For our next
test, we used
Futuremark's controversial benchmark, 3DMark 2003.
NVIDIA's concerns over the validity of 3DMark03 are well
documented, so we won't rehash the arguments here.
With that said, however, over 2.5 million people have
downloaded this benchmark since its release last month.
Two and a half million people can't be wrong...can they?
Controversial Synthetic DX9 Benchmark
Don't read into
the overall 3DMark score, the individual game results and
feature tests tell the real story. In the DX7 Game
Test 1 - Wings of Fury, the GeForce FX 5200 Ultra just
barely missed breaking the 100 FPS barrier. From that
point forward though, things took a dramatic turn for the
worse. In game tests 2 and 3, the GeForce FX 5200
Ultra couldn't hit 10FPS, and in game test 4, 10.1 FPS was
the average. So what does this mean? Well, it's
too early to say. If future games come out using the
rendering methods employed by Futuremark, the 5200 Ultra may
not be very future proof. The more important point to
make is that ATi's current offering of budget graphics cards
can't even complete all of the tests that comprise 3DMark03.
The Radeon 9000, 9100 and 9200 are only DX8 parts.
NVIDIA should be given some praise for bringing the full DX9
feature set to such an affordable price point.
Head-to-Head / Performance With Comanche 4
DirectX Helicopter Combat Sim
Comanche 4 Benchmark, although highly CPU bound, can show
significant variances in performance, at higher resolutions,
especially when AA and Anisotropic Filtering are enabled.
We've stuck to 1024X768 and above here, for our testing
purposes. This benchmark utilizes DirectX 8 Pixel and
Vertex Shader effects as well.
and clear picture is painted here for the GeForce FX 5200
Ultra. Without Anti-Aliasing and Anisotropic Filtering
enabled, the Radeon 9000 Pro took the lead and began to walk
away at higher resolutions. However, enable these
image quality enhancements and the GeForce FX 5200 Ultra
takes the lead handily. This brings to mind a couple
of issues with the GeForce FX 5200 Ultra, namely, immature
drivers and or perhaps, less robust pixel and vertex shader
engines. Frankly, neither of these two cards handle
this benchmark all that well but at 1024X768, the game is
playable. A 2.4GHz Pentium 4 should be enough
horsepower for this benchmark as well, so we're confident
that is not the issue. Finally, the GeForce FX 5200
Ultra does show us here that it has a significantly better
optimized AA engine, such that it can almost double the
Radeon 9000s performance with AA enabled.
Benchmarks / Comparison With Unreal Tournament
Maximum Eye Candy
Antialiasing and Anisotropic filtering disabled, the Radeon
9000 Pro held a significant lead over the GeForce FX 5200
Ultra. We can't say how that performance deltas would
have changed with AA and Aniso enabled, however, because we
could not complete a round a of time-demos with the Radeon,
with AA and Aniso turned on. The benchmark would cause
a BSOD about a minute into the test. We were using a
completely clean system with a fresh install of Windows XP,
and the latest Catalyst drivers... It seems ATi still
has a bit of driver tweaking to do as well. Like all
of the other tests, we suspect the GeForce FX would have run
away with these settings, once AA and Aniso was enabled.
We can't say that officially until we can complete a round
of benchmarks with the 9000 Pro though...
Sammy Pays Us A Visit!