1080p/120Hz is not working correctly. You can, however, achieve proper 1080p 120Hz/144Hz through creating a custom resolution using CVTRB/CVTRBv2 timings. 1440p 60Hz (not sure about higher) should also work (again, CVTRB/CVTRBv2 timings are recommended).

PS: My personal experience is, that while nVidia drivers are aware of your monitor and video card's maximum capabilities, attempting to use resolutions requiring HBR2 will lead to a blank screen. This has caused a lot of headache with using those cards on XP as after installing the driver, on next boot the driver will attempt to use the maximum possible resolution (in case with DP1.2 capable cards, 4K/60Hz) which will not work, and I have to go to safe mode to edit the respective registry entries to reduce the resolution setting to a point that the system can at least show the desktop, and start from there.

Yes, I have found an article about AMD GPU that have DP 1.2 (HBR2) working under WinXP...
Somewhere was noted that nvidia driver rely on some MS library from windows that is needed for HBR2 to work - this is present only in Win7 and newer. Do we know what files/system functions exactly are needed? As AMD was able to write own implementation to bypass missing MS library it should be possible for nvidia too.
BTW I found a report from Linux user that HBR2 is working on GTX960 with nvidia Linux driver 361, this is a bit older version that the latest WinXP drivers. I belived that nvidia has some kind of unified driver architecture that most of codebase is common for Windows and Linux and only some smaller part of code is the specific kernel interface. This would mean that driver 361 could already implement HBR2 for all platforms but as WinXP missing some support API it's just disabled then. It would also mean there should be some Linux kernel or Xorg API that is needed to support HBR2 on Linux. I cannot find some reports for same nvidia driver version running on different kernels if it does a change in HBR2 functionality (if there's some minimal kernel/Xorg version needed support HBR2) or if kernel version doesn't matter. If yes, then it could be possible to track what feature in Linux kernel was added to enable HBR2 and maybe then ported to WinXP...
I also cannot find any confirmation that HBR2 works with Linux nouveau opensource driver.

EDIT: it seems that under Linux the DP code is inside binary driver, it was reported that newer drivers than 450.119.03 switched from HBR2 to HBR3 regardless on other system components but it doesn't imply that the driver doesn't need any support code from kernel.

LSS10999 wrote on 2023-06-20, 06:27:

I'm not sure how ATI/AMD cards perform but I recall seeing reports about those video cards (if supported DP1.2) having working H […]
Show full quote

I'm not sure how ATI/AMD cards perform but I recall seeing reports about those video cards (if supported DP1.2) having working HBR2 on XP thus can show higher resolutions. It was said that ATI/AMD used a different approach than nVidia to achieve that.

For nVidia, however, you're limited to DP1.1 resolutions, which is the maximum output supported by Fermi (the last nVidia generation with a sane VGA BIOS).

Note that for some nVidia cards and drivers, the default 1080p/120Hz is not working correctly. You can, however, achieve proper 1080p 120Hz/144Hz through creating a custom resolution using CVTRB/CVTRBv2 timings. 1440p 60Hz (not sure about higher) should also work (again, CVTRB/CVTRBv2 timings are recommended).

PS: My personal experience is, that while nVidia drivers are aware of your monitor and video card's maximum capabilities, attempting to use resolutions requiring HBR2 will lead to a blank screen. This has caused a lot of headache with using those cards on XP as after installing the driver, on next boot the driver will attempt to use the maximum possible resolution (in case with DP1.2 capable cards, 4K/60Hz) which will not work, and I have to go to safe mode to edit the respective registry entries to reduce the resolution setting to a point that the system can at least show the desktop, and start from there.

