The benchmarking procedure (the list of software and test conditions) is described here. To make the diagrams easier to read, results are represented in percents (100% stands for the result of Intel Core 2 Quad Q9300 in each test). Detailed results in absolute values are published in this Microsoft Excel spreadsheet.
As we expected, there are no performance gains from more than two cores. But the clock rate and CPU architecture do matter. The same concerns cache. However, insufficient cache can be compensated by a higher frequency -- the E7400 catches up with the E8200, while the E7600 outperforms it. So there is nothing surprising that Intel abandoned the expensive E8200 as soon as the same performance level was reached by cheaper processors with higher clock rates. In other respects, we can see that even inexpensive processors suffice for professional 3D Modeling packages. It's true, when a given computer is used solely for its creative component, of course, and the final renders are performed on a dedicated computer or even on a whole render farm.
No dual-core processor can compete with high-speed triple-core and even slow quad-core models here. Performance ratios are so big that in order to catch up with at least the Q8200, dual-core processors would have to operate at 4 GHz, while their current maximum is 3.33 GHz. We do not accept the remark that a Core 2 Duo can be overclocked to this level -- Core 2 Quad can be overclocked as well, and Phenom II X3 720 even has an unlocked multiplier. As we expected, such tasks need as many cores as you can get. Performance gains in rendering are demonstrated even in those cases, when we install eight physical cores running 16 threads simultaneously (e.g. two Xeons). The saturation point in desktop system is even further. Performance gains are not linear, and the clock rate also affects the results (so the Q8200 and the Х3 720 demonstrate almost identical results). But the general picture is apparent.
Scientific and engineering analysis
This group of applications also demonstrates a clear tendency, but it does not favor multiple cores: it's better to have a couple of cores operating at a higher clock rate. Besides, 2M cache is apparently too small, and it spoils results of the Pentium or Core 2 Quad Q8000, to say nothing of the Athlon II, where this cache is shared between cores and cannot be used to exchange data. But over 3M seems to be excessive. Differences between processors are so small that it's not rational to make a choice on the ground of this group of applications -- even a Celeron will be doing fine here, even though it's a serious group of programs, not just home multimedia.
We noticed some performance gains from the increased number of cores, but they are not significant. Results? Pentium E6300 demonstrate the same performance as Core 2 Quad Q8200, and Core 2 Duo E7600 catches up with Core 2 Quad Q9300. Dual-core processors operate at a higher clock rate, of course, than quad-core processors with similar results. But the difference is not that big to consider the latter better suited for such tasks. To put it shortly, even Mid-End dual-core models will suffice for bitmap graphics. And practically the only factor that can turn you against them is the reign of quad-core processors among top performance solutions. This is all familiar -- Intel and AMD used the same method to push single-core processors out of the market. But now it's done in a milder fashion -- for example, top models of Core 2 Duo have higher frequencies than "composite" models so far. Sometimes this difference is quite big, so than can keep their face. However, this tendency is quite pronounced already. It concerns even old familiar platforms, to say nothing of promising ones -- for example, there are already three quad-core processors for LGA1156, dual-core models will appear next year.
These tasks do not need more than two cores, unlike much cache, so our winner is Core 2 Duo E8200. But our comparison of the E7400 and E7600 makes us regret about DDR3 in LGA775. As we have already found out, even the upgrade from DDR2 1066 to DDR3 1333 decreases performance in this group of tests, and DDR3 memory yields really bad results for processors with FSB 1066: such memory frequency is attainable for DDR2 as well, memory bandwidth will be the same, but latencies will be much lower. Why don't we see the same fiasco with the Pentium? The E5300 has FSB 800, and it was tested with DDR2 800. So the E6300 can objectively "stomach" faster memory. But in this case, "all steam was wasted on the whistle" to make up for the harmful effect of DDR3. As a result, we've got equal terms (performance gains appear due to the higher clock rate), that's something at least.
The number of cores, their clock rate, and cache size (to some degree) -- these are components of success. When at least two of these components are present at once, results grow significantly: that's not the first time when we see a high-clocked triple-core processor from AMD compete on a par not only with Core 2 Duo (that's its target competitor), but also with low quad-core devices from both companies. Dual-core processors are much slower. Any of them, especially our Pentium. Huge cache capacity allows the Core 2 Duo E8200 win the whole 400 MHz that separated it from the top representative of Series E7000.
Write a comment below. No registration needed!