Is 5Ghz the limit ??? they seem just to be adding more cores

"5Ghz" means little.

Which is faster/more powerful - a P4 3.4Ghz (2003) or a i5 3.4Ghz (2014)?

The 5960 is 8 cores /16 threads. It doesn't need a lot of speed. It'll run to about 4.4GHz while keeping below 75C - http://www.anandtech.com/show/8426/the-intel-haswell-e-cpu-review-core-i7-5960x-i7-5930k-i7-5820k-tested/3

Easily? No - they don't easily run at those speeds. See the temps and power consumption go up in that article.

They release chips at an advertised clock, because that is what they can guarantee. Some individual chips may be able to be clocked higher, but that is all on you. Go to the store, and buy 10 identical CPUs. In the same case, with the same cooling, with the same motherboard....all will achieve a different absolute maximum.
Now fold in 15 different motherboards, and JoeRandom cable management/cooling situation.....you see why they advertise a specific Ghz.

Oh, and physics!

Multi tasking is something a person describes as doing. A person can single task and multi task. But a computer always multi tasks. It always runs hundreds of processes and functions. Even if you think you are running a single game, that involves many, many processes executed simultaneously. That requires a lot of power.

Intel had an idea with the Netburst architecture back in the early 2000 that they will soon hit 10 GHz on a single core. This proved to be rather impossible, since the electrical consumption in CPUs grows exponentially with GHz. So 2 cores at 2 GHz consume less power than 1 Core at 4 GHz. And 2 cores at 2 GHz consume half the power that one core at 4 GHz consumes (rough estimate by myself, don't ask me to back it up).

So in the end, it proved more efficient to parallelize processes and run them on a midly GHz-ed chips with larger core count.

But still - not all CPUs are created equal. Not all architectures are equal. AMD and Intel cores are not equal. AMD and Intel MHz are not equal. Not all software is created equal. Some is parallelized - some is not.

In an ideal world:

You have an architecture. You have GHz, You have number of cores.

So it makes more sense to make more cores than to go for GHz.

So lets make up an imaginary number:

CPU performance = [(Architecture Productivity)*Ghz]*Core Count+1/3 per core if Hyper threading is present (because in my own experience Hypet hreaded thread has 1/3 productivity of a real core)

Core2Quad Q6800 (2007) = [(65)*2.93]*(4)
I7 2700K Sandy Bridge (2011) = [(100)*3.5]*(4+4/3)
I7 3770K Ivy Bridge (2012) = [(105)]*3.5]*(4+4/3)
I7 4790K Haswell (2014) = [(110)*4.0]*(4+4/3)

This would lead you believe that a Core2Quad running at 2.93 GHz is the same as an Sandy Bridge I3 at 3.0 GHz

Because [(65)*2.93]*(4) = [(100)*3.0*]*(2*2/3) or almost equal. But then you forget that the I3 has 2 real cores and consumes half the power that the Q6800 takes. That is because the architecture is much more advanced. There is much more behind the hood of a CPU than number of cores and GHz.

---------------

TL;DR version:

For practical reasons, the top overclock maximum for daily usage caps out at around 4.5/5.0 GHz.

They make changes in architecture where each clock gets more work done. Then the processing is quicker with lower clocks, power consumption and temps.

every new generation has a new architecture which adds and increases Single Core performance, a Pentium 4 is about 4x slower than the current hashwell CPu in single threaded application so if you were to add the extra 3 cores a pentium 4 is about 15x slower than an i5, you cant compare Ghz vs Ghz unless they are using the same architecture.

for example a i3 4130 at 3ghz is still clearly much faster than a i3 2100 at 3ghz, frequency doesn't mean much.

Moore's law simply states:

"The number of transistors in a chip doubles every 3 years"

Moore's law does not say anything about performance, core count or GHz. Nothing at all. A core can be 10 million transistors, a core can be 100 million transistors. Nothing to do with Moore's law at all.

Moore's law is not leveling off yet, just because clock speeds aren't increasing exponentially doesn't mean the amounts of transistors aren't still increasing!

No. The package size (typical CPU) has not really changed. Performance has changed hugely.

Cores that are not used are switched off. It is also known as "parking" cores. This means the core is turned off and does not drain electricity.