What are the registry entries? I messed up a couple of installs and ended up using DP to HDMI adapter which bypassed it.

shaq wrote on 2023-06-23, 08:39:

LSS10999 wrote on 2023-06-20, 06:27:

I'm not sure how ATI/AMD cards perform but I recall seeing reports about those video cards (if supported DP1.2) having working H […]
Show full quote

I'm not sure how ATI/AMD cards perform but I recall seeing reports about those video cards (if supported DP1.2) having working HBR2 on XP thus can show higher resolutions. It was said that ATI/AMD used a different approach than nVidia to achieve that.

For nVidia, however, you're limited to DP1.1 resolutions, which is the maximum output supported by Fermi (the last nVidia generation with a sane VGA BIOS).

Note that for some nVidia cards and drivers, the default 1080p/120Hz is not working correctly. You can, however, achieve proper 1080p 120Hz/144Hz through creating a custom resolution using CVTRB/CVTRBv2 timings. 1440p 60Hz (not sure about higher) should also work (again, CVTRB/CVTRBv2 timings are recommended).

PS: My personal experience is, that while nVidia drivers are aware of your monitor and video card's maximum capabilities, attempting to use resolutions requiring HBR2 will lead to a blank screen. This has caused a lot of headache with using those cards on XP as after installing the driver, on next boot the driver will attempt to use the maximum possible resolution (in case with DP1.2 capable cards, 4K/60Hz) which will not work, and I have to go to safe mode to edit the respective registry entries to reduce the resolution setting to a point that the system can at least show the desktop, and start from there.

What are the registry entries? I messed up a couple of installs and ended up using DP to HDMI adapter which bypassed it.

I think it's in

HKEY_CURRENT_CONFIG\System\CurrentControlSet\Control\Video

Note that there are more than one set of monitor settings. You need to actually find the one corresponding to the nVidia video driver. Usually the one with a high, widescreen resolution (depending on your monitor), and more contents than others. Other sets were intended for the default VGA driver and should not be touched under normal circumstances.

Change the right one from Safe Mode to a value that's within HBR range and reboot to normal mode. If done correctly, the system should be able to boot and you should be able to see the desktop using the resolution you have put in the registry.

This was posted on MSFN which was grabbed from wp.xin.at
https://msfn.org/board/topic/184219-solved-ge … n-lcd-monitors/
http://wp.xin.at/archives/5616

This is for increasing the refresh rate by modifying the driver. Supposedly works for both HDMI and DVI.
May be worth it to make a list of the must have drivers for 2000/XP for Nvidia and see how difficult it would be to patch.

Hello, thanks for your information about the version of nVIDIA drivers. […]
Show full quote

Hello, thanks for your information about the version of nVIDIA drivers.

I just wanna display as WQHD@60Hz by GT710(10DE:128B), I found and modified the upper limit codes of pixel clock via HDMI in nVIDIA drivers for Windows XP.

like this (in v355.98):

0.Strip the signature.(Just in case)
nv4_mini.sys(32bit)
000001A0: 00A14BC0H -> 00000000H
000001A4: 00001E78H -> 00000000H

nv4_mini.sys(64bit)
000001A0: 00859700H -> 00000000H
000001A4: 00001E78H -> 00000000H

And, in the case of 32bit, remove from 00A14BC0 to the end, in the case of 64bit, remove 00859700 to the end of file.
(For now, the CheckSum turns 00A15BC0H in 32bit, 0085A700H in 64bit.)

1.Increase the limit values up to HDMI 1.4.
nv4_disp.dll(32bit) 00031C9D:
nv4_mini.sys(32bit) 004E402C:
nv4_disp.dll(64bit) 0004B1D2:
nv4_mini.sys(64bit) 001CA9A3 and 001CA9AE:
00028488H -> 00053020H (165000 -> 340000[KHz])

(By the way, nearby jump-code seems by me like crash one will never run.)

