VOGONS


First post, by maximus

User metadata
Rank Member
Rank
Member

About a year ago, I decided to start collecting benchmark results for various video cards using the following mid-spec Windows 98 machine:

Dell Dimension 4100
Intel 815E chipset
Pentium III 1000
256 MB PC-133 RAM
20 GB HDD

Many hours of benchmarking later, I'm finally ready to share the results. I'm posting the most interesting graphs below, but this represents only a small fraction of the data I collected. I also have results for 640x480 and 800x600, as well as 16-bit color. I've published the full results on my website, along with analysis and details of the hardware, software, and methodology used:

Windows 98 VGA Charts - PCGames9505

pubchart?oid=327655779&format=image

pubchart?oid=1076882992&format=image

pubchart?oid=1633283799&format=image

pubchart?oid=360549122&format=image

pubchart?oid=730969476&format=image

pubchart?oid=1793727078&format=image

pubchart?oid=1765917677&format=image

A note on the cards:

I thought it would be interesting to test newer, more advanced, low-end video cards against older, less advanced, high-end ones, so I chose cards with this goal in mind. More generally, I also chose cards with low-ish power requirements and good Windows 98 driver support. I don't know if the test system's power supply could handle a GeForce Ti 4600, GeForce FX 5950 Ultra, GeForce 6800 Ultra, or Radeon 9800 Pro, so I left these cards out. (They would all be severely bottlenecked by the Pentium III anyway.) I also got into collecting too late to score many 3dfx cards for good prices, which is why the Voodoo2 and Voodoo5 5500 are not on the charts, and probably never will be.

PCGames9505

Reply 1 of 11, by MMaximus

User metadata
Rank Oldbie
Rank
Oldbie

Thanks for doing this. How come the voodoo 3 3000 shows "0" in all the results though? I have one in my PIII rig and am curious about how it compares to the rest of the bunch...

Hard Disk Sounds

Reply 2 of 11, by clueless1

User metadata
Rank l33t
Rank
l33t
MMaximus wrote:

Thanks for doing this. How come the voodoo 3 3000 shows "0" in all the results though? I have one in my PIII rig and am curious about how it compares to the rest of the bunch...

Because he's only showing results at 32bpp. 🙁

OP - thank you for posting this. I will go to your site to check the rest of the results. When results take months (let alone a year) to collect, I always give the utmost respect. Cheers, and thanks!

The more I learn, the more I realize how much I don't know.
OPL3 FM vs. Roland MT-32 vs. General MIDI DOS Game Comparison
Let's benchmark our systems with cache disabled
DOS PCI Graphics Card Benchmarks

Reply 3 of 11, by RogueTrip2012

User metadata
Rank Oldbie
Rank
Oldbie

Great work. I'm surprised by the mx440 beating a GeForce 3!

If I lived closer I'd loan you my cards for testing. No way I got time for that testing.

> W98SE . P3 1.4S . 512MB . Q.FX3K . SB Live! . 64GB SSD
>WXP/W8.1 . AMD 960T . 8GB . GTX285 . SB X-Fi . 128GB SSD
> Win XI . i7 12700k . 32GB . GTX1070TI . 512GB NVME

Reply 5 of 11, by CkRtech

User metadata
Rank Oldbie
Rank
Oldbie

Thanks for doing this and sharing it with the community. We greatly benefit from the blood, sweat, tears, and time that goes into this endeavor.

Displaced Gamers (YouTube) - DOS Gaming Aspect Ratio - 320x200 || The History of 240p || Dithering on the Sega Genesis with Composite Video

Reply 6 of 11, by Srandista

User metadata
Rank Oldbie
Rank
Oldbie

Thank you for nice set of tests. Just few notes from my side.

In Unreal Benchmark, you probably have a mistake, with resolution 640x480x32, you probably meant x32, not x16: "Enabling 32-bit color in Unreal Gold causes a significant performance drop across the board, even at 640x480x16."

And it's pity, that you tested all ATI cards with 6.2 driver, as it seams, that on slower HW, it's driver with highest overhead, at least according to testing of this fellow Vogoner: ATI Radeon 9800 Pro, quick test.. The 6.2 version is only needed for 9550.

