Call of Duty: (OpenGL)
Map: VL_brecourt
| 1024x768 |
| |
NoAA/AF
|
2xAA/AF
|
4xAA/AF
|
4xAA/AF
|
| N5900 |
59.1
|
58.7
|
57.5
|
50.9
|
| 9600XT |
106.4
|
96.5
|
97.9
|
96.2
|
| FX5600U |
81.3
|
51.7
|
13.8
|
13.6
|
| 1280x1024 |
| |
NoAA/AF
|
2xAA/AF
|
4xAA/AF
|
4xAA/AF
|
| N5900 |
58.2
|
54.9
|
52.5
|
50.9
|
| 9600XT |
79.3
|
67.8
|
68
|
67.2
|
| FX5600U |
56.4
|
38.6
|
27.1
|
26.4
|
This is Viper Lair's homemade demo so we could get
a more realistic gameplay result. The results of which, well,
lets say they were less than flattering to the N5900.
As you can see, the 9600XT beat up on the N5900
hands down, not by just a little bit mind you, we are talking
close to a 45% gain here. Even the FX5600U managed a better frame
rate then the N5900. Not only did I run this test with our demo,
I ran it with demobta00 (from Benchemall)
only to get the same results. Different drivers, be it ASUS specific
or NVIDIA returned the same puzzling results as well. The only
conclusion that I can come to, is there is something in the PCIe
scenario and how CoD draws out that causes an issue...
Quake III: (OpenGL)
Map: four.dm_67
| 1024x768 |
| |
NoAA/AF
|
2xAA/AF
|
4xAA/AF
|
4xAA/AF
|
| N5900 |
344.9
|
315.4
|
289.1
|
281
|
| 9600XT |
202.1
|
174
|
174.9
|
174.8
|
| FX5600U |
141.2
|
123.8
|
99.8
|
99.7
|
| 1280x1024 |
| |
NoAA/AF
|
2xAA/AF
|
4xAA/AF
|
4xAA/AF
|
| N5900 |
307.2
|
245.8
|
207.5
|
201.4
|
| 9600XT |
144.7
|
115.8
|
115.8
|
114.6
|
| FX5600U |
111.9
|
98.1
|
76.8
|
75.1
|
The framerates here are high, even with AA/AF turned
up we get nice returns from all of the graphics cards.
Even though CoD uses a modified Quake 3 engine, it appears the
N5900 has no issues in pumping out the framerates in the original.
Over 200 FPS with 4xAA/8xAF is very impressive, a near 100% gain
once again over the 9600xt and almost 150% gain over the FX5600U.
Overclocking
ASUS was nice enough to supply us (and you) with the software
required to overclock the N5900 on the fly. I tested the GPU first
and found it's maximum by slowly ramping up the frequency and
running a few benchmarks looking for artifacting. I used UT2K4
for the testing as it has a brighter and rich color which shows
artifacting much quicker then the other tests. Once I attained
my maximum GPU performance I then reset to base and did the same
for the memory. From this I abstracted a best case scenario of
GPU:memory overclock and proceeded to benchmark from that point.
Since we had the hardest time with Far Cry and Doom 3, I figure
those are two good games to test the overclock with (nicely fitting
one DirectX based and one OpenGL based as well) The best scenario
I was able to achieve was a 390MHz GPU paired to a 728MHz Memory.
I could go much further, with both mind you, this produced the
best results.
| Doom 3 - 1280x1024, No AA/AF |
| |
1024x768
|
1280x1024
|
| 375/700 |
39.7
|
27.8
|
| 390/728 |
21.7
|
13.9
|
| Far Cry - 1280x1024, No AA/AF |
| |
1024x768
|
1280x1024
|
| 375/700 |
41.11
|
39.6
|
| 390/728 |
27.9
|
18.67
|
A very disappointing showing period! On Doom 3 we lost performance
at any rate of overclock attempted. Inside the game, it felt choppier
than at stock speed and during heavy action sequences it had moments
of pure freeze and release. Far Cry faired about the same, no
overclock even matched what we were able to do base. This is a
very perplexing turn of events. I have never seen an issue where
overclocking produces negative results, and it makes me wonder
why. Did ASUS do something with their rendition of the GPU to
give it the greatest performance at stock speeds? Is this why
they slowed the GPU down. Unfortunately I can not answer any of
these questions within this reviews time frame. As always, overclocking
can damage your equipment, you use the included ASUS software
and our results at your own risk.
NEXT
