Intel K CPUs have higher results than non-K. Why?

[purple]

According to pretty much everyone, Intel K and non-K CPUs are physically identical with a single difference that the K CPUs are unlocked. Some will also add that K CPUs are those that achieve the top testing results. So why is there a noticeable benchmark result difference (on Passmark charts) between non-K and K CPUs that both run at the same clock speeds? One example is i7-2600 vs i7-2600k. Single Thread Rating: 1919 vs 1943. CPU Mark: 8239 vs 8496. There is a 3.1% increase in CPU Mark score from 2600 to 2600k. Some might say that this increase is negligible but then again when comparing 2600k to 2700k (which actually is an upgrade) there is a 3.5% increase in performance. Similarly, Single Thread Rating increase from i5-3570 to i5-3570k: 0.8% and from i5-3570k to i7-3770: 2%

So how is this performance difference between K and non-K CPUs explained? Are the K CPUs really performing better? Is it worth the extra money to buy a used K when not intending to overclock just because they perform better on benchmarks, or better to be sure that I'll get a CPU that has never been overclocked?

[David (PassMark)]

3% is negligible. Some upgrades also offer negligible performance improvements as well.

But you are right, the K units are often very slightly better in our benchmark, but they shouldn't be (in theory).

My theory is that

1. The margin for error accounts for some of it. Especially when we don't have a large number of samples. This can swing results up or down however.
2. The people who buy the "K" parts care for their PC more. So spend more time making sure it is all running nice. e.g. no crapware running in the background.
3. We exclude overclocked results from the averages, BUT very small overclocks (e.g. a 1% increase in clock speed) do not exceed the threshold we have set for declaring a CPU as overclocked and so these units get included in the main average.
4. People who buy the "K" parts are also more likely to spend more money on their other components. especially faster RAM. Fast RAM doesn't count for much, but it is more than enough to move the results a percent or two.

So in real life (on average) a machine with a K CPU is slightly faster.
BUT, you should be able to buy a non-K unit with fast RAM, a good motherboard and match the K unit's speed.

[purple]

Very helpful response. Thanks once again.