Curios: What makes CPUs better than others? [closed]

I have been wondering about this for a long while now and was hoping someone here could answer it pretty easily. If I was looking for the most powerful CPU what should I really be looking at? There are so many different parameters of a CPU and I am wanting to know what each thing does and what really matters. Basically this: What is the deal with cores? If I take using optimized applications out of the mix would it theoretically better to get quad core 1.0GHz CPU or a 1 core 4 GHz CPU? Also, what is the difference between maybe an Sandy Bridge CPU versus an Ivy Bridge CPU? If they both were had the same clock speed and number of cores would the Ivy Bridge perform better? Does an older Xeon with an equal clock speed and number of cores to a new i7 really perform worse/slower? Does size matter? Why would I go with a 22nm CPU over a 32nm when the size difference is so trivial? What about the cache? When does the cache come into play with performance?

Without a good understanding of underlying CPU technologies and the applications that use them, it is unfortunately non-trivial to compare two processors from different lines or determine a 'most powerful' CPU.

Why would I go with a 22nm CPU over a 32nm when the size difference is so trivial?

It is hardly trivial in percentage terms; as that is a 1/3rd reduction. The smaller size means that:

• more complex circuitry can fit on the same die size
• less heat generation, which allows usage of smaller heat sinks or more overclocking
• the 22nm one is likely newer, and has other architectural improvements integrated as well

What is the deal with cores? If I take using optimized applications out of the mix would it theoretically better to get quad core 1.0GHz CPU or a 1 core 4 GHz CPU?

Multiple cores allows the system to run more than 1 application without having to context switch, it can also help with overall responsiveness as your foreground application doesn't starve the OS of CPU resources.

Most typical computer applications don't max out each core, so more of them will improve performance. Conversely, in some cases like games which make heavy use of a single core but aren't optimized for multiple cores, the performance is better with higher clock speed per core.

If they both were had the same clock speed and number of cores [... which does better?]

As a rule of thumb, the newer one should do better as it has the latest architectural improvements (better pipeline prediction, lower branching penalty etc.) integrated.