2.Apply the CheckSum.
nv4_disp.dll(32bit)
00000150: 00356780H -> 0036022BH
nv4_mini.sys(32bit)
00000160: 00A193ECH -> 00A2075AH

nv4_disp.dll(64bit)
000002C8: 00BA2A80H -> 00BAD61AH
nv4_mini.sys(64bit)
00000150: 0085D87EH -> 0085ED46H

I haven’t tested other resolutions over HDMI 1.4 like 4K or 8K and so on, so I tested only the pixel clock for WQHD@60Hz in 32bit mode.(I haven’t in 64bit mode.)
I also haven’t investigated precisery that has any other limits include via DVI-D.

Just for reference.

Reply
thrawn
thrawn says:
November 25, 2022 at 00:26
Hello Yas,

Awesome! Thank you for sharing! 😀

How did you even determine all that? Did you disassemble the driver (Ada?) and analyze the code by yourself? If yes, then you have my respect! 😀 If no, you still have my thanks! 😎

I was so fixated on the HBR2 limitations of DisplayPort that I hadn’t even considered that all you’d have to do for the HDMI side of things was to boost the pixel clock while staying on TMDS!

Still interesting that it’s enough for 3840×1440 @ 60 Hz though. According to Tom Verbeure’s calculator it’s just barely out of range even when using CVT-RBv2.

Reply

Yas says:
June 21, 2023 at 17:06
Hello.

Yes, by myself. I used dumpbin in MSVC and a hex editor, WinPE for overwriting, and stuff.
So simply, finding the value with my intuitions, and checking the disassembled codes, also verifying the behaviors, that’s all.

That modified value has just adjusted for safety to the public spec that works certainly on Windows 7 or later.
It doesn’t mean a limit that the GPU has, so I think it will works well if the value is less than really limit, but no any guarantee.
Ofcouse it needs recalculated the CheckSum. (by binutils codes, or some kind of tools, including MapFileAndCheckSum() function in Win32API)

And, I have new some reports.

First, installed GT710(10DE:128B) by modified inf was wrong with CUDA effect,
so I soon turned GT730(10DE:1287), has also worked completely with that patched limit.

Next, that immediate value in nv4_mini.sys is assigned into two variables at the same time (is divided in 64bit), it may imply that has two limitations, because it has had an effect not only via HDMI but also via Single Link DVI-D(Out of standards).
It might be divided by type of the port, or the common limitation for Single Link TMDS might be each assigned to the port.

My monitor has three HDMI, all ports keep each aspect ratio, to take advantage of those situations, I’ve been using to switch between the passive DVI-D SL to HDMI cable as 4:3(1920×1440 or less) and the HDMI cable as 16:9(2560×1440 or less).

Another in nv4_disp.dll, that value seems a limit for rejection at system boot.

Reply

MistaH says:
October 14, 2023 at 15:14
Hi

Sorry to be a pest.. But do you have a hex edited version of Nvidia’s 355.98 graphics drivers for XP 64, I can download from anywhere?

Reply
thrawn
thrawn says:
October 16, 2023 at 09:51
Hello MistaH,

Just in case Yas doesn’t reply, I attempted to apply his hack to the NVIDIA GeForce 355.98 driver for XP x64.

Before you do anything with this, please understand that this is completely untested and poses according risks.

Get the files (unpack with [7-Zip]):

[HDMI-hacked 355.98 driver file components]
Workflow: Install the regular NVIDIA 355.95 driver on XP x64 if not done already, and make sure to use this exact version! Boot into safe mode or alternatively, boot a Linux live distro, then backup (!) and replace the files %WINDIR%\system32\drivers\nv4_mini.sys and %WINDIR%\system32\nv4_disp.dll with the ones from the archive. Finally, pray to all the demons of the nine hells and then reboot.

If I made just the tiniest mistake, this may give you BSOD loops or all kinds of other horrible issues, because as said, it’s untested. If stuff breaks, boot from a live Linux or into safe mode again and restore the files from your backup.

Sadly, I can’t test this myself, as I currently don’t have a machine with an NVIDIA card that can run 355.98.

Reply

MistaH says:
October 20, 2023 at 15:33
Thrawn THANK YOU!

