Intel Core 2 Extreme QX9770 in Real-Life Apps
|
Scientific applications
|
DDR3-800 |
DDR3-1600 |
Gain |
X4 9850 |
Maple |
0,0314 |
0,0314 |
0% |
1% |
Mathematica |
Internal |
4.3570 |
4.5320 |
4% |
4% |
MMA |
1.6172 |
1.6452 |
2% |
2% |
MATLAB |
LU |
0.0482 |
0.0438 |
10% |
4% |
FFT |
0.1204 |
0.1041 |
16% |
6% |
ODE |
0.1487 |
0.1486 |
0% |
2% |
Sparse |
0.2587 |
0.2355 |
10% |
3% |
2D |
0.2076 |
0.2070 |
0% |
2% |
3D |
0.1026 |
0.1025 |
0% |
0% |
Group score |
134 |
140 |
5% |
3% |
MATLAB provides the most interesting results: there are three tests here, which demonstrate 10% or higher performance gains with the QX9770, and two tests, which demonstrate some performance gain with the X4 9850 and no gain with the QX9770 at all. On the whole, response of both systems to the increased memory performance reveals more differences than similarities. Only one explanation comes to our mind: different cache sizes and organizations in Core 2 Extreme QX9770 and Phenom X4 9850 make them respond differently to changes in stream access speed and latency. We can assume that relatively high latency on the Intel platform results in deeper performance drops. In other words, it's not that DDR3-1600 is fast, it's DDR3-800 that makes some operations too slow.
Web server
|
DDR3-800 |
DDR3-1600 |
Gain |
X4 9850 |
PHP Calculator |
244 |
246 |
1% |
0% |
PHPSpeed |
Synthetic PHP |
3853 |
3877 |
1% |
0% |
Synthetic MySQL |
2973 |
2976 |
0% |
0% |
Synthetic Read/Write |
2405 |
2411 |
0% |
4% |
Real World PHP |
6637 |
6764 |
2% |
4% |
Real World PHP & MySQL |
4312 |
4334 |
1% |
4% |
Server |
2526 |
2549 |
1% |
2% |
Group score |
137 |
138 |
1% |
2% |
QX9770 depends less on memory performance in this group of tests. And this tendency appears practically in all tests here. As in compile tests, the easiest explanation to such behavior is relatively large L2 Cache.
Archivers
|
DDR3-800 |
DDR3-1600 |
Gain |
X4 9850 |
7-Zip |
0:02:07 |
0:01:50 |
15% |
10% |
WinRAR |
0:01:04 |
0:00:56 |
14% |
7% |
Ultimate ZIP |
0:01:39 |
0:01:39 |
0% |
0% |
Group score |
125 |
137 |
10% |
6% |
Both systems are very sensitive to memory bandwidth. "Champion tests" are absolutely identical, but performance gains on the Intel platform are significantly higher.
Encoding
|
DDR3-800 |
DDR3-1600 |
Gain |
X4 9850 |
FLAC |
0:00:54 |
0:00:53 |
2% |
0% |
LAME |
0:01:26 |
0:01:26 |
0% |
0% |
Musepack |
0:01:34 |
0:01:34 |
0% |
0% |
Vorbis |
0:02:44 |
0:02:43 |
1% |
0% |
Canopus ProCoder |
0:04:35 |
0:04:31 |
1% |
2% |
DivX |
0:00:55 |
0:00:54 |
2% |
4% |
x264 |
0:01:54 |
0:01:54 |
0% |
2% |
XviD |
0:05:30 |
0:05:28 |
1% |
1% |
Group score |
137 |
138 |
1% |
1% |
No comments needed.
Games
|
DDR3-800 |
DDR3-1600 |
Gain |
X4 9850 |
Call of Duty 4 |
119 |
130 |
9% |
0% |
Company of Heroes |
57 |
57 |
0% |
2% |
Call of Juarez |
51 |
52 |
2% |
2% |
Crysis |
24.48 |
25.57 |
4% |
5% |
S.T.A.L.K.E.R. |
128 |
128 |
0% |
4% |
Unreal Tournament 3 |
126 |
145 |
15% |
22% |
World in Conflict |
60 |
65 |
8% |
10% |
Group score |
109 |
115 |
5% |
6% |
Strange as it may seem, the Intel platform is less sensitive to memory bandwidth in games, except for Call of Duty 4. Why does Phenom "ignore" this game? We can only repeat our above-mentioned assumption: Phenom is OK with access latencies in both cases, while QX9770 responds well to their reduction.
Non-professional photo processing
|
DDR3-800 |
DDR3-1600 |
Gain |
X4 9850 |
ACDSee |
0:06:54 |
0:06:34 |
5% |
1% |
IrfanView |
0:13:32 |
0:13:30 |
0% |
1% |
Paint.NET |
19150 |
18966 |
1% |
1% |
xat.com Image Optimizer |
0:28:13 |
0:28:05 |
0% |
1% |
XnView |
0:15:48 |
0:15:44 |
0% |
1% |
Group score |
131 |
136 |
1% |
1% |
ACDSee demonstrates a strange result, but it does not affect the overall situation much. In fact, when we speak of the 1-2% difference, you shouldn't forget about inevitable measurement errors.
Conclusions
|
DDR3-800 |
DDR3-1600 |
Gain |
X4 9850 |
PRO SCORE |
132 |
137 |
3% |
3% |
HOME SCORE |
125 |
131 |
4% |
3% |
OVERALL SCORE |
129 |
134 |
4% |
3% |
In broad terms, we acknowledge that Intel Core 2 Extreme QX9770 benefits from using faster memory in more cases than AMD Phenom X4 9850. In other words, QX9770 is more dependent on memory performance. Feel free to choose any definition of these two, depending on whether you want to praise QX9770 or criticize it.
However, these are purely theoretical conclusions, scores and percentages, even their fractions. In practice, it will be very difficult to notice the difference between DDR3-800 and DDR3-1600 even with the fastest quad-core processor from Intel. A couple of games, archivers, some activities in Photoshop and MATLAB -- that's the complete list of situations, where you can try to feel a slightest performance gain. It's up to you to decide whether you want to pay much money for this transient sensation.
Write a comment below. No registration needed!
|
|
|
|
|