 |
Untitled Document
Synthetic benchmarks and quick performance comparisons..
About the benchmarks
For those who are not familiar with the benchmarks used here, a quick explanation
is in order. 3dmark 2001 and 3dmark 2003 are synthetic gaming benchmarks based
on a range of real game engines. The scores are only comparable with other scores
from the same version. The main difference between the 2001 and 2003 versions
is that 2003 uses DirectX 9, whilst 2001 uses 8.1.
Aquamark is a synthetic gaming benchmark using DirectX 9 and the Krass game
engine.
All of these benchmarks are run at default settings, 32 bit colour with no
antialiasing, at 1024 by 768. This table shows approximately what scores are
needed for gaming at these settings, with low, medium, and high detail levels.
| |
Low |
Medium |
High |
| 3dmark 2001 |
4,000 |
8,000 |
12,000 |
| 3dmark 2003 |
1,000 |
2,500 |
4,000 |
| Aquamark 3 |
3,000 |
7,000 |
20,000 |
The Results and Comparisons
The difference column shows the difference between the 3.2 GHz score and the
1.6 GHz score, as a percentage, giving you some idea what the performance penalty
is like for running each card on a slower system.
The Quick Comparison column shows the relative performance of each card, at
3.2 Ghz, with the fastest card being shown as 100%, and the rest as a percentage
of that score.
The Combined Quick Comparison is the average of all three percentages, giving
a comparison that encompasses both DirectX 8.1 and DirectX 9.0 scores.
| 3dmark 2001 |
|
3dmark 2003 |
|
Aquamark 3 |
|
Combined Quick
Comparison |
| |
1.6 GHz |
Difference |
3.2 GHz |
Quick Comparison |
|
1.6 GHz |
Difference |
3.2 GHz |
Quick Comparison |
|
1.6 GHz |
Difference |
3.2 GHz |
Quick Comparison |
| Asus 9800XT |
12,197 |
33% |
18,286 |
100% |
|
5,897 |
8% |
6,423 |
100% |
|
35,245 |
23% |
46,156 |
100% |
|
100% |
| Sparkle 5900 |
10,702 |
33% |
15,870 |
86% |
|
4,297 |
9% |
4,889 |
76% |
|
34,260 |
11% |
38,168 |
82% |
|
81% |
| Gainward 5900 |
10,775 |
31% |
15,703 |
85% |
|
4,199 |
11% |
4,717 |
73% |
|
33,897 |
10% |
37,566 |
81% |
|
79% |
| Asus 9600 XT |
10,515 |
20% |
13,283 |
72% |
|
3628 |
5% |
3,838 |
59% |
|
28,395 |
10% |
31,308 |
67% |
|
66% |
| Sapphire
9600 pro |
9,806 |
17% |
11,885 |
64% |
|
3,240 |
4% |
3,397 |
52% |
|
24,374 |
7% |
26,188 |
56% |
|
57% |
| Sparkle 5600 |
8,277 |
20% |
10,345 |
56% |
|
2,116 |
5% |
2,223 |
34% |
|
17,285 |
1% |
17,331 |
37% |
|
42% |
| Terratec 5600 |
7,939 |
20% |
9,872 |
53% |
|
1,998 |
4% |
2,081 |
32% |
|
16,840 |
1% |
16,876 |
36% |
|
40% |
| Terratec 5200 |
5,185 |
7% |
5,555 |
30% |
|
873 |
4% |
900 |
14% |
|
7,669 |
0% |
7,669 |
16% |
|
20% |
A few common trends are shown up by these figures. Higher end cards are more
heavily crippled by a slower PC, but slow cards do not perform significantly
better with a fast PC. It seems that the speed of the PC only acts as a bottleneck,
which will determine the upper limit of your possible graphics performance,
unless the limits of the graphics card are reached first.
We can also see that 3dmark 2003 is a less CPU sensitive benchmark than 3dmark
2001 and Aquamark is even more CPU independent. This is useful as we have more
chance of testing the graphics card rather than the overall system performance.
<< Back | 1 | | 2 | | 3 | | 4 | | 5 | | 6 | | 7 | | 8 | | 9 | | 10 | Next >>
|
|