My ultimate goal.. Is to be able to get a single 4k display running at 60Hz.. Which is why a purchased a used (though very minty) FirePro W8000 as I heard these cards will work.. On a display with a native display port input etc..

THIS however.. Turned out to be a BIG mistake..

I’ve owned ATI/AMD graphics hardware before.. And while there hardware always been “fine on paper” it was always there damn drivers.. That gave me issues.. I’d end up going through several crappy sets of there Catalyst drivers.. Before I’d finally get one that seemed to work.. So I’d end up stuck on that particular version for a while.. Before I’d bother upgrading to a newer version etc..

Now roll on a few years.. And it seems.. NOTHING has really changed.. The “quality” of there FirePro V8.982 drivers could only be rather bluntly descried as.. “dog shit” in which hardly ANYTHING seems to work in them.. Apps that use OpenGL.. Perform TERRIBLY.. And crash after a few seconds of running.. And while stuff that uses D3D seems to be fine.. They’ll only run for a few minutes at a time.. Before the whole system hard locks.. And I have to reach over for the reset button.. Once more etc..

So I need a graphics card.. That ISN’T just acting as a “bitch basic” frame buffer etc.. I’m not sure if purchasing a consumer grade Radeon card.. Might be a better idea though.. As it would grant me the use of the slightly newer Catalyst 14.4 drivers.. Which given the two years between them and the older FirePro V8.982 drivers.. I’d at least HOPE.. They’d used that time to “iron out” most of the major issues with them etc..

In the mean time though.. I’ve gone crawling back to green company of “envy” run by the “mob boss” of graphics card business.. As my trusty old Quadro M6000.. Just.. Works.. Even though I’m stuck with a maximum resolution of 2560×1440 with the current drivers I’m running on..

I may see IF I can purchase a consumer grade Radeon card in the near future.. Funds permitting of course.. Though in the mean time.. IF that doesn’t work out.. I guess I’ll just have to “stomachache” 4k at 30.. With the Quadro.. As their drivers.. Just.. Work.. With no bullshit.. No issues.. Etc.

Reply
thrawn
thrawn says:
October 23, 2023 at 14:41
Don’t say “thank you” yet! 😉 I still don’t know if I really did it right.

The most confusing part was that the endianness / byte order appeared to be inverted. The sequences Yas posted were all there but in reverse order. Or maybe it’s due to right-to-left reading direction? I hope I didn’t do anything wrong when I followed that exact reverse order, writing the bytes in reverse as well.

Also: Don’t get your hopes up too much about those older Catalyst drivers. They are pretty crappy as well. I mean it’s not as bad as you describe it, but I have some graphical glitches in a truly simple OpenGL 2.1 game: “Into the Breach”. It’s annoying and wasn’t there on my GTX Titan Black. I’ve never before seen that game glitch on any card, so there you go. I didn’t really test other games though, I only wanted the Radeon R9 280X for the higher refresh rates it provides.

After all, Windows XP’s GDI+ UI looks pretty bad at low refresh rates, with all that screen tearing going on when you move windows around with no VSync support being present… That was seriously no fun at 40 Hz on the GeForce.

Re

Holly Molly, I hoped that the nvidioted HBR2 limitation could be some stupid constant like this that could be easily patched instead of need implementing a new code...

Well, from that I readed on MSFN, it seems this patch unlock HDMI output only, not the DisplayPort, well?
Did/can someone try this if it will work also for 4k resolution @60Hz? They did it because wanting 144Hz in some lower res. but 4k seems to be untested yet. I would like to try this but I have access to 4k monitor only at work and my GTX970 is in home PC and both things are not small enought to take it in a pocket or bag 😀

Hey, I got some partial success with hacked drivers (Yas patch) but more hacking is needed to unlock true 4k (3840x2160).

