|
Posted by Thomas
McGuire on December 07, 2001
Manufacturer: Abit Product:
KG7-RAID Socket A Motherboard
System performance
was tested with a few varied applications, which hopefully
will cover a few different aspects of the system. It’s
worth noting though that given XP is only relatively
recently released a lot of the Drivers may not be as well
optimised as could have been hoped for (Compared to say,
using Windows 98 instead).
Serious
Sam was used to test both CPU & Graphics Card
performance, with the Dunes timedemo being the Graphically
reliant one & Memphis Suburbs relying on the CPU.
Partially thanks to Croteam
there was a bit of a deviation in the settings used due to
the recent release of the 1.05 patch. Initially benchmarks
were run on the Vivid! XS using the settings that can be downloaded
here (As used in the Vivid! XS review). Results being as
follows;
Map
|
Dunes
|
Memphis
Suburbs
|
Audio
Mode
|
EAX
|
DirectSound
|
Disabled
|
EAX
|
DirectSound
|
Disabled
|
Resolution
|
|
|
|
|
|
|
|
1024
x 768 x 32
|
87.2
|
88.7
|
89.8
|
66.6
|
65.1
|
70.7
|
|
800
x 600 x 32
|
96.6
|
97.7
|
101.2
|
68.6
|
68.5
|
73.2
|
|
640
x480 x 32
|
99.4
|
99.5
|
103.1
|
68.5
|
68.9
|
73.5
|
Next
up was the Creative Geforce 3 Titanium 200 which used the
1.05 version of Serious Sam. Given changes to the PersistentSymbols.ini
files which stores configuration data I felt it best to
create new settings to test with instead (Which can be
summed as calling them higher quality image settings). These
can be downloaded
here. Now for the results;
Map
|
Dunes
|
Memphis
Suburbs
|
Audio
Mode
|
EAX
|
DirectSound
|
Disabled
|
EAX
|
DirectSound
|
Disabled
|
Resolution
|
|
|
|
|
|
|
|
1024
x 768 x 32
|
104
|
104.5
|
109.2
|
82.9
|
83
|
88.1
|
|
800
x 600 x 32
|
107.9
|
108.8
|
114
|
85.1
|
84.1
|
89.7
|
|
640
x480 x 32
|
109.5
|
109.5
|
115.8
|
85.1
|
84.9
|
90.7
|
Results
are good, with the Geforce 3 out performing the Kyro 2, as
you could have expected – even though it is using settings
which provide better image quality.
Further
testing OpenGL performance I used the timedemo provided in
the latest Quake 3 patch. Settings used can be downloaded
here. Unlike the Serious Sam tests the exact same
version & settings were used for both the Kyro 2 &
Geforce 3.
Graphics
Card
|
Vivid!
XS Kyro 2
|
Creative
Geforce 3 Ti 200
|
|
Audio
Mode
|
High
quality
|
Disabled
|
High
quality
|
Disabled
|
|
Resolution
|
|
|
|
|
|
1024
x 768 x 32
|
93.9
|
95.8
|
119.7
|
146.1
|
|
800
x 600 x 32
|
111.5
|
131.8
|
121
|
154.5
|
|
640
x 480 x 32
|
112.5
|
142.4
|
121.1
|
156.8
|
Again,
some good results with both Graphics cards offering high
frame rates. Although as can be seen most prominently with
the Geforce 3 results the Audigy WDM Drivers seem to be
lowering the frame rate quite a lot (Also noticeable on the
Kyro 2 at lower resolutions). Some frame rate loss was to be
expected with Sound enabled, although suffering a 30 frame
rate hit just to playback 22 KHz, 16-Bit seems just a bit
unreasonable. This, of course, could be a Game problem too
potentially – compare the results to Serious Sam which
suffered a minimal 5 or 6 frame reduction from Audio being disabled
to using 44KHz, 16-Bit, EAX sound.
3D Mark 2001 was
untested given issues with it & Kyro Graphics cards (The
High detail Lobby test for example, runs with everything
rendered regardless of visibility – this of course will
adversely affect the results obtained – rendering them
useless). Although this issue has been resolved, Mad Onion
haven’t & won’t be releasing the patch with this
update in it anytime soon, nor are they release the Kyro fix
separately. You can read further details here.
As a result 3D Mark 2001 is effectively a useless Benchmark
program to use for testing this system (Or anyone with a
Kyro 1/2 Graphics card in it for that matter).
For the same reasons
stated above the Geforce 3 was also untested in 3D Mark
2001, again, the program cannot provide a valid comparison
& so I feel any results obtained by the Geforce 3 would
be similarly unable to be used for comparative purposes.
Power
VR’s Villagemark
benchmark was used to gauge Direct3D performance with the
Motherboard & Graphics cards. This benchmark uses a good
deal of overdraw & so is a good test of any culling
optimisations a Graphics card may have & also supports
Hardware T&L and DXTC.
|
Resolution
|
Vivid!
XS Kyro 2
|
Creative
Geforce 3
|
|
640
x 480 x 32
|
154
|
106
|
|
800
x 600 x 32
|
165
|
79
|
|
1024
x 768 x 32
|
132
|
55
|
Here the Kyro 2 performs a great deal better than the Geforce 3 (Well
over 50% higher frame rates at 640 x 480 to a whopping 240%
higher at 1024 x 768). Whereas the Kyro 2 features HSR
(Hidden
Surface
Removal)
via Tile based rendering (Using the on-chip tile buffer) the
Geforce 3 supports Z-Occlusion
Culling
for HSR. So if both support HSR then why the large
difference? Well, its all down to order, more specially the
HSR on the Kyro isn’t order dependant, whereas the Geforce
3’s HSR is. To put things into perspective the majority of
Games use a more random rendering order & so cannot
benefit from the HSR available with the Geforce 3.
|
