For example, GLQuake drops from 71.1 fps to 59.9; Quake III drops from 19.6 fps to 17.5 fps; Unreal drops from 20.7 fps to 17.4 fps; Turok2 from 29.3 fps to 23.1 fps; the Quake 2 score did not drop.

Good to know that at least early Radeon 7500/8500 drivers are somewhat decent. Now I need to test Radeon 7000 64mb on a Pentium MMX system again.

I also tested the GF2MX on this system and will publish the charts in the next few days - Radeon 7000, GF2MX, G450, and S4 Savage4 Pro+. I want to re-run some of the S4 benchmarks with Vsync disabled and ensure I can get 3D acceleration working in NT4 and W2K before settling on a card. I'm tempted to run the Radeon 9250, but I think it's too new and I'm spent.

Serpent Rider, do you know how to disable OpenGL Vsync for the Diamond S540 card? When I open PowerStrip, there is only the Direct3D option to disable VSync. All other graphic cards that I used, Powerstrip showed the option to disable OpenGL Vsync. Thanks!

S3Tweak 1.16.

Thanks. Disabling Vsync on the S540 didn't alter the scores at all.

It also looks like the GPU is running at 125.28 MHz, as is the memory. Upping the memory to 144 MHz, which I think is standard for the AGP version, didn't increase the scores either. I tried 144 MHz for the GPU as well, but the scores did not change.

feipoa wrote:

I'm looking for the Diamond Stealth III S540 Xtreme Drivers for W2K and NT4

nt4: s540_nt4_81909.exe , cannot find

Collection drivers for S3 Savage 4.
https://yadi.sk/d/dU1g7bpFqdiif

Fantastic, thank you!

What is unfortunate, though, is that W2K and even NT4 frame rates are well below those from Win9x. I didn't notice this kind of drop when I used the Matrox G200 in Win9x vs. NT4. The frame rates were about the same.

I've finished tabulating the results for these 4 graphics cards. I was wondering if anyone knows if these games use D3D or OpenGL: Incoming, Dark Forces 2, and Expendable?

Also, for the S3 Savage4 Pro+ card, altering D3D and OpenGL Vsync on/off settings did not change any of the benchmark scores. So I cannot really determine if VSync is hardware enabled or disable. For the Matrox G450, Radeon 7000, and GF2, turning Vsync on/off changed the benchmark results. For the same of preparing the charts, should I assume my S3 Savage results are with Vsync disabled or enabled when I compare them against the other graphic cards?

dark forces 2 is direct 3d

Incoming and Expendable use Direct3D.

So I cannot really determine if VSync is hardware enabled or disable

Look for screen tearing.

Thanks.

I'll check.

I've compiled some charts. Here is the information on the test system.

TEST SYSTEM
Intel VS440FX Pentium Pro motherboard using BIOS 1.00.18.CS1
Pentium Pro 200 w/512K running at 233 MHz
128 MB of non-parity EDO memory
Windows 98SE w/Unofficial Service Pack 2.1a

Diamond Stealth III S540 32 MB (S3 Savage4 Pro+)
graphics core/memory = 125 / 125 MHz
S540xtreme driver v4.12.01.0258

Matrox G450 32 MB
graphics core/memory = 112 / 279 MHz
Driver package v6.83

ATI Radeon 7000 32 MB (RV100)
graphics core = 166 / 166 MHz
Driver v4.13.7199; Utilities v6.13.2522

GeForce2 MX400 64 MB
graphics core/memory = 120 / 120 MHz
Nvidia driver package v45.23

Resolutions, API type, test type, driver versions can be found in enclosed table.

Note on Matrox G450: GLQuake requires the -nomtex startup flag set to avoid stuttering in-game. Quake II needed gl_dynamic'0' set in-game to avoid stuttering. [If gl_dynamic '1' then, w/vsync=16.5 and w/out vsync=18.2]

For games which didn't play correctly, e.g. missing textures, or stuttering so bad that its not playable, the results given for that game were 0 fps.

As adjusting Vsync on/off for, both, Direct3D and OpenGL did not result in changes in the benchmark scores for the Savage4 card only. Based on the absense of screen tearing, it looks as if the Savage4 card is insisting on using Vsync. Therefore, all cards are being compared with Vsync enabled. For most games, disabling Vsync did not improve results. The notable exceptions are GLQuake, Quake 2, GL Hexen II, and to a minor exent, MDK2, but only with the GF2 card.

Individualised charts here:

Combined chart here:

And the average of OpenGL, Direct3D, and all games here:

The Radeon 7000 performed surprisingly well when using the oldest drivers. It was, on average of all games, the fastest of the 4 cards tested. The Savage4 Pro+ came in a close second. It was hard to choose between these two cards, but the Savage4 Pro+ is less period incorrect compared to the Radeon 7000 and I think the use of the Metal API makes it all the more interested. Unreal with Metal looked better than all the other cards in my opinion.

EDIT (25 Nov. 2019): I found a typo. For the game Shogo, the Savage4 framerate should be 27.6, not 25.15. Thus the Average for the Savage4 in the Combined Chart should be 29.9, not 29.7. In the Direct3D chart, the Savage4 value should be 28.9, not 28.7. The difference is small, so I don't plan on updating the charts.

Very nice! Didn't think the Savage4 would do well compared to the GeForce 2 MX400. The Dark Forces 2 Matrox result really stands out, not sure what is going on there.

You might get to squeeze a tid bit more performance out of the Savage4 with Blood by using S3VBEFIX /B+. I would have expected the Matrox to pull ahead here because of its memory speed and my own experience with Matrox/Build engine games. Did you see any tearing when benching it?

graphics core/memory = 120 / 120 MHz

That's too low for GeForce 2 MX400. It must be 200/166(200).

I had to say some Nvidia results are quite low and not any better than Pentium MMX 233 with 430TX chipset. These tests were done without sound?

GeForce2 MX400 200MMX

I'm weary to directly compare a mmx and non mmx cpu in games.

I'm weary to directly compare a mmx and non mmx cpu in games.

How so? Pentium Pro should be noticeably faster at least in Quake 2. Although it's probably driver fault. Detonator 45.23 is too new.

Wasn't improved gaming performance and video a major impotus for MMX? So any game/driver that leverages MMX would obviously be faster on a 200mmx VS the ppro.

That's why I covet my overdrive CPUs.

I'm not sure which games can benefit from MMX... but at least in GLQuake, Q3A and UT, PPro is faster than Pentium MMX. http://hw-museum.cz/article/5/cpu-history-tou … 1995---1999-/12

luckybob wrote:

Wasn't improved gaming performance and video a major impotus for MMX? So any game/driver that leverages MMX would obviously be faster on a 200mmx VS the ppro.

That's why I covet my overdrive CPUs.

that was the selling point for MMX, as for reality I'm not sure it was ever utilized to anywhere near that level of benefit