I had to pass throuh some obstacles. I just brought a testing MB with my GTX970,.SSD and PSU at work to make test setup beside 4k LCD. I had to do a fresh install of WinXP SP3 yesterday because cloned system from very similar MB hanged at boot when logon screen appeared. Safe mode too. First problem was that I forgot to install USB HID drivers because I use PS/2 KB&mouse at home (no USB legacy feature in BIOS doesn't work in XP). Damn, nobody at work had legacy KB. Fortunatelly old MS IntelliMouse Optical saved my ass. I know it works with simple USB/PS2 passive adapter so I got an USB connector and 4 wires and made one on the fly. Then I used mouse to install HID drivers. Well, test setup was ready. 😀

1st I tried the DP output. What really surprised me is that when I booted to DOS the OSD on LCD reports it's running at 3840x2160@60Hz upscaling the textmode. Even it offers this resolution as VESA mode for 8 and 16bpp (not 32bpp). So I have 4k gfx mode under DOS-cool! I then tried WinNT4 with installed VBEMP VESA driver, it booted at 1600x1200/32bpp and offerd the 4k mode but when I tried to switch it crashed to BSOD. Tried to change some options but no way...

Then I booted to WinXP. Max listed mode was 3840x2160@60Hz but no 30Hz option and when I set it monitor went black saying no signal goes from DP. I tried to.make custom mode with 30Hz but failed immediatelly at testing, monitor even not recognized any change. So I have switch to lower mode 2560x1440@60Hz that works. Then I googled a solution that it is needed to switch off the multistream or switch from DP 1.2 to 1.1. On this Dell monitor it's a bit hidded. You have to press menu button for 10s at input selector menu then the option appears. After that I got 3840x2160@30Hz as highest mode and it works as expected. But we want more 😀

As I understand the Yas hack doesn't affect the DP so nothing more to be done here. So I switched to HDMI output. Unfortunatelly under DOS the support for 4k and 1600x1200 mode was droped and I got only 1280x 1024 😒. Under WinXP it was a similar story as with DP 1.2. 3840x2160@60Hz was listes but didn't work, again it maxed to 2560x1440. I tried the same trick in OSD and I got option to switch from HDMI 2.0 to 1.4, so I did. But damn I still couldn't get even 3840x2160@30Hz working, monitor complained that mode is invalid and shut down. Then I tried to make a custom mode again but it refused 3840x2160 with any timing. After some trial-error I found that lowering the Y-res helps. I got success with up to 1600p. I can increase refresh rate even up to 89Hz with reduced blanking - I hit 600MHz pixel clock! OSD info confirmed it really runs at 89Hz. So at least the bandwitdth limit is fixed by the patch. I belive that 1600p limit is artifical and needs to be pached too. I'd like to contact Yas because he already disassembled the code around this and he would probably find it more easily than me... At least I could try to search for "1600" 32bit constant close to patched area. Just walking through cold winter night with my GTX970 in my bag to put it back to home PC :p

Crosslink to another WXP 4k thread: How to get 4k on WinXP ?

Anyone has contact to Yas?

I got this patch to work with 355.98, but I'd like to try some older driver versions (specifically 285.58, 301.42 and 314.22) for compatibility. I was actually able to get it to partially work in multiple different older drivers, but there are problems. For one, any version prior has 3 total bytes containing the 88840200 byte in the nv4_disp.dll that you need to change to 20300500. However, while changing the resolution to 2560x1440@60Hz works, modifying the nv4_disp.dll no matter which variation of bytes changed (I tried them all) causes a BSOD or a game to not launch when trying to launch a 3D game. So, I tried only modifying the nv4_mini.sys driver only while leaving the nv4_disp.dll stock, and it actually lets you change the resolution and launch a game with that resolution, but when you reboot it causes a BSOD and you have to go into safe mode and reverse it. So patching the nv4_mini.sys works, but keeping the resolution and booting into it doesn't. I'm not sure what I need to do to make it work like the 355.98 patch where you can set the resolution permanently.