I made a video or two over 12 months ago of the two systems side by side. It doesn't offer much information but does clearly show the loading time gap between both of them when the hard disk is accessed heavily (IDE vs SCSI)

HEXEN Cyrix 5x86 120MHz IDE vs Cyrix 5x86 133MHz SCSI.

https://www.youtube.com/watch?v=y3VWzHj3QUc

I have never heard of that network card. Best to try something common place, like the 3Com Etherlink III ISA.

I think your Cyrix 5x86-133/4x was set to 3.60 V because you have a chip which has "3.6V" printed on it.

In my tests, Windows 98SE yielded the same benchmark results as Windows 95c. I use Windows 95C rather than Windows 98SE because it is more vintage and reminds me of the times. I never used Win98 when it was in production (I transitions from Win95a to NT4).

Today I spent the better part of an afternoon installing Windows 95 drivers and fault finding. After sorting out Windows 95/Drivers/Games/Hard Drive issues I attempted to run Grand Theft Auto within 3DFX mode and I ended up getting this every time:

Next I loaded up an original Voodoo 2 3DFX "GAME PACK" CD with 15 or so games on it and I tried Frogger 3D and got something similar. It didn't matter if I used DirectX or Glide both failed. I swapped out my LCD monitor and tried an old KTX and Hansol CRT I have laying around - I basically got the same error even if it looked a little different. I made some changes testing other drivers with no success. I started to suspect a faulty Voodoo 2 card, I swapped the Voodoo 2 card with a spare I had and Frogger 3D worked perfectly the first time.

I've spent a couple of hours playing games using the Voodoo2 and playing 256/320k mp3's and haven't had a single crash. I'm really impressed with the Voodoo2 card, finally Grand Theft Auto and other games are extremely playable in higher resolutions. I was surprised at how responsive Quake and GTA are now @ 800x600.

glQuake -width 800 -height 600 -nosound -nocdaudio -nonet -nojoy +timedemo demo2
glQuake -width 640 -height 480 -nosound -nocdaudio -nonet -nojoy +timedemo demo2

◙ GLQuake @ 800x600 yields 27.7 frames per second.
◙ GLQuake @ 640x480 yields 27.7 frames per second.

glQuake -width 800 -height 600 -waveonly -nomouse -nosound -nocdaudio -nonet -nojoy +timedemo demo2
glQuake -width 640 -height 480 -waveonly -nomouse -nosound -nocdaudio -nonet -nojoy +timedemo demo2

◙ GLQuake @ 800x600 yields 28.0 frames per second.
◙ GLQuake @ 640x480 yields 27.9 frames per second.

Surely that's close to the maximum performance I'll get with this hardware combination. I did attempt to load Windows 95 a couple of times with the Cyrix enhancement Branch Prediction turned on but I get a Windows Protection Fault during boot every time.

Its nice to see someone else running this configuration. What is your GLQuake result with sound enabled?

With these settings, I get:

-width 800 -height 600 -bpp 16 +timedemo demo1 = 25.8 fps
-width 800 -height 600 -bpp 16 -waveonly +timedemo demo1 = 25.7 fps (I seem to recall the -waveonly settings being more stable)
-width 800 -height 600 -bpp 16 -waveonly -nomouse +timedemo demo1 = 26.0 fps
-width 800 -height 600 -bpp 16 -waveonly -nomouse -nosound +timedemo demo1 = 29.0 fps

I consider 29.0 fps to be my results using the default quake settings. System is an IBM 5x86-133/2x with 64 MB FPM and 1024 KB cache. The audio card I am using is a Yamaha YMF718-S (Audican32 plus w/XR385). I recall the Audican32 having quite a bit more loss in FPS with sound enabled compared to the Create AWE64Gold. I think using the AWE64Gold over the Audican32 should give an extra ~1.2 fps.

Sorry about the sound volume but you get the gist of the speed and responsiveness of the games at 640x480 and 800x600.

