 |
|
| |
|
| |
     |
| |
|
| |
      
|
| |
|
|
|
| |
|
| |
|
|
|
| |
|
You
need to consider several factors when testing 3-D game
cards. The most important is how well it renders complex
3-D operations and how fast it plays games. But almost as
important, in the case of stand-alone cards, is whether
or not the 3-D card will supply acceptable DOS and
Windows 2-D performance. Finally, we were curious about
the impact of different processors on 3-D performance, so
we used two machines. Our primary testing station was a
Pentium 100 with 16MB of memory. But we also spot-checked
our 3-D numbers and gameplay on a Pentium 200 with 32MB
of memory. |
| |
|
|
|
| |
DOS Testing
 |
| |
|
To test DOS
performance, we ran three separate tests. The first two
were Chris Dial's Bench and SVGA bench, which rotates a
cube in VGA and SVGA modes respectively, outputting a
score. Our third and more important test was running an
automated Quake demo that would generate a frame rate
count. This Quake test was executed in several DOS
resolution modes; if these were not supported, we
installed Scitech's Display Doctor which, for certain VGA
chips, allows the user to raise resolutions in DOS mode. |
| |
|
|
|
| |
Windows Testing
|
| |
|
For Windows
testing, we first ran Steve Gibson's Torque, a simple
test which measures how fast the card can move rectangles
from main computer memory onto itself (across the PCI
bus), and how quickly the card can move rectangles inside
of it's own frame buffer. We then did some real-world
testing, scrolling a Word and Photoshop image and timing
how long it took. All 2-D testing took place in two
resolutions and color bit depths. |
| |
|
|
|
| |
3-D Testing, Part One
|
| |
|
When
evaluating 3-D performance, it's necessary to conduct two
completely different types of testing. First, to
determine the strength of the hardware itself, we needed
to find titles that have been ported to as many
ASIC-types as possible. Luckily, Activision's MechWarrior
2 has been ported directly to the hardware of every
single product in our roundup save the three
Rendition-based boards. Interplay's Descent II has also
been ported to the Virge and the Voodoo chips, which are
used in four of the boards in our roundup. (By the time
you read this, there will also be a Rendition-ready
Descent II available.) Finally, Criterion's Scorched
Planet proved helpful, with versions compatible with the
Mystique, Rendition, and 3Dfx chips. Finally, of course,
we tested all games bundled with each product, rating
their performance based on more subjective criteria. |
| |
|
|
|
| |
3-D Testing, Part Two
|
| |
|
The second
aspect of 3-D testing was just how well each board
accelerated Direct3D, Microsoft's quick and
device-independent API and hardware interface. To find
out, we ran three tests from the Direct3D Software
Developers Kit - the Direct3D test, Tunnel, and Twist -
in several settings. Specifically, the Direct3D tests
measures a board's "fill rate," the ability of
the ASIC to move high number pixel textures onto the
screen, as well as the "polygon score," which
twists and turns a series of fancy polygons around and
measures how fast the board is capable of rendering them
as well. Finally, we just played Monster Truck Madness, a
Direct3D truck racing title from Microsoft. Together, all
these tests gave us an excellent idea of the strength of
each 3-D hardware product. |
| |
|
|
| |
     |
| |
|
