mkarcher wrote on 2024-03-16, 20:41:

So, that settles the HiNT vs. PLX question. In fact, HiNT got bought by PLX in 2003. The G450 PCI DVI I recently had at hand uses a HiNT branded PCI/PCI bridge. To get it run stable at 40MHz (e.g. for an 5x86 at 160MHz), I needed to add a heatsink to that bridge chip.

Makes sense as my Parhelia 256 pci uses a HiNT branded chip but most of the Parhelia 256 pci I see use the PLX branded ones, I always figured my card was a earlier version than most I see.

I am facing a very strange problem with my G200PCI too.

Within win98 I can not use a dos-fullscreen program, everytime I go back to the desktop the machine freezes completely.
I tested different drivers, different directx-versionen, bios-defaults, loading the vesa-dos-driver or without that driver, it stays the same.

So I took out the G200-PCI and put a 2064W Millenium in that machine (TX97-XE with K6 III+) and it shows the same issues....strange.
I did not have changed the PCI-Slot, but I don't think that that is the issue...

I took out the millenium too and put an old ATI Rage Pro PCI (Mach64) in that machine and everything seems to be fine...strange!

Any idea where the problem might come from ?

Enabeling/Disabling busmaster in the driver did not change a thing, also checked for delayed transaction...still the same, lockup back from dos.

I'd start with a different distribution of Win98.
Then move to testing on a completely different machine... if possible. Even completely switching manufacturers (intel, AMD, via, etc...). Also monitor, peripherals, everything.

On the topic of stuff behaving weird:
I have a Win98SE CD, that crashes on boot PCs that have a Voodoo2 with the latest official drivers installed.
I've seed cases of PC's not posting because of a mouse, keyboard, or a USB device..
I also happened to witness first hand, a long time ago, a K6-2 system not booting Win98 while plugged into one 220V wall socket (with several attempts made)... but worked just fine in the other socket that was 2 meters away.

Sooo... yeah... IMO, nothing is to small to rule out.

Thx, there are lot's of thing to consider... this exact version of win98se I have installed hundred times in several machines...but I will give a go for the first and not the se version.

I will try a mga mystique next...

To be fair, Matrox G450 PCI has problems with VIA only in Windows 98. It works fine in Win2k+ environment. So Matrox probably did not pay much attention to Win9x at that point.

I went for a ATI Rage Pro, this is 100% working with no issues in Win9x...now Amithlon is the problem...damn! VESA with new kernel, no acceleration...

I recently installed a Matrox Millennium G450 DualHead DVI PCI (32MB) in my HP Vectra VLi8 (Pentium III 650MHz) and it works great so far. The mobo is a FIC KZM-6120 (Intel 440ZX chipset) with PCI 2.1 slots and a Matrox MGA-G200 AGP on-board.

I dual boot Windows 98SE and SCO UnixWare 7.1.1. Under Windows, I use Matrox driver package 6.83.017. VBIOS is 1.3 b05. Bus mastering is enabled. I tested DESCENT3 and PowerDVD V4, both worked flawlessly.

From UnixWare:

1# /usr/X/bin/VideoHelp 
2 IBM COMPATIBLE MATROX/Matrox G450 VGA/VBE BIOS (V1.3) b05
3Copyright (C) 1998, Matrox Graphics Inc.
4Copyright (C) LSI Logic Corporation 1990-1991
5DATE: 12/12/01
6Revision: 0.05 
7VGA/VBE BIOS, Version V1.3

The video quality is great and I really like the fact that I can drive a HDMI display (via DVI) at 1376x768x32bpp (1.79:1).

Starcat wrote on 2024-10-20, 06:54:

I recently installed a Matrox Millennium G450 DualHead DVI PCI (32MB) in my HP Vectra VLi8 (Pentium III 650MHz) and it works gre […]
Show full quote

I recently installed a Matrox Millennium G450 DualHead DVI PCI (32MB) in my HP Vectra VLi8 (Pentium III 650MHz) and it works great so far. The mobo is a FIC KZM-6120 (Intel 440ZX chipset) with PCI 2.1 slots and a Matrox MGA-G200 AGP on-board.

