Do you have a board with AGP so you can compare all Matrox cards, e.g. add P650 AGP, P750 AGP, Parhelia (PCI & AGP), P690 PCI, G250?

In AGP I have only G100; G200; G200A; G400; G400 Max; G550 and Parhelia 128MB retail. But the Parhelia 128MB retail I cannot insert into this P100 benchmark because it not works on AGP 3.3V. But later I plan a DOS VGA benchmark with VIA C3 Ezra-T 1000MHz with most of my cards (about 170 cards).

AGP Matrox cards also has small differences is DOS. G200; G200A; G400 and G400 MAX AGP cards are the fastest and they has full the same performance.
G100; G450 and G550 AGP cards a bit slower and Parhelia 128MB retail AGP a bit more slower especially in Doom (12%).
It is interesting, but in same motherboard the G450 PCI versus AGP gives a bit mixed advantage: PCI better in PCPBench, AGP better in 3DBench and Doom, and Quake is the same. Also interesting that G450 PCI was better than G200 PCI on Intel VX, but G200 AGP is better than G450 AGP on VIA MVP3. It seems G450 and G550 are optimalised to Intel chipsets.

Originally I want to test the Parhelia 128MB retail AGP in an ASUS P3V4X motherboard and PII 233 CPU but the card not start in it. Finally I test it in an AGP 8X board.

Curious. Does the MVP3 chipset generally show faster DOOM results, or is it just with the Millennium series? For examplel, if you used a TNT card on the 430VX vs. MVP3, would the MVP3 still show a bit faster results?

The results look fairly static in DOS. Perhaps doing a similar comparison in Direct3D and OpenGL would be more meaningful?

The Parhelia AGP has two variants - one is keyed for 2x/4x/8x while the other is 4x/8x only. Do you happen to have the Parhelia keyed for 2x?

feipoa wrote on 2022-11-11, 03:55:

Curious. Does the MVP3 chipset generally show faster DOOM results, or is it just with the Millennium series? For examplel, if you used a TNT card on the 430VX vs. MVP3, would the MVP3 still show a bit faster results?

The results look fairly static in DOS. Perhaps doing a similar comparison in Direct3D and OpenGL would be more meaningful?

The Parhelia AGP has two variants - one is keyed for 2x/4x/8x while the other is 4x/8x only. Do you happen to have the Parhelia keyed for 2x?

There seems to be a lot to do with the bios timings in regards to ss7 performance. Ss7 usually has better memory performance also.

K6 3+ aiming for the stars thread talks about this.

feipoa wrote on 2022-11-11, 03:55:

Curious. Does the MVP3 chipset generally show faster DOOM results, or is it just with the Millennium series? For examplel, if you used a TNT card on the 430VX vs. MVP3, would the MVP3 still show a bit faster results?

I have to correct my earlier statement: Also it seems VIA MVP3 is better in Doom, but Intel 430VX/HX/TX are better in all other. to this: It seems VIA MVP3 with AGP versions of the cards is better in Doom, but Intel 430VX/HX/TX (with PCI versions of the cards) are better in all other. With same PCI cards also the Intel 430VX/HX/TX are faster, only smaller the difference in Doom compared to the other benchmarks. I have benched best DOS performer cards and this small difference is common: between 1.59% to 2,52%. The only bit bigger difference the 2.84% is the G450 PCI, because Intel chipset optimalisation. I have something about this optimalisation:
"What features will not be supported for the combination of G450 PCI and non-Intel chipsets?
DVDmax and OpenGL will not be supported for the combination of G450 PCI and non-Intel chipsets.
Will bus-mastering ever be supported with the G450 PCI cards and non-Intel chipsets?
Matrox does not foresee much demand for PCI graphics products for systems with non-Intel chipsets"
(https://web.archive.org/web/20040622004143/ht … 450/pci_faq.cfm)
Here are my Doom results:

feipoa wrote on 2022-11-11, 03:55:

The results look fairly static in DOS. Perhaps doing a similar comparison in Direct3D and OpenGL would be more meaningful?

Direct3D and OpenGL would be full different. For example nVidia AGP cards from Riva128 AGP to the GeForce3 Ti 500 has full the same DOS benchmark results (laters should be same or slower), but in Direct3D and OpenGL GF3 is much faster than Riva128. Also S3 Trio64 is faster in Doom than later S3 cards. It is interesting that in Doom the S3 Virge/GX a bit drop behind although Virge/GX is the only chip can fast enough for DOS Duke Nukem 3D is 1600x1200 (if the CPU is strong enough like P4). I have tested lot cards on the same P4 3GHz and Duke 3D was only playable on Virge/GX in this resulution.
But here Direct3D and OpenGL benchmarks are not interesting because we were wondering what is the fastest chipset for the Pentium 100 (and we were surprised that it is Intel 430VX/HX/TX not the SS7 chipsetted boards), and I would not play Direct3D or OpenGL game on a P100 machine.

feipoa wrote on 2022-11-11, 03:55:

The Parhelia AGP has two variants - one is keyed for 2x/4x/8x while the other is 4x/8x only. Do you happen to have the Parhelia keyed for 2x?

