Voodoo2 should be always faster than a G200, unless Matrox really upped their game with the later drivers. It's roughly equivalent to a Riva TNT (the V2) and the TNT2 is roughly twice as fast, which 2xVoodoo2 cards can match (or alternatively a Voodoo3). The G400 compares very favorably in D3D to the TNT2 and Voodoo3 and apparently the G400 MAX is faster than both, but not by a whole lot.

I don't know if you can bring up a framerate display in Turok 2 in normal gameplay, but you can benchmark the title by adding "-benchmark" in the game's shortcut, I believe. Glide will always be faster and this was a huge advantage when you had a slower CPU. What system are you using these cards in out of curiosity?

Are you saying that the Voodoo2 when run in Direct3D mode should always be faster than the Matrox G200 when run in Direct3D mode? I'm only considering Direct3D as it is pretty obvious that 3dfx glide runs faster. Are there some bugs with Voodoo2's D3D capability I should be aware of? If the Voodoo2's D3D is faster than the G200's and there aren't any notable bugs with using V2 D3D, then the G200 would only get used as a Windows accelerated 2D card or if wanting to play D3D titles at 1014x768 or higher.

Did the Voodoo1 also work well with Direct3D titles, or was it strictly 3dfx glide?

I'll try the '-benchmark' idea later this evening.

I'm using them on a seriously underpowered CPU - a fully optimised IBM 5x86c at 133 MHz, so depending on the ALU/FPU/Pentium-optimisation mixture used in the game, around a P75 - P100. I played Turok 2 last night for about an hour and to me, it was playable w/Voodoo2 glide.

Yes, a Voodoo2 should always be faster than a G200 when run in D3D mode. I don't know how the underpowered CPU affects your results, but the G200 is in general a weaker card when it comes to performance in D3D (and OpenGL).
There are no bugs with either Voodoo 1's or Voodoo 2's D3D driver/capabilities. 3Dfx cards were really the go to cards for gamers in the late 90's and as such, all developers made sure that their games were running properly on them, they pretty much set the standard.
That being said, you can run into some issues when running newer drivers with some games. For example, one of my favorite games is Warhammer Dark Omen from 1998 and it is quite picky with newer drivers (and newer video cards for that matter). I know for a fact that the latest Voodoo 2 drivers simply crash the game and I've marked an earlier release if I ever want to play the game again. This also applies with the Voodoo3 for example.

If you like visual differences, you'll notice Voodoo2 in D3D always uses a 2x2 dither matrix without subtraction. In games where it'll give you the option between both, with their Glide renderer, it may be sharper....... unless that game also specified a 2x2 dither in Glide (i.e. Unreal games) as opposed to a 4x4 matrix with dither subtraction.

On later cards such as the Voodoo3 this doesn't matter and can be overidden with either dither matrix in the control panel.

-benchmark works for Turok 1 and 2. I ran a small comparison with Turok 2 in my underpowered system, using just the first scene cycle which autoplays when starting the game. The average frame rates are:

3dfx glide at 800x600x16 = 11.4 fps
3dfx d3d at 800x600x16 = 9.4 fps
G200 d3d at 800x600x16 = 9.2 fps

So, not all that much slow-down between G200 and 3dfx in D3D mode with this system. I did note earlier that I felt the game playable with glide, but not so much with D3D. Highest frame rate I saw was 28 fps.

For some reason, Turok 1 hangs at about 30-60 seconds into the game, irregardless of what graphics card I am using (G200 D3D, 3dfx D3D, 3dfx glide). With 3dfx D3D mode at 800x600x16, I get 12.3 fps before it crashes. With 3dfx glide mode at 640x480x16, I get 16.9 fps at crash. Any idea why Turok 1 crashes, but not Turok 2?

I have no clue why Turok 1 crashes. Turok 2 shows no difference in D3D between G200 and Voodoo2 simply because the processor is way too slow for this game. Ideally, you want to start at something like a Pentium II 400 or a K6-2+/K6-III+ 550MHz to see far better framerates (and frametimes) and start enjoying the game. I guess it's not too far off the N64 version with this setup though 😊 .

At this point, I'm not too interested in testing the games on faster motherboards/CPUs. I am trying to relive and make tweeks to my past from '97/'98. I was still using an Am5x86-133 for all of '97 and '98, though I did start a large purchase of a Dell Workstation in October-November 1998 (dual PII-400). It arrived late December 1998, but had a defective SCSI cable and the issue wasn't resolved until early January '99.

I did have a Mystique 220 on my 486 later on, but never really understood certain games worked with certain hardware in accelerated mode. I didn't really play games much anyway as I was particularly studious and academically focused. I did play some Subspace here and there. I'm trying to go back in time and see what it could have been like to make hardware modificiations, e.g. if I had a Cyrix 133 instead of an AMD, and if I had a Voodoo card, to see how I may have played games. Quite a few games came my way from friends at some point and I recall trying to play some of them and thought how slow they ran. One in particular was Need for Speed II, which ran like a slideshow. A few nights ago I tried running NFS2 with a Cyrix 5x86-133 in software mode, and again, it ran like a slideshow, however when I used the Voodoo2 there was a night and day difference. I personally thought the game was playable, but I have never been spoiled by high frame rates. I think even Subspace played at around 20 fps.

I have been testing all the popular 3D game titles from 1995 to 1998 to see which I feel are playable. I've tested several dozen so far and have dozens to go. I'm hoping to compile a list of all games which have some sort of frame rate display and to come up with a strategy to compare different CPUs using this information. If there isn't a benchmark, then at least, perhaps, turn in a circle, or walk some very specific route, attack, or whatever. Its better than nothing. I might rerun the 486/686 CPU benchmarks with these games as it seems game benchmarks acquire the greatest interest. I'd likely use the Voodoo2.

Once I feel like I am finished with game testing on the Cyrix 5x86-133, I'll probably spoil myself with running the same games on my PII to relive the hardware transition. Obviously, if I spoil myself now, I probably won't want to run the games on my Cyrix 5x86-133. The Cyrix 5x86-133 came out in 1996, as did the Voodoo1, but I'm using the Voodoo2 because the display is clearer and it can run at 800x600. Seems like with some tweaking, the Voodoo2 can run most of what was intended for the Voodoo1.