We will compare the i5 2500(K) and i7 2600(K) on one side, and the i5 760 and i7 930 on the other.
Note that the only differences between the i5 and i7 Sandy Bridge (2500 k and 2600 k) are:
A frequency 100 MHz higher for the i7
An additional 2 MB L3 cache for the i7
Hyper-Threading technology of the i7
The differences between the i5 and i7 Nehalem are minimal, since the frequencies are identical, and the Turbo Boost of the i7 is negligible. The differences between the i7 and the i5 are depicted below:
The i7 has the Hyper Threading
The i7 has Triple Channel has a mode (only on the i7 9xx)
The "Uncore" part of the i7 is somehow better than the i5 (2400 MHz against 2133 MHz)
The small improvements of the Uncore and memory controller (Nehalem), cache and frequency (Sandy Bridge) do not provide significant gains (in practice), we will *primarily focus on the effect of the Hyper Threading on games for these two architectures.
Effect of Hyper-Threading in games
By comparing the gaming performances ( Crysis, Far Cry 2 and Dirt) on a i7 980X, the average gain is less than 1% when HT is enabled.
The consequences are less appreciated on other games (Oblivion, World in Conflict, FEAR and ETQW), since global losses are within a range of 0.2 to 4% when HT is enabled.
Some games are indifferent to the activation of the Hyper-Threading, others react negatively when this option is activated (ArmA II, where the drop in performance is greater than 10%!).
Concerning Metro 2033 the minimum FPS is much lower on the i7: the lags will be more marked on the i7.
In conclusion, we observe that on the contrary to professional applications where the HT technology is very useful, the games really are not taking advantage of Hyper-Threading (except for Battlefield Bad Company 2)