Good morning / afternoon / evening! This is my first post here, and I bring you some benchmarks that I have been doing these days, since much cannot be done in quarantine, it is an ideal time to build some good builds at home.
My idea with these benchmarks was to compare two VGAs, both PCI, since this build does not have AGP support (the BIOS has support, but the mother does not have the AGP port), in particular the Matrox Millennium II 2MB vs the Diamond S3 Stealth 2000 2MB and being able to come to the conclusion of which is the best of both, separating what the specs of both say.
The PC on which I am running these benchmarks is a project box, that is, it may not be as faithful to what it should have for its time and be of the correct period, but it serves the purpose that I am using it.

The machine specs are as follows:
Motherboard: VIA VT8601A
Processor: Intel Pentium III EB Coppermine 866Mhz
Ram: 128MB PC 133
HD: WD 80GB
Sound: Sound Blaster Live! Value CT4830
Monitor: Apple Studio Display "17 CRT
System: Windows 98 SE

The VGAs I tested as I mentioned are the Matrox Millennium II:

Made by: Matrox
Codename: 2164W (My specific model is 2164WP-C)
Bus: PCI, AGP 1x
Memory Size: 4MB, 8MB (There were also 2MB models like this model that I own)
Max Memory Size: 16MB
Memory Type: WRAM
Year: 1997
Announce Date: Tuesday, 13 May 1997
Card Type: VGA
Manufacturer: NEC
Process tech: 350nm
Outputs: 15 pin D‐sub
Price $: 399
Video Acceleration: MPEG-1 (VCD)
DirectX: DX3/5
Core Clock (MHz): 66
Core: 64bit
Effective Memory Clock (MHz): 66
Memory Bus Width: 64bit
Max Core Clock (MHz): 66
Min Core Clock (MHz): 62
Ramdac (MHz): 250

http://www.vgamuseum.info/index.php/technolog … w-millennium-ii

And the Diamond S3 Stealth 2000:

Made by: S3
Codename: 86C325
Bus: PCI
Memory Size: 2MB, 4MB
Max Memory Size: 4MB
Memory Type: EDO
Year: 1996
Announce Date: Monday, 01 April 1996
Card Type: VGA
Family: ViRGE
Manufacturer: UMC
Process tech: 500nm
Made in: Taiwan
Outputs: 15 pin D‐sub
Price $: 199
Video Acceleration: MPEG-1 (VCD)
DirectX: DX3/5
Core Clock (MHz): 55
Core: 64bit
Effective Memory Clock (MHz): 55
Memory Bus Width: 64bit
Pixel Pipelines: 1
Texture Units: 1
Sold by: Diamond
Max Core Clock (MHz): 55
Max Memory Clock (MHz): 55
Ramdac (MHz): 135

http://www.vgamuseum.info/index.php/cpu/item/ … d-2000-s3-virge

The tests that I carried out were focused specifically on graphic tasks, in particular the software used was
3D Mark 99 Max, tests by Phil Computer Labs, and a gameplay of Quake 2 version Q2Test and Q2Demo in Software and OpenGL mode.

The results were the following:

3D Mark 99 Max
640x480 in 16bpp with Double Frame Buffer:

Matrox Millennium II - 2249 3DMarks

Diamond S3 Stealth 2000 - 1357 3DMarks

800x600 in 16bpp with Single Frame Buffer:
Matrox Millennium II - 1763 3DMarks

Diamond S3 Stealth 2000 - 925 3DMarks

640x480 in 32bpp (Matrox) and 24bpp (S3) with Single Frame Buffer:
Matrox Millennium II - 2096 3DMarks

Diamond S3 Stealth 2000 - 767 3DMarks

Phil Computer Labs Benchs
3DBench 1.0C:
Matrox Millennium II - 329,4

Diamond S3 Stealth 2000 - 330,8

Chris´s 3D Benchmark:
Matrox Millennium II - 892.7

Diamond S3 Stealth 2000 - 634.4

PC Player Benchmark 640x480:
Matrox Millennium II - 77,3

Diamond S3 Stealth 2000 - 41,6

Doom max. details:
Matrox Millennium II - 2134 gametics in 863 realtics

Diamond S3 Stealth 2000 - 2143 gametics in 888 realitics

Quake timedemo 640x480 (Matrox) 360x480 (S3):
Matrox Millennium II - 52,5 FPS in 640x480

Diamond S3 Stealth 2000 - 47,3 FPS in 360x480
(Forgot the photograph)

Quake 2 (Q2Test)
OpenGL in 320x240:

Matrox Millennium II - Averange of 7FPS

Diamond S3 Stealth 2000 - N/A

Quake 2 (Q2Demo)
Software Mode (Q2Demo) in 800x600:

Matrox Millennium II - 29,1 FPS

Diamond S3 Stealth 2000 - 25,4 FPS

Software Mode (Q2Demo) in 1024x768:
Matrox Millennium II - 19,6 FPS

Diamond S3 Stealth 2000 - 16,8 FPS

I leave all the photos corresponding to the benchmarks presented here in order, starting with the Matrox, and also photos of the system and the VGAs.

My conclusions were as follows:

Both being 2MB video cards, one with EDO technology (S3) and the other with WRAM technology (Matrox), and having a year of difference almost between both, the results given by the benchmarks, and also visually and from the quality they deliver, they are very radical from each other, and a pleasant surprise for the Matrox on my part, which I mistakenly believed, that it was going to be the worst in performance, and then when seeing the detailed specs of both, and see also that one is from a 500nm (S3) process and the other from 350nm (Matrox), the difference in frequencies between the two and memory technology, make the difference in these cases.
I was surprised that despite everything that the Matrox could run Quake 2 in OpenGL mode with the lowest possible resolution, and with the maximum textures at a framerate greater than 5FPS. The downside of this video card is that despite supporting a higher resolution and definition than the S3, it suffers from several specific graphic bugs, which for certain applications does not make it as reliable as the S3 despite the higher graphic quality.
And that Quake 1 couldn't run on S3 at 640x480 and Q2Test in OpenGL mode didn't run properly at 320x240 was also somewhat disappointing.
I know that this build would do much better with a more advanced board such as a VooDoo 4 4500 or something similar and a larger Ram Memory, but at the moment it is what I had on hand, and the idea was only to test which of these two VGAs was better .

What are your opinions or conclusions in this regard?