I dual boot Windows 98SE and SCO UnixWare 7.1.1. Under Windows, I use Matrox driver package 6.83.017. VBIOS is 1.3 b05. Bus mastering is enabled. I tested DESCENT3 and PowerDVD V4, both worked flawlessly.

From UnixWare:

Copy code to clipboard

1# /usr/X/bin/VideoHelp 
2 IBM COMPATIBLE MATROX/Matrox G450 VGA/VBE BIOS (V1.3) b05
3Copyright (C) 1998, Matrox Graphics Inc.
4Copyright (C) LSI Logic Corporation 1990-1991
5DATE: 12/12/01
6Revision: 0.05 
7VGA/VBE BIOS, Version V1.3

The video quality is great and I really like the fact that I can drive a HDMI display (via DVI) at 1376x768x32bpp (1.79:1).

Please do 2 benchmark rounds in 3Dmark2000: with, and without Bus Mastering turned on from the BIOS.
Normally, Bus Mastering enabled, should boost the performance by at least 50%, than with it turned off.

But if the results will be the same, then it means that somehow the drivers (or the motherboard) disables Bus Mastering regardless of what is set in the BIOS.

B24Fox wrote on 2024-10-20, 19:29:

Please do 2 benchmark rounds in 3Dmark2000: with, and without Bus Mastering turned on from the BIOS. Normally, Bus Mastering e […]
Show full quote

Starcat wrote on 2024-10-20, 06:54:

I recently installed a Matrox Millennium G450 DualHead DVI PCI (32MB) in my HP Vectra VLi8 (Pentium III 650MHz) and it works gre […]
Show full quote

I recently installed a Matrox Millennium G450 DualHead DVI PCI (32MB) in my HP Vectra VLi8 (Pentium III 650MHz) and it works great so far. The mobo is a FIC KZM-6120 (Intel 440ZX chipset) with PCI 2.1 slots and a Matrox MGA-G200 AGP on-board.

I dual boot Windows 98SE and SCO UnixWare 7.1.1. Under Windows, I use Matrox driver package 6.83.017. VBIOS is 1.3 b05. Bus mastering is enabled. I tested DESCENT3 and PowerDVD V4, both worked flawlessly.

From UnixWare:

Copy code to clipboard

1# /usr/X/bin/VideoHelp 
2 IBM COMPATIBLE MATROX/Matrox G450 VGA/VBE BIOS (V1.3) b05
3Copyright (C) 1998, Matrox Graphics Inc.
4Copyright (C) LSI Logic Corporation 1990-1991
5DATE: 12/12/01
6Revision: 0.05 
7VGA/VBE BIOS, Version V1.3

The video quality is great and I really like the fact that I can drive a HDMI display (via DVI) at 1376x768x32bpp (1.79:1).

Please do 2 benchmark rounds in 3Dmark2000: with, and without Bus Mastering turned on from the BIOS.
Normally, Bus Mastering enabled, should boost the performance by at least 50%, than with it turned off.

But if the results will be the same, then it means that somehow the drivers (or the motherboard) disables Bus Mastering regardless of what is set in the BIOS.

Both runs were done with a fresh load of Windows 98SE, DirectX 9.0c, and 3DMark2000.

Run 1, Bus mastering enabled in the driver, 1785 3D marks
Run 2, Bus mastering disabled in the driver, 849 3D marks

That's a 52.4% performance decrease when bus mastering is disabled.

Perhaps worth noting, two other PCI bus masters are installed. An Adaptec AHA-2940U2W SCSI HBA and a 3Com 3C900B-COMBO network card.

