Make your own free website on

Voodoo Shack


voodoo6000_table.jpg (53436 bytes)



Image Archive


x3dfx Discussion




Voodoo Shack: When Voodoo is what you do™

3dfxtears_small.gif (2175 bytes)


This review is far more ambitious and thorough than my first comparison, and yet I still find myself coming to the same overall conclusions.  Even though the V4 has a far more robust feature set, in terms of clock for clock gaming speed, the single VSA100 is fairly equal to the older V3. Although the V4 edges out the V3 slightly in many of the benchmarks conducted this time around, the differences are minimal and negligable to gameplay.  The V3 3000 and the V4 4500 perform nearly equal in most benchmarks at the same clock speed of 166MHz, especially in GLIDE and D3D.  This excludes Q3A, where the VSA100 is far superior to the V3.  I'm not sure if this is because of 3dfx driver optimisations for the VSA100 in this game title or something else, but the other newer OpenGL title tested (Serious Sam) didn't show this extreme performance differential.   Regardless, the V4 4500 is faster in Serious Sam, and this could be an indication that the larger amount of SDRAM and support for large textures is paying off in newer game titles.  The V3 3500 will actually be a little faster than the V4 in older games, and should perform equally in new games.  Overclocking yields performance benefits for both the V3 and the V4, and both cards have certain strengths that a gamer should study before making a decision to upgrade.  If, for example, you like to play UT in GLIDE, you'd be better off with your V3 and overclocking it. 

Image quality was similar for all the cards tested (excellent), especially setting the post filter settings to "High/Sharper" while running in 16bit color.  All the cards performed brilliantly in all the titles I tested, and there was really no difference in "seat-of-the-pants" game feel when switching from the V3, V4, or V5 in single chip mode. The exception to this was the inability of the V3 to run UT in D3D at 1600x1200, but since this resolution is unplayable on either card anyways it's really not that critical.  Since both the cards become fill-rate limited around 1024x768x16, using at least a 750MHz CPU should allow the cards to perform at their peak potential when using higher resolutions.  If you want to use FSAA or 32bit color with the V4 4500, you can probably get away with using an even slower CPU and not notice a performance penalty.  If you like playing your games in low-res 16bit with no FSAA, the V4 should scale up to and beyond a 1GHz CPU.   

Bottom line:  Overall, the V4 4500 is superior to any V3, as well it should be.  Quake3 engine games, large texture support, texture compression, 32bit color, and FSAA are all strong advantages for the V4, but when running FPS games in 16bit color, the no-frills V3 will hang in there right beside the V4 most of the time.  If you already have a V3 3000 or better, keep it and overclock it as much as you can until your next big card upgrade.  There just isn't any tangible performance benefit by "upgrading" to a V4 4500, and it won't be worth the switch if you're looking for speed.  Although newer games may continue to run a little better on the V4 4500, you can somewhat compensate for this by overclocking your V3 as mentioned.  The V4 4500 is a great secondary card if you can find one cheap on Ebay, however, and I will be using it in my other rig on my home gaming LAN for the time being (clocked at a comfortable 183MHz).  Who knows, I may build a third system for the old LAN so that I can continue to use my super-fly 200MHz V3.


[Intro | Test System/Scaling | Benchmarks | Overclocking | Conclusion]

Copyrightę 2002 Nightstormer Productions