Disabling multitexturing fixes the lighting problem with Unreal. I'm still searching for a fix for OpenGL.

The issue with Quake 3 was that it had been configured too high fot another card. Setting to to 640x480 and lower tender options made it playable.

OpenGL issues with GL Quake and Quake 2 were solved by using the OpenGL ICD from the previous driver release version 6.82.016.

I still have issues with the card sometimes using some kind of output that my LCD doesn't like on the VGA connector at some resolutions. This can often lead to the necessity of a hard reset, as I've lost all video and can't get it back with an alt-tab or ctrl-alt-del.

Overall, I would say that this card is "fine", once it has been tweaked and configured correctly. It needs a lot of hand holding to get everything right before it will behave. And even then you need to keep an eye on it, because it's a spoiled little child video card and and might throw a tantrum at any moment.

The 'Cannot Display This Video Mode'-error is likely caused by the video card trying to push a refresh rate that your monitor can't handle. Be sure to go over the settings for different resolutions in Windows' Display Properties and set the default (rather than optimal) refresh rate for different resolutions and color depths if you don't have a monitor driver installed and those are the only two options you get.

Quake probably needs to have some renderer options set via console (~) or in the default config, it wasn't that uncommon that you'd need special settings depending on your video card.

Unreal/Unreal Tournament also have quite terrible OpenGL out-of-box. For Unreal, I think OpenGL 2.1.0.6 was the one that worked on Win9x for TNT2 and older video cards with OpenGL). For UT, this is the one I think you want.

Fun fact - the first Direct3D renderer for Unreal was a beta made for the G200. 😁

D3D is probably the best way to play it on a G200.

UT needs the multi texturing disabled too.

Kahenraz wrote on 2022-06-12, 20:53:

Disabling multitexturing fixes the lighting problem with Unreal. I'm still searching for a fix for OpenGL.

That seems a bit odd as the G200 does not even support multi-texturing, AFAIK. Maybe it's a driver bug ?

I won't be trying anytime as my only G200 variant is integrated into my Dell r710 server which I need for other, non-gaming, uses .

There is a Matrox G200 integrated onboard my Socket G34 server motherboard as well, is currently in use, and has been for years. Running Linux of course, not Windows 98. 😀

https://www.supermicro.com/Aplus/motherboard/ … 6x0/H8DGi-F.cfm

Kahenraz wrote on 2022-06-12, 21:23:

OpenGL issues with GL Quake and Quake 2 we're solved by using the OpenGL ICD from the previous driver release version 6.82.016.

As far as I know the best OpenGL ICD for G200 is actually found in the G400 driver pack, funnily enough..

Yeah that G200 ICD included with the G400 drivers has some bug fixes and they never bothered to package it into a new set of G200 drivers. G200 was not really supported anymore.

batracio wrote on 2010-11-13, 07:27:

Many years ago, I posted about that G200 transparency bug on Matrox tech support forum, a bug report was raised and problem was eventually fixed, but oddly enough, Matrox only released the updated G200 OpenGL ICD inside G400 Windows XP driver. You can still download xp2k_596_004.exe archive, expand g200icd.dll file and use it under Windows 95, 98, 2K and XP with fixed transparency and other improvements.

I never would have thought to try using files from the XP version with Wibdows 98. I'll try this, thank you.

I tried the g200icd.dll ICD from the XP driver and it works great. I didn't ever see an issue with transparency, and didn't notice anything different after swapping it out. But I'm sure that I just didn't have a proper example to test with.

This card is definitely a solid pick for early DirectX (up to version 5), and the final OpenGL implementation is solid. It's a shame that it took Matrox so many years to get it all ironed out.

On a side note, the G200 is also the last PCI Matrox card to support Windows 3.1. while there is driver support for the G400, I've never seen one in PCI form.

I've read that the G450 is very similar to the G400. I wonder if it would work if the G400 drivers?

Quake 2's water is opaque without that OpenGL driver. One example. But that wasn't always a problem. It was a regression that happened at some point and nobody noticed I guess.

Does anyone know how/if the better 400 driver works with the server 'e' and 'ew' versions. I think they are 16MB, or so. I wonder if unused ones would be worth vfio passing to a vm?

That's a very good question. If it is possible to get it to pass through, it would be amazing, since there are drivers for Windows 3.1 and up, with very reasonable DirectX and OpenGL support.

The same question could be asked about server motherboard which have a Rage XL chip. Lots of good support there, although not as capable as the G200.

Kahenraz wrote on 2022-06-14, 02:16:

That's a very good question.

Your right, broad early support.
Ok, I'll try someday. In my rotation of use is my X8 XP, I think I used 5.82 for XP and leave it disabled in XP with a full card as primary. As the boot gpu, linux used it and passed the slotted gpu. Never tried reversing it...
I need some fiddle time...the OS graduated to this computer, and does not have IOMMU group listing it...ok let me see what I can do, it'll take awhile. I have record of the bus, G200eW WPCM450, and a record of the mode line for 1920x1080.
Interesting

What keeps me from earlier than XP vm's is the acpi or lack of that pegs a core for no reason...