Has there ever been an unofficial windows 98 PCI-E patch for GPU cards? If not... is it possible for someone on GitHub to possibly develop one?

If you mean some unofficial 3D Drivers, its probably hardest task as you can get.. and nobody yet done it.. even Vmware Windows 98 driver is 2D only, Virtualbox had some very basic experimental Direct3D patch in the past.. if remember correctly but it wasnt good.. There is only some small hack to last Win98 Nvidia driver to support, Geforce 7 cards.. instead Geforce 6 and they are very similar.
There are only VBE9x 2D drivers which are working on newer cards, but they have lots of bugs too. Same people which are doing Linux gpu drivers could probably do it, but why they would do it for such small retro community. Well there is always chance, maybe it would be in 5,10,15 years.. and at the start compatibility would not be great. Nice project for Google decade of code.

The Nvidia GeForce 6-series/7-series and ATI (now AMD) x-series (x300, x600,x800, etc.) of PCI-E video cards have Windows 98 drivers. The Nvidia drivers also work with Windows 95 as long as you don't need the control panel apps.

I really like my Radeon x800 XL with Windows 98. Very stable with excellent DX8 and DX9 performance.

You need to modify the GeForce driver's .INF file to identify the PCI ID of the specific PCI-E card. These drivers, unfortunately, have a lot of instability in Windows 98, though. They're far more stable with Windows 95 (probably because the control panel apps don't work).

The Matrox G550 PCI-E card also works with Windows 98 using the Matrox drivers, but its DX performance won't impress anyone.

The Matrox G550 PCI-E card also works with Windows 98 using the Matrox drivers

Huh, it actually works in Win9x environment? If so, that's Riva TNT2 level card (with similar feature set), which may come in handy.

The Serpent Rider wrote on 2021-05-29, 15:51:

The Matrox G550 PCI-E card also works with Windows 98 using the Matrox drivers

Huh, it actually works in Win9x environment? If so, that's Riva TNT2 level card (with similar feature set), which may come in handy.

Yup. Here are the drivers:

https://www.matrox.com/fr/video/apps/drivers/ … les/w9x_683.php

I used these with the PCI-E version of the card. I don't remember if I had to edit the .INF file, though.

Update: I just tried to install the Matrox driver on Windows Me. I did have to modify G550.INF to include the PCI Subsystem ID for my PCI-E card.

Update 2: You also need to disable "Use bus mastering" in the Matrox Graphics Options tab to prevent random freezes. Here's a screenshot from Windows 98 SE:

What set of drivers and OS updates should be used with PCI-E cards under Windows 98 for best results?

What results do you get for example in the Tom2D benchmark?

Do I need to forcibly enable PCIMP.PCI (For AGP cards, for example, it is used, significantly increasing performance.) Or maybe there is some analogue?

Has Microsoft (or third-party developers) released any updates to implement PCI-E support from the OS?

And another question, I see someone here uses ATI cards, tell me, do you also show the absence of "AGP Support" in the dxgiag properties?

PCI-E works in Windows 98 without any special OS drivers because the OS thinks the device is on a 32-bit PCI bus. You will need drivers for the device (video card, USB, etc.) to get it working.

I made a screen capture of the display properties of my Radeon X800 XL in Windows 98 SE. The AGP Texture entry in the DXDIAG.EXE shows "Enabled", but I don't know if that has any significance for the PCI-E drivers. The X800 is a native PCI-E device, so I suppose ATI wrote the driver with the intention to interface over PCI-E instead of AGP.

fosterwj03 wrote on 2021-12-28, 03:02:

PCI-E works in Windows 98 without any special OS drivers because the OS thinks the device is on a 32-bit PCI bus. You will need drivers for the device (video card, USB, etc.) to get it working.

As well as AGP video cards by default, without a chipset of the GART driver (For Intel, this is PCIMP.PCI), with all the ensuing consequences for performance (Working in PCI mode). To use a PCI-E video card fully, there must be support from the operating system, otherwise it works like a regular PCI card on a very fast PCI bus.

Maybe there is some kind of patch or driver (PCIMP.PCI analog) for Win9x, which allows you to get the same 2D speed as with a similar AGP card.

What results of the Tom2D benchmark are the norm for PCI-E video cards under Win9x?

With AGP video cards, I get 15000 byls, and with PCI-E cards (I tried different nVidia cards with different drivers) about 5000, that is, three times less! I thought, maybe I'm just missing something in the system, some kind of patch is needed, a driver or a setting. Climbed into DXDIAG and saw "AGP Support: Not Available". It writes in the same way when the GART driver is not installed on the chipset with the AGP card (PCI mode). I thought there must be some connection here.

Is it something I'm missing in the system, or is this the norm and you have exactly the same results with PCI-E in Tom2D?

