At this point in time,
I wanted to see what this increase in core/memory would “buy me.” Before we take
a look at the numbers, let’s establish a Quake3 performance baseline, using a
Max-Quality configuration, without enabling any additional hardware
visual tweaks. With all the Titanium reviews out, this graph should look pretty
Now that we’ve established a performance baseline, let’s
enable some of the hidden features of the GeForce3. In particular, I’m talking
about anisotropic filtering and FSAA. I won't bore you guys with any glossary
definitions and/or screenshots. Suffice to say, the difference in visual quality
Mike Chambers, and others, have proclaimed the anisotropic
filtering of the GeForce3 to be the real golden nugget, in the absence of
DirectX8 games. I couldn’t agree more. In fact, I would probably go a little
further and say that it’s a must-have feature with today’s 3D chips. I used
RivaTuner to choose various feature combinations.
Let’s take a look at 32-Tap performance.
As you can see, the
boost in the core/memory has provided a substantial increase, particularly with
some of the more aggressive settings. There’s no combination of settings which
would be regarded as too “harsh,” as for as performance is concerned. 32-Tap
looks great, but what about 64-Tap? Let’s take a look.
If you subscribe to the “60 FPS” minimum, then you would
have to conclude that this boost would allow for any combination of
anisotropic filtering and FSAA. Personally, I think that 64-Tap + 2xFSAA is the
way to go. The additional boost in performance certainly does give that little
“push over the cliff” (a-la Spinal Tap).
Some people have asked, on various forums, what the
performance might look like at some of the higher-resolutions, namely,
1600x1200x32. Given the resources needed to perform some of these operations, I
think it’s a little too much to ask of a 3D chip (right now at least) to provide
anisotropic filtering and FSAA at this resolution. In any event, let’s
just see what kind of performance gain we get in going from the standard
clocking, to our stable 250/560.
In my opinion, a high anisotropic setting and FSAA
is a little bit too much, even for the GeForce3. Nonetheless, it still delivers
acceptable performance. There’s obviously no issue whatsoever in enabling
32/64-Tap anisotropic filtering (without FSAA) at this resolution. Recall, we
achieved 112.2 FPS without enabling any additional hardware features. With a
GeForce3 clocked at 250/560, you’re looking at dropping between 20-30 FPS using
32/64-Tap anisotropic filtering. That may seem high, but when you have well over
100 FPS to begin with, isn’t it nice to sacrifice some performance for some
quality? I sure do.
Next up is Serious Sam, one of my personal favorites. This
is one sweet looking game, with tons of tweaks available to the gamer. I’m only
going to show the anisotropic filtering performance this time around, due to
No matter how you slice it, your performance is still
ridiculous when enabling this hardware feature. What this performance boost
does, essentially, is provide 32-Tap at almost no cost.
Last on the OpenGL front is DronezMark. Let’s take a look:
The numbers are pretty similar to Serious Sam. It’s almost
like getting 32-Tap at the same performance level as a stock GeForce3.
Let’s turn to DX8 performance. First up is Giants.
This game is extremely odd. I only use Windows2000 and
WinXP (final) at home for gaming. I did, however, have a small Win98
installation on my system up until a few months ago. There is *something*
holding the newer 2000/XP kernel(s) back in the performance department, because
I was attaining nearly 100 FPS several months ago on a Win98 machine. At any
rate, we gained a small performance gain.
Finally, let’s take a look at 3DMark2001
The results are pretty obvious. It’s pretty rewarding to
exit this benchmark to the following window…