er..as i know the voodoo4 is clocked at 166mhz and voodoo3 at 143/166/183 (concerning the version). if you own a voodoo3 3500 you won't get a speed-up when updating to a voodoo4, except 32bit support and the other stuff..from which vendor(s) are your cards from? i can remember there were voodoo3s & 4s from powercolor..or have you just o/c? then try again with normal clockrates, even if 3dmark isn't the best way to bench out a voodoo...try with standalone ut (--> press       during gameplay, type stat fps & press enter)

I had a similar problem when upgrading from voodoo² 12mb 90mhz to voodoo³ 16mb 166mhz. The only problem is that those games that make use of the more texure-ram/higher resolutions need more cpu power than i have. As Falconfly said in the corresponding thread "the cpu must be fast enough to feed the 3dboard"!

Regards,
Andreas

Sorry, my Voodoo4 is a PCI version.

I ran it in my girlfriend's machine (AMD XP 2200+, 512 Mb 2100 RAM, MSI AGP8x Mobo KT400, Radeon 9000 Pro2) with GTA3, and it ran pretty well compared to the Radeon 9000.    It was about 10fps down on the Radeon in highly complex areas, but still playable at the same resolution of 1024x768, with full distance detail drawing and trails.  Used with AmigaMerlin 3.0 drivers on WinXP.
Then I tried my Voodoo5 5500 PCI, hoping to better that! Boy was I wrong! 
The game was unplayble! It ran at 2fps! Same settings, better card, worst performance!  How? Why?
Again, used the AmigaMerlin 3.0 drivers on WinXP.  I even turned the distance detail drawing down to quite near, and still no difference, even in simple texture areas of the game.  Why is this?  ???
The setting that always stayed ON throughout all 3 tests was the Frame Limiter in GTA3, but why would that make the V5 perform different to the V4 and Radeon?

Hey, i noticed there is a sweet spot when overcloacking my voodoo's, i get an incress to a certain point in the 2001se test but after a point it starts to slow down even though it runs the test just fine, maybe you have hit that point, id try it set to default and work your way up till you see a decress and then back off a bit from there, to close to the edge for to long just burns these cards out   plus frame rates dont really jump up like you think they would with a 20 to 30 bump up, ive run my v5500 at 200 and it wasnt much better than at 183 where i keep it for gamming

The most visible effects of overclocking are slowly drops in performance (when you have done a fresh start of the system and imediately started the game and you have 20fps average, as oposed to after 30 minutes of playing in the same area, when you get to 11fps), and sudden "glitches".

Also a random lockup (after 5 minutes sometimes, after 12 minutes at other time, or 2 minutes at other times), no matter what settings you have, and no matter what 3D scene is displayed, MIGHT suggest an overheating problem.

The first thing to keep on mind is that most of the time the card can produce quite a lot of heating. You need a reasonable heatsink, but most of the time the lack of a fan fails to provide a proper cooling. The airflow inside the case is also very important. It is thus recommended to arrange the cables inside the case so that around the CPU and the video card you have plenty of space for air to travel.

And the videocard heatsink and fan is...?

However, I think that the Banshee example might be worth a look. The Banshee, even if it is clocked at 90Mhz (not overclocked or forced in speed) it still runs pretty hot, even with the heatsink and fan combination. If the fan fails, then the card is freezing or performance drops a lot.

The point is that some videocards can heat quite a lot because of their design. And some of them (The Banshee) can give serious problems even when the fan is working, but not at full speed. There will be interesting to find out if the fan/heatsink combination is working perfectly on the V4 card on this thread.