◙ Grand Theft Auto using 3DFX with Cyrix 5x86 133MHz https://youtu.be/kQMpZ1M5vBc
◙ Tomb Raider 2 using 3DFX with Cyrix 5x86 133MHz https://youtu.be/cW2PJkMR9Eg
◙ Frogger using 3DFX Cyrix 5x86 133MHz https://youtu.be/W3bV7CBuNvQ
◙ GL Quake using 3DFX with Cyrix5x86 133MHz https://youtu.be/7R8ZC3Ii8so
◙ Virtua Squad 2 using 3DFX with Cyrix 5x86MHz https://youtu.be/CQqjODd3Q2U

Nice results feipoa, here they are with mine:

GLQUAKE:

My Cyrix 5x86-133/4x :
◙ width 800 -height 600 -bpp 16 +timedemo demo1 = 23.5 fps
◙ width 800 -height 600 -bpp 16 -waveonly +timedemo demo1 = 23.5 fps
◙ width 800 -height 600 -bpp 16 -waveonly -nomouse +timedemo demo1 = 23.7 fps
◙ width 800 -height 600 -bpp 16 -waveonly -nomouse -nosound +timedemo demo1 = 26.3 fps

Feipoa's IBM 5x86-133/2x :
◙ width 800 -height 600 -bpp 16 +timedemo demo1 = 25.8 fps
◙ width 800 -height 600 -bpp 16 -waveonly +timedemo demo1 = 25.7 fps
◙ width 800 -height 600 -bpp 16 -waveonly -nomouse +timedemo demo1 = 26.0 fps
◙ width 800 -height 600 -bpp 16 -waveonly -nomouse -nosound +timedemo demo1 = 29.0 fps

So I had a blast playing some more 3DFX games today, here's a 3DFX title that looks and sounds stunning running between 9.00 - 66.0 frames per second on my Cyrix 5x86 133MHz with 3DFX / Voodoo 2 and all display options on high. Check out the video or you probably wouldn't believe it! Spectacular!

Forsaken 3DFX on a 486! 640x480
https://youtu.be/3yGAttirh1c
https://youtu.be/sDFbu-dtfxo

60.10fps

30.98fps

30.45fps

Forsaken is d3d only, by the way...

it's "-wavonly" not "-waveonly"

rgart, did you find a hidden 3dfx patch for Forsaken, or is that the work of the Matrox G200?

Your GLQuake results weren't as good as I thought they would be compared to mine. Did you set vcache?
If so then I guess the slightly faster RAM/cache speed of 66 MHz FSB system pulls it this much ahead. Or perhaps the 1024K vs. 256K? Write-back vs. write-through cache? Or the sum of small elements add up well.

How did you get 27.7 fps previously?

leileilol, thanks.

Turning off parity check made the fps a little faster. Here's a comparison with NT4.

Biostar MB-8433 - IBM 5x86c-133/2x - 1024 KB write-back cache - 64MB RAM - Voodoo2 - Audican32plus
Win95c
-width 800 -height 600 -bpp 16 -wavonly +timedemo demo1 = 26.1fps
-width 800 -height 600 -bpp 16 -wavonly -nomouse -nosound +timedemo demo1 = 29.1 fps

WinNT4
-width 800 -height 600 -bpp 16 -wavonly +timedemo demo1 = 26.8fps
-width 800 -height 600 -bpp 16 -wavonly -nomouse -nosound +timedemo demo1 = 29.0 fps

NT4 results with sound are a little better, however when actually playing the game, I feel Win95c to be smoother.

DTK PKM-0033S - Cyrix 5x86-133/4x - 1024 KB write-through - 128 MB - Voodoo3 3000 - AWE64Gold
Win95c
-width 800 -height 600 -bpp 16 -wavonly +timedemo demo1 = 25.4 fps
-width 800 -height 600 -bpp 16 -wavonly -nomouse -nosound +timedemo demo1 = 28.2 fps

So the Biostar board is only slightly faster than the DTK board in these configurations.

you also might want to try -snoforceformat, -simsound and -primarysound

-simsound's like -nosound, but still runs the sound mixing

Adding +loadas8bit 1 may also help. Usually this is automatically 1 when there's low (8mb) memory.

feipoa wrote:

rgart, did you find a hidden 3dfx patch for Forsaken, or is that the work of the Matrox G200? […]
Show full quote

rgart, did you find a hidden 3dfx patch for Forsaken, or is that the work of the Matrox G200?

