you also have to factor in microarch like i the i7 compared to the Phenom II it beats it in almost everything, by a very small insignifacant margin in games but a bigger margin in calculations
it doesnt equal performance directly at all, hell sometimes performance gets worse (indirect cause but still, eg 90nm vs 65nm A64's with higher L2 latency, and Northwood 130nm vs Prescott 90nm P4's etc)
what it does (usually) equal is smaller transistors = more can be packed into the same/less space = higher feature/performance/core count = lower heat/power = higher clock speeds within an equal or lesser thermal spec = INCREASED PERFORMANCE
nm can be seen as what 'chip generation' you are talking about. 90nm chips are produced using older fabs than the newest 32nm chips. Simpler chips are often made on older factories, while high-performance high-profile chips like AMD/Intel CPUs are made using the newest generation fabs.
Generally, you should avoid buying older products when newer ones are available. For example, many still bought the E6600 produced on 65nm while the E8400 on 45nm was also available, albeit a bit more expensive. In return though, you should get the benefits of more modern production technology: less heat, capability of higher clock speeds, new instruction sets, less current leakage, more overclock potential, etc.
Generally, whenever a newer - smaller - production fab is ready, there is a choice by the manufacturer to either increase performance by a more complex design or higher clockspeeds -OR- to reduce energy consumption if the other factors stay the same (chip design and clockspeeds). Often a combination of these, and chips produced with newer production methods may also significantly reduce idle power drain causing massive energy savings - even if their thermal envelope says 125W TDP.
Several people here have come up with answers to this question. The only answer that seems similar to any other answer is that is is the size of the transistor. I'm going to look around some more. Thanks for your help.
Nanometer in CPUs specify the distance between two transistors connected in sequence. To improve the performance of a processor data from one transistor to next should be available as soon as possible which results in higher processing speed due to less time delay. The logic here is that the speed of light in vacuum is 3x10^8m/s, we cannot exceed this limit. So, if the signal is traveling even at this speed it has to cover some distance, lesser the distance less will be the time delay. So, if we reduce the distance between two consecutive transistors we can get higher processing speed due to reduce time delays.