How are slower but newer CPUs better than older but faster CPUs?

There's A LOT more to it than this, but basically, speed itself is only comparable in the same generation/manufacturer of CPU's. Otherwise, it's useless to try to compare the speeds of CPU's from different generations and different manufactures. Different generations and manufacturers = Different architectures (usually), meaning totally incomparable based on speed alone.

Just know that speed isn't everything. A 3.6 Ghz P4 would get destroyed be a 3.1 Ghz first gen Core CPU, for comparison, and it gets worse the newer you go (just for comparisons sake. I know that's not what you're talking about here).

Edit: This post is heavily edited from it's original form, lol, but I feel the explanation is better than it was at first.

there are two issues you have raised.
One is the performance per Ghz, if you consider that if different architectures carry out different numbers of instructions per clock cycle then you can easily see that a lower speed but with a better number of instructions per cycle could be faster than a higher speed with a lower number of instructions per cycle.
Following the end of the P4 the Ghz race was halted by processors that were able to do a lot more per cycle, hence the first core 2 duo's were faster than the fastest P4's.
The issue about price, is partly a branding issue, also the ultrabooks have other features that are charged for, i.e. thinness and lightness, and the thinness affects what processors can be fitted based on their thermal performance.

Thank you both for answering! I understand now that with a new line of processors there is a new, different architecture and other features of the processors are added, etc. Never really thought about the number instructions per cycle though, so that is something I will look into when I'm looking at a processor. Thanks again for you help  

Don't bother looking into it as a buying decision, the benchmarks, processor to processor is all you need as cache and other elements come into play.

Also, in the case of laptops, battery life also comes into play.

Newer architectures draw less power and are more efficient. EX: Draws 1/2 power but performs 2x faster!

can we give up the capacity and somehow increase the speed....... before i go to sleep with chrome loading 7 pages or just starting up?
please?

In terms of "ghz" this metric is only a valid measurement of speed when comparing processor of the same family/architecture. This is because IPC(Instrctions per clock) vary wildly from one processor architecture to another. For instance a "Sandy Bridge"(i.e intel 2600k) completes 30%+ more work in a single clock cycle than say a processor from AMD's Phenom 2 lineup (i.e P2 965 BE). Therefor whne comparing those 2 very different processors "ghz" would not be an accurate measurement of performance.

A 4.0ghz Sandly Bridge is going to be faster than say a 4.4 Ghz Phenom 2. WHen you start throwing in technologies like Hyper threading it gets even more confusing. Just remember that GHZ can only be a true measure of performance when comparing 2 different processors of the same exact architecture, core count, extensions, features, etc...

From SB to IB and IB to Haswell, it looks more like 5-10% faster while using 10-20% less power unless you count Haswell's low-latency sleep state which Intel says will reduce the amount of power wasted going in/out of sleep by ~10X vs IB.

Assuming perfect core scaling and 100% full load, the simplified performance equation you can use to calculate the relative performance between two processors:

Theoretical Max Speed = (Instructions Per Clock * Clockspeed) * Number of Cores

So, going even from a Pentium 4 (3.4GHz) to a Core 2 Duo (1.7GHz), the Core 2 is faster, because the Instructions Per Clock of the Core 2 is SIGNIFICANTLY faster then the Pentium 4.

EG, assuming the Core 2 arch can do about 1.5x the Instructions Per Clock as the Pentium 4:

Pentium 4:
Theoretical Max Speed = (Instructions Per Clock * Clockspeed) * Number of Cores
Theoretical Max Speed = (1 * 3.4) * 1 = 3.4

Core 2:
Theoretical Max Speed = (Instructions Per Clock * Clockspeed) * Number of Cores
Theoretical Max Speed = (1.5 * 1.7) * 2 = 5.1

5.1 (Core 2) / 3.4 (Pentium 4) = The Core 2 is 1.5x the speed of the Pentium 4, using these figures. Even though you doubled the cores and halved the speed, the Core 2's superior IPC drives performance ahead of the "faster" Pentium 4.

Problem is, IPC generally isn't a flat number, and usually has to be reverse calculated from benchmarks. But you can usually "guestimate" the figure based on performance metrics.

EDIT

Interestingly, you see that the equation is associative, so reducing one component to maximize another generally doesn't yield significant performance changes. AMD's BD/PC was an example of this: IPC was reduced in order to increase the Number of Cores, which lead to typical performance about the same as their previous architecture. Its also worth noting that the more cores you have, the harder it is to attain maximum processor performance, so IPC increases will generally yield superior performance to the number of processing cores.