faugusztin wrote:You base your asumption on false statement. You see a 3.3GHz and asume that i5 2500 has standard x33 multiplier. While it is theoretically true, technically it is not. i5 2500 has minimum multipler of 16, maximum of 37 and maximum unlocked of 41 that is it (for 2500K, it is maximum unlocked of 57). In case of H67, your multipliers are managed automatically between 16 and 37 (or 38 in case of 2600K), and if you have selectable multiplier in BIOS, then you can fix it at that multiplier i guess.
First of all, I guess we're talking about different things.
As far as I can understand, the above depicted scenario is somehow what really happens with SB.
But our talking started from this state of yours: "If it can fix the 3.7/3.8GHz for all 4 cores, it gives you a slight overclock, but nothing extraordinary and you don't need a K-type CPU for that, as you will get same "overclocking posibility" with 2500 and 2600, without K.".
I read this phrase as: "If it had been a slight but true overclock, you wouldn't have need a K-type CPU for get it, because you would have get the same "overclocking possibility" with 2500 and 2600, without K.".
With reference to that, now we know the condition "if it can fix" isn't true, as it shouldn't fix anything for any core, because - according to NTNgod - even fixing it, that multiplier goes down when the load goes up (as in Turbo mode schemes): so definitely it isn't an overclock, but for our purposes it doesn't matter; what really matters on the contrary is that if it had actually set the multiplier to a fixed (say) 37x, only -K would have to agree to work continuously at that frequency. I'm saying so as at anyway there should be a "stock", or "standard", or "steady state" multiplier, sort of a flag set somewhere in the CPU microcode which actually differentiate -K SKUs from the others one, providing that in this latters (talking about a 2500, for example) any multiplier higher than 33 (so up to 37/41) may be reached just dynamically, ONLY in Turbo modes.
Eventually we could marvel that this limit coincides with that of Turbo, but I guess it's because the limit of the running fixed/static multiplier is somehow actually twofold: there is a limit that the chipset may set (to say, 37x for the H67), and the limit that the CPU can accept (33x or 57x respectively for not-K or -K CPUs).