Probably because PCI-E is running in PCI mode -133 MB/s share for all device include discs, although some very modern MB has multiple parallel PCI buses to improve that.. AGP is up to 2048 MB/s (AGP 8x) and there even some better AGP specific code which give you than more much more brute force.

ruthan wrote on 2021-12-28, 18:57:

Probably because PCI-E is running in PCI mode -133 MB/s share for all device include discs, although some very modern MB has multiple parallel PCI buses to improve that.. AGP is up to 2048 MB/s (AGP 8x) and there even some better AGP specific code which give you than more much more brute force.

Then the games would run slowly. But this is not so. Check out the 3DMARKs results on the first pages. The games are fine.

fosterwj03 wrote on 2021-12-28, 03:43:

I made a screen capture of the display properties of my Radeon X800 XL in Windows 98 SE. The AGP Texture entry in the DXDIAG.EXE shows "Enabled", but I don't know if that has any significance for the PCI-E drivers. The X800 is a native PCI-E device, so I suppose ATI wrote the driver with the intention to interface over PCI-E instead of AGP.

Is the Radeon X800 XL connected to the PCI-E connector on the motherboard or to the AGP?
If to PCI-E, then AMD should be fine with this. Thanks for the information.

Then the games would run slowly. But this is not so. Check out the 3DMARKs results on the first pages. The games are fine.

Not really because some synthetic benchmark could be a lot about bandwidth games are not.. if you will search benchmark AGP vs PCI for games, difference is not huge.

DoZator wrote on 2021-12-28, 21:49:

fosterwj03 wrote on 2021-12-28, 03:43:

I made a screen capture of the display properties of my Radeon X800 XL in Windows 98 SE. The AGP Texture entry in the DXDIAG.EXE shows "Enabled", but I don't know if that has any significance for the PCI-E drivers. The X800 is a native PCI-E device, so I suppose ATI wrote the driver with the intention to interface over PCI-E instead of AGP.

Is the Radeon X800 XL connected to the PCI-E connector on the motherboard or to the AGP?
If to PCI-E, then AMD should be fine with this. Thanks for the information.

I have the x800 in a PCI-E x16 slot.

It runs Call of Duty at 1080p 60Hz without any stutters. I haven't done a benchmark on it, though. If I have time tonight, I'll try to benchmark it. It might have to wait until tomorrow.

Edit: I'm also running it on an Intel B75 Motherboard with a Core i5-3570.

I couldn't find an active link to Tom2D, but I ran 3DMark99 at 800x600x16 z16 TB. I got a GPU score of 27,959 3DMarks with the x800 XL at stock speed. Not too shabby for Windows 98.

For fun, I also ran it at 1920x1080x32 z24 TB. I got a score of 19,407 3DMarks. I think it's running properly.

fosterwj03 wrote on 2021-12-29, 02:51:

I couldn't find an active link to Tom2D, but I ran 3DMark99 at 800x600x16 z16 TB. I got a GPU score of 27,959 3DMarks with the x800 XL at stock speed. Not too shabby for Windows 98.

For fun, I also ran it at 1920x1080x32 z24 TB. I got a score of 19,407 3DMarks. I think it's running properly.

The link is at the end of the post:
Re: 2D graphics benchmark
If your results are good (especially "Stretching", which is lame for me at all), it will probably mean that this is a problem specifically for nVidia cards, related to flaws in the drivers. Then it seems to me that there will be nothing left but to switch to ATI PCI-E. Or return to AGP...

I ran Tom2D at a resolution of 800x600x16 and got a score of 1866. It looks like the stretching portion hurt the score quite a bit. I don't know enough about this benchmark to identify the cause of that issue.

Yeah. This means that an AGP bus is still needed for full 2D speed in Windows 98? Nothing good will come of PCI-E?

I first noticed this problem when I saw the jerks while stretching the photo. I remembered exactly that on the old AGP system the stretching was perfectly smooth under 9x. And it looked especially great when compared to XP-10 systems (where there were jerks).

For clarity of the problem, for stretching, I used the standard Windows 98 photo previewer "C: \ WINDOWS \ KODAKPRV.EXE" (Used GDI32.DLL) and JPG photos with medium / large resolutions.

Stretched the picture diagonally back and forth (From small to large). With an AGP video card on an AGP bus, it was perfectly smooth. And with a PCI-E video card, even on a more powerful system - terrible jerks!

In general, can we conclude that the lag of GDI 2D in 9x (and stretching in particular) is a problem that does not depend on the manufacturer (AMD / NVIDIA) and applies to all cards connected via the PCI-E bus?

The solution, apparently, as I understand it, does not exist (Except for returning back to the AGP system) 😁

Old AGP system, Windows 98:

New PCI-e x16 system, Windows 98:

But with games and 3DMark tests under Windows 98, it is also acceptable: 3DMark99, 3DMark2000

Spoiler