By,
Dave Altavilla
February 1, 2004
Those of you
that like to skip those hard working business and
professional application benchmarks and get straight to the
fun factor, can rejoice here, as we round out of testing
suite with an all out gaming benchmark blitz.
|
Wolfenstein Enemy Territory |
OpenGL Quake Engine Gaming |
|
We've finally
moved on from Quake 3 here in the HotHardware testing labs
but somehow a specter of its gaming engine still haunts the
hallways and test benches in our work areas.
Wolfenstein Enemy Territory is a highly modified Quake 3
engine based game, with impressive all new lighting,
textures and high polygon count models filling the
landscape. We turned the resolution way down, in an
effort to offload the graphics subsystem and let the
processors we tested stretch their legs.
There's no doubt
about it, if you like to game in Quake engine based games
and need extreme performance regardless of price, then the
P4 Extreme Editions can be enticing. However, a second
glance at this series of scores, reports only about an 8%
lead between the fastest 3.4GHz P4 EE score here and good ol'
Northwood at 3.2GHz. In addition, this is obviously a
game engine that favors the P4, with all Athlon 64s
reporting in behind the slowest Pentium 4 score.
Finally, note that Prescott posts in at about a 3% deficit
under the 3.2GHz Northwood score we took here.
|
3DMark 2003 |
DirectX 8 Gaming Performance |
|
We'll step
quickly through FutureMark's 3DMark 2003 CPU test here, to
provide you some reference points for this popular synthetic
DirectX 8/9 gaming based benchmark.
Here we see
large cache size and low memory latency, afford each of the
competitors their respective ranks. Prescott's larger L2
cache allow it a lead over Northwood but not so much that it
can overtake either of its Extreme Edition counterparts.
The Athlon 64 FX-51 here, with its integrated memory
controller in addition to its 1MB L2 cache, takes even the
3.4GHz P4 EE by a small margin of victory.
|
Comanche 4 |
DirectX 8 Gaming - CPU Limited |
|
Over the years,
although this Combat Helicopter simulation from Novalogic
has begun to show its age, time and time again we come back
to it for consistent repeatable metrics on pure compute
intensive power. This game, with its constant
calculation of moving objects in a given battle scene,
really seems to isolate the CPU and overall system
bandwidth. Although it is a gaming benchmark,
interestingly enough it isn't a very good graphics
benchmark, since it is so CPU limited.
Once again, to coin a phrase
"cache is king", pun intended. Low cache latency is
also a deciding factor here as well, with a strong showing
by the Athlon FX but an even stronger showing from the P4
Extreme Editions. Prescott on the other hand, hits
another roadblock and obviously has issues with its deeper
pipeline and branch miss error rates. Even though it
has a full 512KB cache advantage over the standard 3.2GHz
Pentium 4, it falls in behind by a solid 15% shortcoming.
Although we ran this test repeatedly, the results were the
same. Which brings us to an observation we've made in
previous processor launch articles, when there have been
major architectural changes made to a CPU. Some legacy
code is simply not going to run as efficiently on these new
cores, unless it is recompiled to take advantage of the new
architecture.
Unreal Tournament 2003, X2 The Threat and The Wrap-up
|