3D Accelerator Review Step One - 3D Performance, the Real Deal


Turok is a pretty nice 3D 1st person shooter game with a lot of pretty 3D effects and really beautiful level design. For the benchmark you only need to download the demo version, which is playable as well. To run the benchmark you use '-benchmark' as command line option and if you've got several different 3D cards in your system add the command line option '-alldrivers' as well. This will let you choose which 3D engine you like to use. Turok supports Direct3D for all cards, Glide for cards with the 3Dfx Voodoo and Voodoo Rush chipset, a PowerVR engine for Videologic's Apocalypse cards and Matrox' M3D as well as a special engine for ATI's Rage Pro chip, as found on the Xpert@.. cards and the new All in Wonder Rage Pro card. The beauty of the benchmark is that it runs for quite a while, showing you the actual frame rate, the average frame rate since you started the benchmark and after running through all the levels it also gives you the overall frame rate, called 'TMARK'. Due to the long run time of this benchmark, the results are very stable and reliable. Running it twice will in most cases give you the exact same frame rate.

For running Turok TMARK you need only the free demo version. You can download it here directly from Acclaim's website. As already said, start it with command line option '-benchmark -alldrivers'. Then you can choose the 3D device you want to use and can adjust the performance settings. I ran all tests with all performance settings ticked, except in case of Permedia 2 cards, where 'render overlap' has to be switched off if you want to see a picture. PowerVR was set to 'normal', Glide ran with 'flip at vsync' switched off and the Rage Pro engine ran with 'fog' enabled.

I used this benchmark to show the Direct3D as well as the special engine performance of the tested cards. It should give you a pretty realistic view of how the cards perform in real world games. I also think that Turok will become pretty successful, so that it will be important how well a card runs this game anyhow. 3Dfx used this benchmark at Comdex to show the performance of their upcoming Voodoo2 chip.

I ran these tests on a system with a Pentium II 300, using the Abit LX6, to see how well the cards perform on a high end platform. To have a view at the other end of the CPU performance spectrum I used a system with Cyrix/IBM 6x86MX PR200, which I consider as a pretty bad 3D gaming CPU. This systems used the FIC PA-2012. You can see that the results differ considerably, in some cases even shockingly.

Turok TMARK Penitum II 300 MHZ

Looking at the Pentium II 300 results shows pretty well that in quite a few cases not the CPU, but the 3D adapter was the limiting factor for the 3D performance. Especially Quantum3D's brute 3D accelerating force shows how much you can get out of a PII 300 system as long as you are using the right 3D adapter. The Glide engine is obviously quite a bit faster than the Direct3D engine for 3Dfx cards, visible at results with Quantum3D's monster card as well as the normal Voodoo and the Voodoo Rush cards. Glide rules this benchmark. You can see that Direct3D is still ruled by NVidia's RIVA 128 chip, the five fastest Direct3D performers are all using this chip. Right after that there come the normal Voodoo cards at Direct3D. Then there is the newcomer 'Thriller 3D' with Rendition's Verite 2200 chip, overclocked to 69.8 MHz, followed by the same card at default chip clock (62.6 MHz). You can see that Diamond's Stealth II S220 with Rendition's Verite 2100 is just as fast as the Thriller 3D as long as you clock it at the same speed. In the Stealth's case this means overclocking to 62.6, because the default clock is significantly less. My Stealth II couldn't get clocked faster than that. The cool thing is that the cheap ($99) Stealth II S220 is even faster than Voodoo Rush cards running with the Glide engine as long as you overclock it. The Voodoo Rush cards are a lot more expensive than the Stealth II though. Behind the two Voodoo rush cards you can find three cards with 3DLabs' Permedia 2 chip. The ELSA Winner Office lead is mainly due to the fact that it doesn't display smoke properly, you can clearly see some rasterization, which you cannot in case of the Diamond Fire GL 1000 Pro. The last third is lead on by Videologic's Apocalypse 3Dx, next is a Voodoo Rush card at Direct3D and then there comes the biggest disappointment of this test, the ATI Rage Pro cards. Although Turok is using a special Rage Pro engine, the performance as well as the image quality are pretty sad. Don't even bother running the Rage Pro under the Direct3D engine, it looks really horrible. At the very end you'll find Intergraph's Voodoo Rush card under Direct3D and then a completely unknown card with the SiS chip 6326. This card is certainly very slow, but the image quality is just fine. You wonder how it comes that this unknown card can display Turok well under Direct3D while the ATI Rage Pro cards can't. Hercules' Dynamite 3D/GL is unfortunatley unable to run Turok. Unlike the other cards with 3DLabs Permedia 2 chip it has got some serious driver problems. The last driver (build 254) would run Turok, but without performing any bilinear filtering, which looks pretty ridiculous, the latest driver (286) only shows some blue screen with an arrow somewhere in it. I guess Hercules will have to ask Diamond or Elsa how they got their drivers to work properly.