How CPU Features Affect CPU Performance, Part 6
|
Compiling
There is a bump on the graph at 2.66 GHz again, where real performance exceeds the ideal forecast. However, the difference is not very big (see the table with detailed results), about 2%. So we cannot say whether we deal with the above-mentioned phenomenon or with a measurement error. However, this "error" appeared at 2.66 GHz again, so it rings a bell.
Audio encoding
Quite a strange result, which requires additional tests. We have an impression that the test is limited by something else than a processor. Judging by the previous tests, we shouldn't suspect the memory system. Can it be a hard drive?
Video encoding
| |
1.86 GHz |
2.26 GHz |
2.66 GHz |
3.06 GHz |
| Canopus ProCoder ↓ |
05:28 |
04:33 |
20% |
03:39 |
25% |
03:18 |
11% |
| DivX ↓ |
05:58 |
05:02 |
19% |
04:22 |
15% |
03:53 |
12% |
| Mainconcept VC-1 ↓ |
08:34 |
07:09 |
20% |
06:01 |
19% |
05:26 |
11% |
| x264 ↓ |
09:53 |
08:12 |
21% |
07:02 |
17% |
06:10 |
14% |
| XviD ↓ |
03:40 |
03:05 |
19% |
02:39 |
16% |
02:22 |
12% |
| Group Score ↑ |
97 |
117 |
21% |
138 |
18% |
154 |
12% |
It's one of the best scaled groups in this article. Application graphs are also very tight -- all curves, except one, almost merge into one thick line. Strange as it may seem, the leader in this group is the relatively old Canopus ProCoder. However, this phenomenon can be explained with its low capacity to use multiple cores: modern codecs that can use all eight cores must generate higher load on the memory system, and thus they should depend on it as well. And Canopus ProCoder depends solely on a processor.
Write a comment below. No registration needed!
|
|
 |
|
|
|
