Blaster Control and FastTrax offer the user the choice of aesthetics and stability or speedy updates and convenience. I found myself using the FastTrax drivers most of the time, especially since they are already up to the 5.30 Detonator 2's. Latest and greatest you know...
Everything was not flawless though. Switching from FastTrax to Blaster Control can be rough on your system. It's best change to the standard Windows VGA display adapter, remove Display Director, install Blaster Control and reboot. Install the TNT2W9X.exe drivers from Creative and reboot again, and Blaster Control should recognize your card.
Since this is part one of a two part review, I can't give you a wrap-up of the card, but I will offer some impressions on the card and how it performs. Since this portion of the review mainly covers drivers, I'm going to take on the FSAA issue.
If you play Quake 3, this is the card for you. I've been driving myself nuts trying to figure out what the best balance between image quality versus performance. While OGSS FSAA implemented by NVIDIA is not as elegant as the RGSS implementation of the Voodoo5, it is still effective. I will touch on the settings in more detail in the second part of the review which will cover overclocking, FSAA, and the almighty timedemo.
I would like to comment that 800x600x32 with 2 sample FSAA is extremely playable and enjoyable. Frame rates hover around 55 for the timedemo benchmarks and more importantly, never dip below 25 frames per second. With NVIDIA's implementation of FSAA, the graphics are being rendered internally at 1.5 times the screen resolution and downsampling, so the effective required fillrate is roughly that of 1280x1024x32 when using 2 sample FSAA at 800x600x32. Why not just go with the higher res?
Mainly, pixel popping and texture swimming. Having the added horsepower of the GeForce2 has made me a partial believer in FSAA. I say partial, because 4 sample FSAA (3dfx or NVIDIA) in 32 bit color is not yet ready for prime time. My gaming experiences place the loss in image quality from 32 down to 16 bit color, as too much of a compromise, even when using 4 sample FSAA. Bottom line, it is easier to rail through rocket smoke in 32 bit color. 4 sample FSAA is too much of a fillrate hit in 32 bit color, on either 3dfx or NVIDIA cards.
Isn't 3dfx 2X FSAA better than NVIDIA's? Probably, but one of the conditions I see is that most comparisons, even with corrected color depths, are static. Quake3 is hardly static. In motion, I think gamers will see that each implementation has its strengths and weaknesses, and will iltimately be a matter of personal preference. With NVIDIA's simple 2X FSAA algorithm, the image is cleaner than that of the non-FSAA image and more importantly, non-FSAA at the next higher resolution.
Most important, though, is that it's more consistent. Fewer things draw my eyes away from the target. While some may say that FSAA makes for improved image quality, that's not as important as more consitent image quality when trying to rail the camper across the void.
Direct3D FSAA has come a long way over the past few NVIDIA driver releases. I had a hard time finding a game that didn't agree with the 5.30 Direct3D implementation. Direct3D FSAA performance will continue to increase and image quality will improve. Since FSAA is a feature and not truly a benefit yet, I'm going to call it a draw between FSAA and higher resolutions in Direct3D.
then again, 1600x1200 is an awfully sweet resolution when fragging. :þ
The deciding factor in the Voodoo5 versus GeForce debate is transform and lighting. T&L's main goal, unlike FSAA, is to increase performance. The T&L debate will be addressed in part two of the review as part of the performance numbers. The Voodoo5 5500 may have the edge in FSAA, but the Annihilator2 definitely has the edge in driver support through Blaster Control, FastTrax and Windows 2000 support.
Part Two: Overclocking, In Game Performance, FSAA.
Part Two: Introduction