Your GLQuake results weren't as good as I thought they would be compared to mine. Did you set vcache?
If so then I guess the slightly faster RAM/cache speed of 66 MHz FSB system pulls it this much ahead. Or perhaps the 1024K vs. 256K? Write-back vs. write-through cache? Or the sum of small elements add up well.

How did you get 27.7 fps previously?

leileilol, thanks.

It says above in my posts, no I didn't set vcache. What should I be setting it to? We are also using different version numbers for the G200, OpenGL and DirectX.

glQuake -width 800 -height 600 -nosound -nocdaudio -nonet -nojoy +timedemo demo2
glQuake -width 640 -height 480 -nosound -nocdaudio -nonet -nojoy +timedemo demo2

◙ GLQuake @ 800x600 yields 27.7 frames per second.
◙ GLQuake @ 640x480 yields 27.7 frames per second.

glQuake -width 800 -height 600 -waveonly -nomouse -nosound -nocdaudio -nonet -nojoy +timedemo demo2
glQuake -width 640 -height 480 -waveonly -nomouse -nosound -nocdaudio -nonet -nojoy +timedemo demo2

◙ GLQuake @ 800x600 yields 28.0 frames per second.
◙ GLQuake @ 640x480 yields 27.9 frames per second.

I stole the command line from sledge in this thread: My 486 (150MHz and Voodoo! :)

sledge wrote:

Yep: glQuake -width 640 -heigh 480 -nosound -nocdaudio -nonet -nojoy +timedemo demo2

With Voodoo 2 card there was no difference, bottleneck there is obviously the CPU 🙁 Pentium Overdrive is an option, but I would like to stick with "true" 486 CPU. I know Cyrix 5x86 has few Pentium instructions, but...its stil 486 class CPU 😀 And according benchmarks, it IS considerably faster than Intel/AMD CPUs in some ways.

Gamecollector wrote:

Forsaken is d3d only, by the way...

You can clearly see from the video of Forsaken its using 3DFX. Look at all those colours and how smooth it is at 640x480! You simply wont get that smooth picture and those frame rates with D3D and the Matrox G200. Maybe the final release of Forsaken didn't have a 3DFX2 patch but it appears plenty of patches for different hardware did exist.

Forsaken 3DFX on a 486! 640x480x16
https://youtu.be/3yGAttirh1c
https://youtu.be/sDFbu-dtfxo

Voodoo 2 / 3DFX Game Pack CD with 14 playable demo/alpha/beta 3DFX games. G-Police is the only game I failed to get working, I fear it may have Pentium and AGP instruction built into the code.

leileilol wrote:

it's "-wavonly" not "-waveonly"

leileilol wrote:

you also might want to try -snoforceformat, -simsound and -primarysound

-simsound's like -nosound, but still runs the sound mixing

Adding +loadas8bit 1 may also help. Usually this is automatically 1 when there's low (8mb) memory.

thanks leileilol!

-nojoy -nonet -nocdaudio yielded an extra 0.1 fps. I'm now at 29.2 fps with demo1 and 30.7 fps with demo2.

I also tried Forsaken.

Sitting at idle just after entering the game, the Voodoo2 yields 40.2 fps at 800x600; using the Matrox G200 at the same resolution = 37.5 fps. However, the game detail when using the G200 is a bit better. For me, I would definitely use the G200 over the Voodoo2 when playing this game.

For comparison, using software mode yields 3 fps.

feipoa wrote:

the Voodoo2 yields 40.2 fps at 800x600; using the Matrox G200 at the same resolution = 37.5 fps. However, the game detail when using the G200 is a bit better. For me, I would definitely use the G200 over the Voodoo2 when playing this game.

That's very subjective, I have always found games that use 3DFX / Voodoo2 have a softer edge and more vibrant colours than the alternative.

How did you set your vcache feipoa?

rgart wrote:

That's very subjective, I have always found games that use 3DFX / Voodoo2 have a softer edge and more vibrant colours than the alternative.

A forced default gamma of 1.3 or 1.7 plus a dac filter on top tends to do that, yes. Only 3dfx does that, which warps all its rivals as "dark" and "grainy" in opinions at the time. The pure unadulterated 3dfx framebuffer would be just as dark and grainy 😀