Starcat wrote on 2024-11-01, 21:34:

Both runs were done with a fresh load of Windows 98SE, DirectX 9.0c, and 3DMark2000. […]
Show full quote

B24Fox wrote on 2024-10-20, 19:29:

Please do 2 benchmark rounds in 3Dmark2000: with, and without Bus Mastering turned on from the BIOS. Normally, Bus Mastering e […]
Show full quote

Starcat wrote on 2024-10-20, 06:54:

I recently installed a Matrox Millennium G450 DualHead DVI PCI (32MB) in my HP Vectra VLi8 (Pentium III 650MHz) and it works gre […]
Show full quote

I recently installed a Matrox Millennium G450 DualHead DVI PCI (32MB) in my HP Vectra VLi8 (Pentium III 650MHz) and it works great so far. The mobo is a FIC KZM-6120 (Intel 440ZX chipset) with PCI 2.1 slots and a Matrox MGA-G200 AGP on-board.

I dual boot Windows 98SE and SCO UnixWare 7.1.1. Under Windows, I use Matrox driver package 6.83.017. VBIOS is 1.3 b05. Bus mastering is enabled. I tested DESCENT3 and PowerDVD V4, both worked flawlessly.

From UnixWare:

Copy code to clipboard

1# /usr/X/bin/VideoHelp 
2 IBM COMPATIBLE MATROX/Matrox G450 VGA/VBE BIOS (V1.3) b05
3Copyright (C) 1998, Matrox Graphics Inc.
4Copyright (C) LSI Logic Corporation 1990-1991
5DATE: 12/12/01
6Revision: 0.05 
7VGA/VBE BIOS, Version V1.3

The video quality is great and I really like the fact that I can drive a HDMI display (via DVI) at 1376x768x32bpp (1.79:1).

Please do 2 benchmark rounds in 3Dmark2000: with, and without Bus Mastering turned on from the BIOS.
Normally, Bus Mastering enabled, should boost the performance by at least 50%, than with it turned off.

But if the results will be the same, then it means that somehow the drivers (or the motherboard) disables Bus Mastering regardless of what is set in the BIOS.

Both runs were done with a fresh load of Windows 98SE, DirectX 9.0c, and 3DMark2000.

Run 1, Bus mastering enabled in the driver, 1785 3D marks
Run 2, Bus mastering disabled in the driver, 849 3D marks

That's a 52.4% performance decrease when bus mastering is disabled.

Perhaps worth noting, two other PCI bus masters are installed. An Adaptec AHA-2940U2W SCSI HBA and a 3Com 3C900B-COMBO network card.

This is very good news! Thank you!
This basically confirms what @The Serpent Rider said; that the G450 PCI and VIA chipsets, don't mix (at least in win98)

And it's also very interesting that you got basically the same score that I did in XP Re: Trouble installing Matrox G450 PCI on Win98(SE) (but in your case, with a much lower speed CPU).

The bus mastering problem is the same on Windows XP.

From the Matrox Millennium G450 • G450 LE • G450 LX • G450 PCI • G450 DVI • G450 SH User Guide

Matrox Graphics Card – User Guide 47

„ G450-based PCI graphics cards are recommended for computers with Intel chipsets.
With non-Intel chipsets, features and performance may be limited. Specifically, the bus
mastering feature may not be supported and, as a result, the DualHead DVDMax feature
may not be supported and OpenGL acceleration may not be available.

Bam, it's all right there in the manual! I wonder if they advertised this in their marketing or on the box.

B24Fox wrote on 2024-11-02, 01:03:

This is very good news! Thank you! This basically confirms what @The Serpent Rider said; that the G450 PCI and VIA chipsets, do […]
Show full quote

Starcat wrote on 2024-11-01, 21:34:

Both runs were done with a fresh load of Windows 98SE, DirectX 9.0c, and 3DMark2000. […]
Show full quote

B24Fox wrote on 2024-10-20, 19:29:

Please do 2 benchmark rounds in 3Dmark2000: with, and without Bus Mastering turned on from the BIOS.
Normally, Bus Mastering enabled, should boost the performance by at least 50%, than with it turned off.

But if the results will be the same, then it means that somehow the drivers (or the motherboard) disables Bus Mastering regardless of what is set in the BIOS.

Both runs were done with a fresh load of Windows 98SE, DirectX 9.0c, and 3DMark2000.

