Test System Setup
Processor: AMD Mobile Barton @ 2GHz (200 x 10)
Motherboard: ABIT NF7-S v2.0 (nVidia nForce2)
Memory: 2 x 512MB Buffalo PC-3700 (Winbond BH5) @ 2-2-2-11
Graphics Card: Gigabyte Radeon X800 XT PE @ 520/560
Hard Drive: Seagate Barracuda V 120GB 7,200 RPM PATA
Operating System Used: Microsoft Windows 2000 SP4
Drivers Used: ATI Catalyst 4.9 CCC and DX9c
To give you an idea of the performance differences between 3DMark03 and 3DMark05, we ran a series of tests in both versions on the test system above at 1024 x 768, 1280 x 1024 and 1600 x 1200 with 6x AA and 16x AF (maximum image quality) enabled and disabled at each resolution.
We also included results with the CPU overclocked by around 600MHz (228 x 11.5 = 2629MHz) to help determine how both 3DMark03 and 3DMark05 overall scores scale with CPU clock speeds. We suspect the increase in CPU clock speed won't make much of a difference since for the most part both versions of 3DMark are focused on the GPU to handle most of the intensive 3D tasks.
We have used the ATI Catalyst 4.9 driver which is the newest driver from ATI to gain approval from FutureMark, essentially clearing the driver from any unfair optimizations after the issues with drivers and 3DMark03 last year.
For those wondering how the overall 3DMark score is calculated, you need to use the following formula:
3DMark05 score = (Game Test 1 x Game Test 2 x Game Test 3)^0.33 x 250
3DMark03 vs. 3DMark05
As expected at 1024 x 768 we can see a large difference in scores between both versions of 3DMark with 03 showing a performance increase of almost 160%. Although, when we enable high image quality the difference is only about 53%.
As we increase the resolution the performance gap between both versions is reduced to about 146% and 43% in line with the increased image quality.
At the highest resolution we used, the gap is the smallest (even though it is still very large) between both versions at around 143% with around about a 43% difference with high quality graphics enabled.
It is clear that 3DMark05 pushes even the highest-end GPU of today from ATI to the maximum with a vast difference in scores between the mostly DX8 3DMark03 benchmark compared against the full DX9 3DMark05 benchmark.
CPU Impact on 3DMark Scores
In our first benchmark to investigate how much CPU clock speed influences the 3DMark score, we can see that in 3DMark05 there is only an increase of about 5% and in 3DMark03 we can also see that there is only an increase of a little over 5%.
When we enable high image quality we can see that in 3DMark05 there is only a 1.6% increase in performance and in 3DMark03 we only see an increase in performance by 3.4%.
As we thought and as our results show, both 3DMark03 and 3DMark05 are much more dependent on the GPU over CPU and memory which only played up to around a 5% difference in performance on our test system setup when we overclocked the processor. This isn't by fault of FutureMark since the benchmark is a 3D benchmark more so than an overall system benchmark, otherwise it would probably be called "SystemMark05" rather than 3DMark.
Find the lowest price on Graphics Cards!