And also, it's too bad, that you don't have Radeon 9500 instead of 9550. That card it's real underdog in my eyes, since if you have right version (non-Pro with 128MB RAM), then that card its having 256-bit memory bus width (not 128-bit, as Wikipedia suggest), and also with some luck can be unlocked by Softmod into full blown 9700, and with a bit overclock even surpass 9700 Pro.

But other then that, I give you my highest respect for your dedication during the testing!

Socket 775 - ASRock 4CoreDual-VSTA, Pentium E6500K, 4GB RAM, Radeon 9800XT, ESS Solo-1, Win 98/XP
Socket A - Chaintech CT-7AIA, AMD Athlon XP 2400+, 1GB RAM, Radeon 9600XT, ESS ES1869F, Win 98

Reply 7 of 11, by maximus

User metadata
Rank Member
Rank
Member
Srandista wrote:

In Unreal Benchmark, you probably have a mistake, with resolution 640x480x32, you probably meant x32, not x16: "Enabling 32-bit color in Unreal Gold causes a significant performance drop across the board, even at 640x480x16."

Fixed. Good catch!

Srandista wrote:

And it's pity, that you tested all ATI cards with 6.2 driver, as it seams, that on slower HW, it's driver with highest overhead, at least according to testing of this fellow Vogoner: ATI Radeon 9800 Pro, quick test.. The 6.2 version is only needed for 9550.

Catalyst 6.2 does seem to have an awful lot of overhead. If I ever revisit any of the ATI cards, I'll have to try some of the older drivers and see what kind of impact they have on the benchmarks. I hear bad things about older ATI drivers, though, so I wonder if I'd just be trading high overhead for bugs. 😒

Srandista wrote:

And also, it's too bad, that you don't have Radeon 9500 instead of 9550. That card it's real underdog in my eyes, since if you have right version (non-Pro with 128MB RAM), then that card its having 256-bit memory bus width (not 128-bit, as Wikipedia suggest), and also with some luck can be unlocked by Softmod into full blown 9700, and with a bit overclock even surpass 9700 Pro.

I'm definitely looking to acquire more R300-based cards. While the Radeon 9550 wasn't the fastest card I tested for this project, it did have the unique ability to enable 4x AA and 8x AF with virtually no performance cost. (Not the case for the GeForce 6200!) Driver compatibility was damn near perfect, too. I can imagine that the mid-range R300 cards would be quite nice to have for Windows 98 gaming.

PCGames9505

Reply 8 of 11, by maximus

User metadata
Rank Member
Rank
Member
RogueTrip2012 wrote:

Great work. I'm surprised by the mx440 beating a GeForce 3!

firage wrote:

Always lovely to see the MX440 do well.

The GeForce4 MX440 is pretty much my favorite Windows 98 card. It's as fast as the GeForce2 Ultra (sometimes faster) and doesn't have the wavy horizontal lines problem (discussed here). The GeForce3 beats the MX440 in some tests, but it has the blurry font problem in Need for Speed: Porsche Unleashed, which is basically a deal breaker for me.

If I could only have one video card for Windows 98 gaming, it would probably be a GeForce4 MX440 or something similar (MX460, Quadro4 380XGL).

PCGames9505

Reply 9 of 11, by maximus

User metadata
Rank Member
Rank
Member

Thanks to everyone who's responded so far. It's good to know that other people find value in this kind of stuff. If I ever get the urge to embark on another benchmarking project, I'll be sure to share the results here.

PCGames9505

Reply 10 of 11, by KCompRoom2000

User metadata
Rank Oldbie
Rank
Oldbie
maximus wrote:

Catalyst 6.2 does seem to have an awful lot of overhead. If I ever revisit any of the ATI cards, I'll have to try some of the older drivers and see what kind of impact they have on the benchmarks. I hear bad things about older ATI drivers, though, so I wonder if I'd just be trading high overhead for bugs. 😒

If you're still having trouble finding older ATI drivers, here's a place that may help (scroll all the way down to the bottom of the page on that link for the older 9x-compatible drivers: http://www.oldapps.com/ati.php

I use Catalyst 3.1, so that version may be worth a shot to try on your Radeon 8500 and 7500.

Reply 11 of 11, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

I'm surprised by the mx440 beating a GeForce 3!

That's because all this tests does not include trilinear filtering.

I must be some kind of standard: the anonymous gangbanger of the 21st century.