Run 1, Bus mastering enabled in the driver, 1785 3D marks
Run 2, Bus mastering disabled in the driver, 849 3D marks

That's a 52.4% performance decrease when bus mastering is disabled.

Perhaps worth noting, two other PCI bus masters are installed. An Adaptec AHA-2940U2W SCSI HBA and a 3Com 3C900B-COMBO network card.

This is very good news! Thank you!
This basically confirms what @The Serpent Rider said; that the G450 PCI and VIA chipsets, don't mix (at least in win98)

And it's also very interesting that you got basically the same score that I did in XP Re: Trouble installing Matrox G450 PCI on Win98(SE) (but in your case, with a much lower speed CPU).

You're welcome!

I'll be installing a G450 PCI in a Vectra with a 600MHz Pentium III, Ultra ATA/33 (instead of Ultra2 SCSI), and less SDRAM (128MB vs. 512MB). I'll run 3DMark2000 on that system too and post the results.

Kahenraz wrote on 2024-11-02, 03:25:

I wonder if they advertised this in their marketing or on the box.

I can almost hear them now:
Matrox Graphics Inc. engineers: " 🤣 that's a good one! While we're at it; let's also advertise the G450 actually being weaker clock for clock than the G400!! 🤣 🤣 Better yet; let's light the whole company on fire! I always wanted to work for nVidia anyways 🤣 "

I had a different problem with my Parhelia 256 pci.

If I remember correctly in win2k I had problems with my Parhelia 256 pci not being recognized by the driver as the pci bridge chip was not being properly assigned resources (win2k kept trying to give it resources already being used by motherboard) I think I fixed it by disabling ACPI in Bios. Although in my case bus mastering still worked (ServerWorks LE chipset) but I see the Matrox Parhelia User Guide has no such mention of pci cards are recommended for computers with Intel chipsets.

It’s not about the VIA chipset. I observe the same artifacts with i430TX and it’s getting worse with non Intel CPUs. Here is my list of CPUs that I’ve tried:
Pentium 200MMX (best with almost no artefacts)
Pentium 233MMX
Cyrix MII PR300+ (I tried with lower multipliers down to 133Mhz (PR166+) and it’s almost the same)
Pentium 100
Winchip C6-180
AMD K5 100 (worst, barely usable)
I’ve tried with two motherboards ZIDA 5STX (256Kb) and A-Trend ATC-5000 (512Kb) on Win95OSR2 and Win98. Driver versions 6.23 and 6.83.017. Unchecking option “Use bus mastering” does not help at all. There are no options for bus mastering in the BIOS to try. I will try WinXP with driver version 5.92.006 that I used on another setup with i815 and did not notice any artifacts.

drianov wrote on 2024-11-03, 17:37:

It’s not about the VIA chipset. I observe the same artifacts with i430TX and it’s getting worse with non Intel CPUs. Here is my […]
Show full quote

It’s not about the VIA chipset. I observe the same artifacts with i430TX and it’s getting worse with non Intel CPUs. Here is my list of CPUs that I’ve tried:
Pentium 200MMX (best with almost no artefacts)
Pentium 233MMX
Cyrix MII PR300+ (I tried with lower multipliers down to 133Mhz (PR166+) and it’s almost the same)
Pentium 100
Winchip C6-180
AMD K5 100 (worst, barely usable)
I’ve tried with two motherboards ZIDA 5STX (256Kb) and A-Trend ATC-5000 (512Kb) on Win95OSR2 and Win98. Driver versions 6.23 and 6.83.017. Unchecking option “Use bus mastering” does not help at all. There are no options for bus mastering in the BIOS to try. I will try WinXP with driver version 5.92.006 that I used on another setup with i815 and did not notice any artifacts.

It would be quite sad if the card doesn't work well, right-smack on the SK7 platform.. which would have been exactly it's best match 😕
This card, paired with a Voodoo2, on a 233MMX(with maybe just a "splash" of OC); would have made a very cool & versatile Pentium 1 machine, IMHO.