The new benchmarking procedure will be described in a separate article in the near future. So today we'll offer a kind of a beta preview that will include test results obtained with the new benchmarks. Note that the final version of the new test method will probably include several new tests and some other additions. These will also be described in due time. To make the charts easier to read, the results are represented in percents (in each test 100% is equal to the result of Intel Core 2 Quad Q9300). The detailed results in absolute values are provided in this Microsoft Excel spreadsheet.
3D modeling and rendering
Probably the most remarkable fact about this diagram is the advantage of DDR3 for Phenom II X4 955. As CPU performance grows, fast memory should become increasingly efficient. As you can see, sometimes it happens quite dramatically. The 955 is only 400 MHz as fast as the 925, but benefits of the new memory type are much lower for the latter. By the way, it's an indirect sign that Phenom II scales well as far as its clock rates are concerned. From the practical point of view it means that future models (and overclocked modern ones) will always benefit from DDR3 memory.
What concerns competition, both Phenom II processors fare very well against representatives of the Core 2 Quad family. However, the i7 920 actually outperforms the top Phenom II, but not too much to draw any practical conclusions.
In our previous test procedure this group of tests and professional 3D applications demonstrated similar results. However, our applications updated, the situation has changed. We should mention results of the Phenom II X4 955 with DDR3. As we already said, our Pro/Engineer test uses more than 4GB of memory, so the following comparison with DDR2 is not posed well. Results of this group are lower (113%) in case of a 4-GB memory kit and the same timings. Even the cheaper processor benefits noticeably from using DDR3 memory.
The new compile test actively uses multiprocessing, and the Core i7 demonstrates the best result. By the way, it gets the maximum score here compared to the other tests, as it compiles the source code 1.5 times as fast as the Q9300. The i7 deserves the praise only partially, because the Q9300 won't set the Thames on fire here, so the other processors look well on its background, starting from the Phenom II X4 925. And the top Phenom II does not outperform the Q9550 very much, as the latter takes advantage of its huge 12-MB cache.
Bitmap graphics editors are united into one group in the new test procedure. We've also added two new editors to accompany Adobe Photoshop. Our constant readers may remember that Phenom II is not doing great here. However, it fares well against Core 2 Quad. On the contrary, many of these tests, which can take advantage of multiple cores and Hyper-Threading, favor the i7.
Phenom II puts up excellent performance in our scientific tests, where its floating-point unit may show its worth. Unlike the other groups of tests, where we can praise the 925 model that easily competes with the Q9300, it's the 955 processor that shines here, as it outscores both competitors, even if by a minimal value. In fact, we would have bet on the Q9550 here, as most mathematical tests simply adore large caches, while the total cache size in Phenom II in smaller than in both competitors.
Note that the new version of the MATLAB test is much more stable and predictable in its preferences (that is, it still depends on CPU and memory performance, but not on random factors that distort results of the built-in benchmark, which is not used anymore).
The new test procedure includes SPECjvm2008 test. We can assume that execution of programs written with languages of such a high level, that is separated from hardware, offers lots of opportunities for such technologies as Hyper-Threading and Turbo Boost. That is everything that can benefit from the relatively disorganized operation of CPU cores and their units. So Core i7 shoots forward again, and both Phenom II processors can only boast of being better than Core 2 Quad in random memory access mode. The cheaper Phenom II model almost catches up with the Q9550.
On the other hand, it's difficult to name applications and situations, which could be limited even by much weaker processors than ours, so that we could wish higher performance in Java. However, several users asked us to add something like this to our test procedure. So said so done.
Write a comment below. No registration needed!