Tests
For your convenience the results are represented in percent, 100% being the result of AMD Athlon II X4 620. Absolute results are provided in this Excel spreadsheet.
3D visualization
The very first tests brings the first discoveries. As we already know, such tasks do not require a lot of threads, they depend on the speed at which these two or three threads are handled by processor. That's exactly where the beneficial effect of architecture optimizations may show. And it does: even the Core i5-2300, the cheapest and lowest-end of the new series, outperforms all processors we've tested before, including the previously unconquered Core i7-975 Extreme Edition. Obviously, the rest of the new lineup is even faster, so there's just nothing these processors could compete with.
3D rendering
We believe that Sandy Bridge will have the last word in these tasks after they start supporting AVX instructions. For now this is pure math, very well parallelized at that, so the more threads the better. However, the high efficiency of each thread shows here as well. In particular, the new Core i5 outperform the old ones — with the same amount of cores and at comparable clock rates — by about 10%. Don't forget that the boosted Core i5-760 works at 2.93GHz, while the boosted Core i5-2300 works at 2.9GHz. But the transition to a finer process technology allows the novelties to work at higher clock rates and allows them to compete with both the old Core i7 and hexacore Phenom II X6 CPUs. Even despited the higher clock rate of the latter. As for the hexacore Core i7 solutions, they are naturally unreachable, but they also cost much more. So the second place of the Core i7-2600 is actually a brilliant victory.
Scientific calculations
Most of these tests also have few threads, although there are certain multi-threaded inclusions. The first two places are occupied by LGA1155 solutions. The two of those sharing the first place proves that Hyper-Threading still isn't for free. The cheaper Core i5-2300 only yields to the much more expensive Extreme Editions.
Image editing
We have already mentioned that applications of this group have diverse predilections: Adobe Photoshop favors large numbers of threads, while the three non-professional programs have no need for that many (and may even suffer from such amounts of threads). And since it's three versus one, there's no surprise that the older dual-core high-clock-rate Core i5-600 have used to perform very well in these tests, outdone just by Extreme Editions. So the Core i5-2000 lineup does an even better job, Adobe Photoshop included. It seems we've got the new leaders. The performance of the Core i7-2600 is especially remarkable, because it almost catches up with the much more expensive hexacore Core i7-970 in Adobe Photoshop and is simply unrivalled in the three remaining applications. The Core i5-2400 performs on a par with the Core i5-680 (the previous leader), but outperforms it by almost 1.5 times in Adobe Photoshop, landing among previous leaders. The Core i5-2500 is obviously faster and yields only to the Core i7-2600. In other words, only the junior Core i5-2300 isn't as amazing. But considering that its wholesale price is just $177 and that it competes with processors $100-$400 more expensive, it's still a remarkable result.
Data compression/decompression
As you probably know, 7-Zip can use every core it finds. And all the three tests favor large cache, the latter being especially sensitive to it. So there's no surprise that the new Core i5 processors do not look as impressive here with just four threads and cache reduced to 6MB. But that doesn't mean they perform badly, the novelties easily outperform both AMD solutions and manage to finish close to the old Core i7 CPUs, which costs about $100 more. As for the new Core i7-2600, it both supports Hyper-Threading and has 8MB of cache, so it's only rivalled by the Core i7-980X Extreme Edition.
Compiling
It seems that Visual Studio doesn't favor the new processors above all. Maybe it's because compiling has already been perfectly optimized. However, the Core i5-2300 still outperforms the Core i5-760 a bit, despite smaller cache which is important in this test. This slight performance boost has a strategic meaning: as you might remember, the Phenom II X6 series has used to rule in this application, outperforming the Core i5 lineup and catching up with the junior Core i7 processors. Now the quad-core Core i5-2400 performs on a par with the hexacore — and more expensive — Phenom II X6 1055T. The Core i5-2500 is outperformed by the Phenom II X6 1075T by just one point. The higher-end AMD models remain faster than the new Core i5, they can be compared with Intel's old processor at about $300. But the newly offered one is far ahead, only lagging behind Intel's own hexacore Core i7-980X Extreme Edition by only 10%.
Java
Well, SPECjvm produces somewhat unusual results. We used to name this test as a good example of multi-core optimization. But it seems it's only true for up to 8-10 threads. This has allowed processors with more threads to win in the competition of CPUs with similar architectures and different numbers of cores. But as soon as we have compared models with different performance per thread... The Core i7-980X Extreme Edition is still the fastest, but its superiority over the Core i7-2600 is purely formal. As for the Core i5-2400, it almost catches up with the Core i7-880 despite the face that the latter has twice as many threads and similar clock rate.
Such a performance boost has utterly defeated AMD processors. The Phenom II X4 970 has used to outperform any Core i5 CPU, while Phenom II X6 1090T has been faster than any Core i7-800. Now the Phenom II X4 970 yields to any Core i5 LGA1155, and Phenom II X6 1090T loses to the Core i5-2500 (!). Obviously, AMD's hexacore solutions cannot compete with the new Core i7 LGA1155.
Write a comment below. No registration needed!
