Distribution of benchmark results for Intel i5 2500 CPU
After some questions from users about the i5 2500 CPU, we were doing some research today into benchmark results we have collected.
Over at cpubenchmark.net we have charts showing the average performance of various different CPUs.
The charts are for the most part made up of benchmark results submitted by the general public. For the most common CPUs we might have 10,000+ individual samples, for the very rare CPUs only a few samples.
As these results come from random machines each CPU type has a distribution of results. One would imagine that the results are tightly bunched, with a CPU of a certain type always scoring around the same benchmark result. It turns out this is not always the case. Not all CPUs are equal.
At the extreme ends, the slowest machine may be several times slower than the fastest. At the high end people are overclocking their machines to 5Ghz and above and at the low end people have a collection of issues that kills the performance of their machines.
One question that has popped up several times is why the Intel i5 2500 CPU doesn't perform as well as the Intel i5 2500K in our charts. These CPUs are nearly identical out of the box, with the differences in the integrated video and virtualisation features, not impacting on performance.
The first chart is a graph shows the performance of the 1132 different i5 2500 machines we have tested.
This second chart is a graph shows the performance of the 6479 different i5 2500K machines we have tested. Despite being more expensive the i5 2500K is a much more popular chip. At least among our users.
As can be seen many more people are overclocking the i5 2500K, depending on where you draw the line maybe 30% - 40% of all 2500K owners are overclocking their machines. There are also less badly configured machines. Reflecting the fact that 2500K owners are more likely to be computer enthusiasts. Owner of the i5 2500 don't fair as well with maybe 20% of machines having some type of performance sapping issue.
While the more extreme examples of overclocking are filtered from the averages used in the main charts, mildly overclocked machines can still get into the charts.
These factors combine to lift the overall average of the 2500K above the 2500, despite them being very similar from a specifications point of view.
Similar effects can be seen in our charts with other CPUs that are popular with enthusiasts. They tend to rank slightly better than otherwise equivalent CPUs that aren't as popular with the community.
Update: From today, 2/Feb/2012) we are making changes to the way overclocking is detected to improve the accuracy of the charts. This will move the two averages in the above charts significantly closer.