Editorials - Timing the T-Buffer |
Have you ever
looked at a 3D game and tossed your mouse in frustration because a
shadow didn't fade at its edges? I didn't think so. Soft
Shadows and Reflections will be nice, but things can be too soft
sometimes - just check out the Nintendo 64's occasional blur before
you're sure you want Full-Scene Anti-Aliasing in every game you play.
Does Mario worry about cutting himself on those jagged pixels?
Further, I
can't believe the T-Buffer's advances in blurring technology will make
better games possible. When was the last time you asked
"would this game be better if the perspective shifted so it would
be blurry in places?" Our eyes automatically perform Motion
Blur and Depth of Field Blur - try typing on a laptop computer in a
moving car. The motion outside the car will be blurring past you
while the laptop's monitor will remain clear in your sights. Oh
yeah, and let someone else drive when you try the experiment out :)
The bottom line
is that the T-Buffer's effects will make every Glide accelerated game
you play look either slightly better or have nicer in-engine cutscenes.
However, one has to wonder why 3dfx is working with such slow advances
in their technology. New features are always good, but the
Voodoo 3 was little more than an improved Voodoo 2 SLI. If the
T-Buffer is 3dfx's best new advancement, it looks like their next card
will be little more than a slightly enhanced Voodoo 3.
Then again,
things may not work out that way at all.
Nvida's recent
announcement about their NV10 chip and their strategic alliance with
SGI makes them a 3D force to be reckoned with. 3dfx's release of
the 16-bit Voodoo 3 this year, in spite of 32-bit competition, could
very well be a sign that they have bigger plans for a future chip.
If that is true
and their next chip in development is astonishingly powerful, it may
have been a smart decision to skip 32-bit 3D graphics for the interim.
The Voodoo 3 gives them a steady source of revenue while they can
continue on their next product's Research and Development without the
fear of a Christmas deadline.
This is all
guesswork, though. Nvida's announcements about the upcoming NV10
say it will have twice as many transistors as a Pentium III and some
pretty impressive specs: AGPx4, a 350-MHz RAMDAC, hardware
control of geometry, transform and lighting, insanely high polygon
counts (which may be as high as 25 million triangles per second by
shipping time), and NOT from a proprietary API. Remember that,
with the exception of Full-Scene Anti-Aliasing, the T-Buffer's effects
can only be seen in Glide titles.
I was pretty
shocked to read about the NV10's power and 3dfx's T-Buffer
announcement in so few days' time. Nvida's next card will make
ultra-high polygon models more likely, so the in-game objects will get
closer to the quality of prerendered cutscenes. 3dfx's response
is to give the cutscenes nicer looking camera-like effects, but
apparently not to improve the models themselves. Unfortunately
for 3dfx, high-poly models make the difference between an arm with a
bulging bicep and a texture-mapped rectangle. Without
improvements in the polygon count and speed, the T-Buffer looks like
an "all icing and no cake" approach to 3D graphics.
Perhaps 3dfx
intended to release their 32-bit chip sometime next year, but the
R&D went a little slow. Maybe they have special marketing
and PR negotiations that prevent them from hinting at their new
technology. The T-Buffer may only be a part of their ultra-cool
next-generation Voodoo4/Napalm card, and it could just be the feature
with the most press-friendly buzzwords. Or maybe they are still
stuck in a 16-bit mindset and the T-Buffer really is the best thing
they have right now - but if this is the case, a whole lot of people
will look at the T-Buffer and ask "is that all you've got?"
That's why the
timing of the T-Buffer's press release is so suspect. In my
corporate-conspiracy-filled mind, 3dfx obviously wanted to steal some
of Nvida's thunder with a press release of their own, and it's a good
short-term strategy. Even with a comment of minimal consequence,
any announcement from an industry leader like 3dfx is bound to be
repeated more often than the less-famous competition.
The question
that remains is whether 3dfx just wanted to overshadow Nvida's press
release while perfecting their next chip's details in secret, or if
3dfx actually believes the T-Buffer technology will be more valuable
than an increase in speed and polygon count. Their reputation
may depend on the answer.
Between Guru editorial columns,
David Filip is a freelance writer, musician, and one-man sound studio.
He can be contacted at grimlock@u.washington.edu
|