This is the 2x/4x/8x version.

vetz wrote on 2014-05-04, 22:38:

Introduction […]
Show full quote

Introduction

Inspired by Phil's VGA BenchmarkI've started on this motherboard benchmark comparison. The main purpose is to show how the different chipsets and motherboards perform with a Pentium 100. Earlier there has been a misunderstanding that a Pentium performs equal in Socket 5 & 7 boards, but that is not true as you will see from the results. Hopefully the results will show which chipsets and boards have the best performance and how much cache and type of RAM (EDO/SDRAM) affects the results. The Pentium 100 was chosen since it is supported in all boards.

Adding results
If anyone wants to add their results please read this CAREFULLY:

1. Use Phil's VGA Benchmark and boot MS-DOS without anything extra (HIMEM, EMM386, etc)
2. You need a Socket 5 or 7 board with a Pentium 100. It doesn't matter if you overclock or underclock your Pentium as long as you have 66mhz FSB and 1.5x multiplier. Do not use Intel MMX, Cyrix, AMD or other types of CPUs.
3. You need a Matrox PCI Mystique, Millennium II and/or G200. See results from this thread that shows these cards perform the same on a P100.
4. Amount of RAM does not matter, but I would advice having atleast 16MB and 60ns RAM. Most of the Socket 5 boards accept EDO ram even if they cannot use the performance benefit.
5. Set motherboard timings and settings to their best and tightest! All boards tested by me run with their best settings in BIOS. Normally this is not a problem and if a motherboard have more settings to modify than another which gives extra performance that should work in its favor.
6. Report results along with other relevant information in the Google Spreadsheet document.

LINK:
https://docs.google.com/spreadsheets/d/1-Qqfm … #gid=1710302252

Current results:
motherboards_P100.PNG

Did you have to do any tweak to run quake bench in the M519 motherboard? I have the same board and i can't run it.

CachoAlpuy wrote on 2022-12-12, 19:52:

Did you have to do any tweak to run quake bench in the M519 motherboard? I have the same board and i can't run it.

You got an error message?

Gona wrote on 2022-12-12, 20:09:

CachoAlpuy wrote on 2022-12-12, 19:52:

Did you have to do any tweak to run quake bench in the M519 motherboard? I have the same board and i can't run it.

You got an error message?

Error: Texture 4_7 is not 16 aligned

Unusual error message. Try to set the bios to "BIOS default".

Gona wrote on 2022-12-13, 11:24:

Unusual error message. Try to set the bios to "BIOS default".

That fixed it.

I will be sending a few benchmarks with the G450 Matrox card soon.

What other cards could be used to add in this benchmark? (For boards which dont support the G450)

I have Trio 64V+, trident's, cirrus logic's, Ati Rage Pro PCI..

I will be adding benchmarks in the bottom, you put it in order:
https://docs.google.com/spreadsheets/d/110bSx … it?usp=drivesdk

Can we demystify that Utron VX Pro II is a bad chipset at least in DOS?

CachoAlpuy wrote on 2022-12-28, 01:49:

I will be adding benchmarks in the bottom, you put it in order:
https://docs.google.com/spreadsheets/d/110bSx … it?usp=drivesdk

Can we demystify that Utron VX Pro II is a bad chipset at least in DOS?

Yes, this chipset is fast, the fastest non-Intel chipset in this P100 comparison, and less then 2% slower than the fastest result. The chipset true name is "Utron UT85C501/UT85C502": http://leoandfanny.bplaced.net/vintagecompute … C501/index.html. It seems G450 with this chipest also slower a bit as all other non-Intel chipsets. Are you do the benchmarks with Bios defaults? Because with bios defaults the results are extreme good, especially in Doom and Quake. In this case you should find the fastest settings that are still stable. (Separated Univbe line not needed because this only add VBE resolutions to video bios speed is not affected).

Gona wrote on 2022-12-28, 09:32:

CachoAlpuy wrote on 2022-12-28, 01:49:

I will be adding benchmarks in the bottom, you put it in order:
https://docs.google.com/spreadsheets/d/110bSx … it?usp=drivesdk

Can we demystify that Utron VX Pro II is a bad chipset at least in DOS?

Yes, this chipset is fast, the fastest non-Intel chipset in this P100 comparison, and less then 2% slower than the fastest result. The chipset true name is "Utron UT85C501/UT85C502": http://leoandfanny.bplaced.net/vintagecompute … C501/index.html. It seems G450 with this chipest also slower a bit as all other non-Intel chipsets. Are you do the benchmarks with Bios defaults? Because with bios defaults the results are extreme good, especially in Doom and Quake. In this case you should find the fastest settings that are still stable. (Separated Univbe line not needed because this only add VBE resolutions to video bios speed is not affected).

All settings are set to "Enhanced" which is the best possible, there are not a lot of options in the bios to change in the Chipset tab. This settings make a substantial improvement when compared with default settings.

Yes, the result in games (doom and quake) are excelent.

I will remove the UniVbe row because it is useless.