CPU performance at 3GHZ after normalizing it to single core

Recently I normalized my single core CPU. It offers a series of benchmarks on a single core with HT / off Turbo, for sixteen CPU from the Pentium 4 to 2600K, running at 3GHz. This is not a judgment of actual performance benchmarks, as we all know that the additional cores, Turbo, and Hyper-Threading absolutely make a difference at this point. But it gives a great perspective on how a single-core performance has changed since 2005, helping to contrast dissimilar classes based processor in a clock-for-clock (i.e. several dual-core, multi-core four).

Interestingly the first generation Core E8600 is beating i5 and i3 dual-core in many situations. I would not have imagined. I wonder how he would quads i5. It is also interesting to remember how bad (ie, inefficient, the advantage of clock speed to one side), the P4 was. Well, depending on the reference, the Phenom II is actually much faster: As pointed out Wut Axel, in general, the difference is only 4%, however. The fact is that the differences between the Athlon and Phenom (ie, cache level 3) occur only in certain circumstances (especially games). Not a totally different design.

In favor of AMD, should consider the unique Athlon 64 X 2 designs was much more well-organized than the Pentium 4, AMD, but could not get the clock speeds up to 3GHz quite some time. So we see the fantastic 3800 + 4800 + CPU here, which were leaders in their time, because they simply could not get to 3GHz. I'm sure I could beat Pentium 4 single core banks, even with a deficit of large clock speed, and destroyed Pentium 4 multi-core applications.

The Phenom X4 955 II has about a 10% gain from the start next to an Athlon X4 II 640 and costs $ 20 extra, also you can OC to 4 GHz on average, instead of the average OC 3.7GHz Athlon II X4 640. If you want to take advantage of them, about 15% faster in general, because something some people overlook is that Deneb scales much improved after 3 GHz Propus. According to me it’s a good deal for only $ 20 more. If it is for the average person however it will be better served by the Athlon X4 II 640, due to lower power consumption, heat, and therefore noise.

The Phenom Athlon II and II are more or less the same; the Athlon only lacks the 6 MB of L3 cache. In one core 2 MB level 2 cache on the Athlon is probably enough to keep the CPU from positions where only one core. That's why performance is similar to that of the Phenom II. Once you start adding more cores do not scale II Athlon and Phenom II. Due to the lack of cache means the CPU is spending more time getting data from RAM much slower.

Unfortunately you cannot simply extrapolate these results to how many cores your program uses. Architectures scale differently once you start focusing on the entire chip. For example, the results of Conroe and Kentsfield by the association will not fit as well as results once Nehalem FSB begins to be overloaded. Complete design of Nehalem will be largely based around the creation of a monolithic CPU performance remains high when all cores are stressed.

Only a small number of multi-threaded programs can be treated as part programs that run on its own CPU. Those who do are those where the Phenom II still performs well. Most have dependencies among threads, so what we're doing a thread at this time that thread B can affect you the next moment, and C thread could be waiting for them both. In most programs with multiple threads, cache updates, and reading from the cache now, there may be significant performance problems.