F
or benchmark fiends, this has been
one hell of a summer. First came id’s
system-stressing Doom 3 , and now
Futuremark is close to releasing its latest
benchmark opus: 3DMark05 , which is
built to bring even the fastest videocards
to their knees.
The new release of one of Maximum
PC’s favorite benchmarks includes only
three main scenes, but each is far more
powerful than anything we’ve seen
before. 3DMark05 uses Shader Model 3.0
but will bump down to earlier shader
models if a card is incapable of the higher
version. (Currently, only nVidia’s 6-series
cards are capable of Shader Model 3.0.)
Users can manually choose among Pixel
Shader 2.0, 2.a, 2.b and Vertex Shaders
2.0 and 2.a.
While the audio performance test
in 3DMark03 has been dropped, the
CPU tests remain and have been made
even more substantial. Although we
had permission to test with the beta
version, we’re going to wait until we
see the final code before rendering any
public judgments on its capabilities
as a benchmark. A recent chat with
Futuremark revealed some interesting
facts about the upcoming benchmark:MPC: What gets lost when a videocard is
running Pixel Shader 2.0 instead of 3.0?
FM : Nothing. The rendering quality
is exactly the same when using Pixel
Shader 2.0 or Pixel Shader 3.0. The only
possible difference is performance, as
PS3.0 can do some shaders with fewer
instructions than PS2.x. The difference
becomes even less evident if the Pixel
Shader 2.x card is capable of PS2.a or
PS2.b performance.MPC: What videocard features are you
testing in 3DMark05?
FM : Basically pixel shaders, vertexshaders, and other features
exposed in DirectX9.0c.
We try to stay away from
the specific features of
independent hardware
vendors (IHV) as it wouldn’t
be an apples-to-apples
comparison.MPC: How do you balance
nVidia’s Shader Model 3.0
versus ATI’s Shader Model 2.x?
FM : Of course, we have tried
to maximize all shaders’
performance, as the newer
shader models (PS2.a, PS2.
b, and PS3.0) can do longer
shaders in one pass, but still
the shaders are designed for
SM2.0. As we said earlier,
the SM3.0-capable cards may
benefit from a performance
increase because some
shaders can be done with
fewer instructions while
still maintaining the same
rendering output/quality. We
haven’t made any specific
IHV optimizations, so as
more cards come out that are capable
of SM3.0, they will automatically use
SM3.0.MPC: How does the CPU test work?
Are there AI or physics calculations?
FM : The CPU tests are much like those
in 3DMark03 , with one exception. In one
of the tests, the CPU not only calculates
the vertex shaders, but also continuously
calculates the flight path of the air ship.
The air ship actually flies the same path
every time, but the calculations are
performed as if it would intelligently steer
according to the canyon shape and other
obstacles (like the sea monster jumping
up from the water).MPC: How are the visual quality tests in
3DMark05 different from that of previous
versions?
FM : We have improved all the image
quality tools from 3DMark03. We have
added more options and settings, and we
combined the filtering test tool with an
antialiasing quality tool. We think the
image quality tools in 3DMark05 are top
notch, and very useful when comparing
different graphics cards’ image quality,
filtering quality, and antialiasing quality.A free version of the benchmark should be
available by the time you read this, with a
pro version costing $19.95.In the Lab A behind-the-scenes look at Maximum PC testing
3DMark05: Tough
on Videocards
On the eve of releasing a brand-new version of its 3 DMark
test, Futuremark explains what’s new—and whyMA XIMUMPC NOVEMBER 2004
A Gatling-gun wielding soldier assaults a space
station in Demo 1 of the new 3DMark05.Game 3 features a Jules Verne-like airship
flying through a beautiful water-filled valley.