Using Skywell Magic TNT drivers
(12/98)
|
640x480 |
800x600 |
1024x768 |
GLQuake demo2 |
101.9 |
68.5 |
44.1 |
Quake II demo1 |
73.0 |
59.2 |
38.4 |
Quake II crusher |
37.8 |
34.5 |
28.1 |
3DMark 99 Lite |
2755 3DMarks |
2123 3DMarks |
N/A |
The performance of the Skywell
Magic TNT was just as high as expected. In almost
all cases, performance was very close to that
of other TNT cards. The OpenGL performance was
almost exactly the same as tests I had previously
done with an STB Velocity v4400 card. Like I said
before, this was expected, so I had no reason
to complain.
However, the Direct3D performance
with 3DMark 99 Lite was not as high as anticipated.
Although it was very close, I was expecting the
card to hit at least 2800 3DMarks using 640x480.
I thought the 800x600 score would reach at least
2300, but it was well over 100 points shy.
I believe the reason for
less than perfect Direct3D performance was based
solely on the drivers. So I ran all the tests
again, this time using NVIDIA's v.48 reference
drivers. This was to prove that with TNT based
cards, the actual card is not as important as
the drivers.
Using NVIDIA v.48 reference
drivers
|
640x480 |
800x600 |
1024x768 |
GLQuake demo2 |
101.4 |
67.9 |
44.0 |
Quake II demo1 |
74.1 |
59.4 |
38.5 |
Quake II crusher |
38.2 |
35.4 |
29.7 |
3DMark 99 Lite |
2852 3DMarks |
2330 3DMarks |
N/A |
The OpenGL results are almost
identical between the drivers, but the 3DMark
score's increased by a significant amount. So,
if you are the owner or future owner of the Magic
TNT, I recommend you either use the NVIDIA reference
drivers or wait for better drivers from Skywell.
Overclocking
No Tweak3D review
would be complete without a little tweaking,
so I decided to try pushing the Skywell
Magic TNT to the limit. Keep in mind that
this card has no cooling other than a
heatsink on the TNT chip. The memory is
8 ns 125 MHz SDRAM.
Here are the 1024x768
overclocking results. 90/110 is the default
setting (90 MHz core, 110 MHz memory)
The numbers
listed are the core speed/memory speed.
Results are provided in the middle. (the
only cooling used was the heatsink!)
|
90/110 |
95/115
|
100/115 |
105/1155 |
100/120
|
100/125 |
100/130 |
GLQuake
demo2 |
44.0 |
45.4 |
46.6 |
failed |
48.2 |
49.7 |
51.1 |
Quake
II demo1 |
38.5 |
39.5 |
40.6 |
41.7 |
41.3 |
42.3 |
failed |
Quake
II crusher |
29.7 |
30.0 |
30.5 |
failed |
30.8 |
31.1 |
failed |
Due to overheating,
no test passed with the core speed higher
than 105 MHz, and only the Quake II demo1
test passed at this speed. Due to memory
restrictions, no test passed with the
memory speed higher than 130 MHz, and
only the GLQuake demo2 test passed at
this speed.
Conclusion