|
Untitled Document
Chipset Performance Comparison
Let's move on to some performance comparisons between the different chipsets
on offer. We have used a range of tests, mostly focused on gaming, because this
is the most common use for these boards. Just like the previous
tests, these are tested on a 9800XT with no AA or AF.
Sisoft Sandra 2004
| |
nForce 3 250 |
SiS 755 |
VIA K8T800 |
| Memory
Bandwidth |
3023 MB/s |
2524 MB/s |
3021 MB/s |
You can see here that the SiS chipset is lagging behind considerably. This is
because our particular test board was buggy and would only run at 333MHz memory
speed even when 400MHz is selected (despite the chipset being quite capable) and
this will unfortunately skew the rest of the results. The nForce 3 and Via Chipset
results show just how independent memory performance is of the north bridge.
Sysmark 2002
| |
nForce 3 250 |
SiS 755 |
VIA K8T800 |
| Internet
Content Creation |
368 |
356 |
353 |
| Office
Productivity |
241 |
244 |
241 |
Sysmark results are incredibly close across the board, even on the SiS,
showing how little disadvantage 333MHz memory gives. Results were averaged
across 3 runs because there is as much as a 10 point variance in results,
this makes it hard to draw any conclusions.
3D Mark
|
|
nForce 3 250
|
SiS 755 |
VIA K8T800 |
|
Overall
3D Mark |
6454 |
6353 |
6467 |
|
CPU
test 1 (800x600) |
76.6 |
73.1 |
81.4 |
|
CPU
test 2 (800x600) |
10.9 |
10.1 |
11.3 |
Once again, we see very little difference, only enough to attribute to result
variance. The Via chipset does seem to have a slight edge though.
UT2004 Software Mode
| |
nForce 3 250 |
SiS 755 |
VIA K8T800 |
| 800x600 |
32.34 |
28.39 |
31.99 |
| 1024x768 |
21.99 |
21.00 |
23.98 |
| 1280x1024 |
17.25 |
14.90 |
17.07 |
| 1600x1200 |
12.99 |
11.17 |
12.86 |
This test really stresses the CPU and memory quite a considerable amount and
due to the software rendering without letting the GPU hold it back. We can see
results are very similar across the board, with nothing worth mentioning.
Far Cry
|
|
nForce
3 250 |
SiS 755 |
VIA K8T800 |
|
800x600 |
93.92 |
90.46 |
101.21 |
|
1024x768 |
94.21 |
90.80 |
101.79 |
|
1280x1024 |
92.68 |
90.40 |
101.32 |
|
1600x1200 |
80.02 |
79.97 |
81.54 |
Last of all, we have Far Cry performance. In the lower resolutions, the VIA
chipset has a consistent lead of around 10% which is excellent. In 1600x1200
it is almost the same across the board, where the GPU becomes the limiting factor.
The results have been very interesting, in most cases variance in results can
be attributed to the differences we are seeing. But, in Far Cry and 3DMark,
the Via chipset certainly had a minor advantage. There was no advantage in the
UT2004 software tests though, so we can assume that VIA either has better drivers,
or a better AGP system, which is amusing considering nVidia specialise in making
AGP solutions. We wonder if the results would have been the same if we had used
an nVidia card instead of an ATi.
This graphical advantage will probably not actually help anyone in particular.
Most people will push the game settings until the GPU is the limiting factor
rather than the surrounding system, as is shown when running Far Cry at 1600x1200.
Conclusion
As you can quite clearly see, performance differences are small at best. Memory
performance has nothing to do with the chipset and is entirely dependant on
the performance of the memory itself. We can also conclude that HyperTransport
implementation makes very little difference, as bandwidth is already in excess.
In realistic gaming situations, there are no measurable difference between
the chipsets, since so much is GPU dependent. This is a good argument for saving
as much money as you can on the motherboard, to spend on a better graphics card.
If the last 1% of gaming performance is not your top priority, then it would
be wisest to choose your motherboard entirely on features, quality and usability
issues.
On the CPU front, an extra 200MHz showed an overall 8-9% performance increase,
which translated to roughly 4-5% extra gaming performance. So think twice before
shelling out an extra £50 for the next CPU jump when that could also be
saved and put into a better graphics card. This is where the Semprons should
offer excellent value for money.
Be sure to join our mailing list on the front page, to be informed of when
our 10 motherboard group test will follow.
<< Back | 1 | | 2 | | 3 | | 4 | | 5 | | 6 | Next >>
|
