The basics of RAM:
RAM is temporary storage. There are many many RAM cells, but they all work exactly the same. A charge is applied to either wipe the cell, or write a value of one. The cell does not retain that value infinitely (it slowly leaks out), so a refresh of the value is provided occasionally. This refresh cycle introduces an unusable cycle, but allows the very volatile RAM to accurately store values.
Each RAM cell has an address, which the operating system assigns it. The pool of addresses is determined by two things, the operating system and the amount of bits the system can process at once (8, 16, 32, and 64 bit OSs are currently commercially available).
Looking at a potential example, we can have three different computers, with exactly the same physical setup. The first has a 32 bit high end OS, the second a 64 bit low end OS, and the final a high end 64 bit OS. Assume each has the same processor, RAM quantity (32 GB), and other peripherals. You would see the following amount of RAM in each system:
1) 4 GB
2) 16 GB
3) 32 GB
Note that the first system can only address 4 GB. This is due to a physical limitation on addresses, because of the 32 bit OS.
Now, the second and third OSs should read the full 32 GB, right? Well, no. If you're using a Microsoft OS then there are artificial limitations on the addressable RAM. A practical example is that the Professional windows 7 x64 variant can address 192 GB of RAM, while the Home Premium can only address 16 GB of RAM.
Moving on to timings and frequency:
Modern RAM has three significant values, which distinguish similar RAM modules from one another. Voltage, operational frequency, and timings determine what RAM is expensive and what is significantly cheaper to buy.
Voltages are determined by the manufacturing process. Each RAM module is designed to operate at a specific voltage, determined by what materials they are constructed with. This value is generally given as a single voltage, but functionally is a range. Manufacturers generally specify DDR3 as 1.5 or 1.65 volts, depending upon many different factors.
Operational frequency is a huge variable as far as system performance. The frequency determines how fast a RAM module can read or write values. Higher frequencies mean there are more read and write cycles, but memory cells can only read and write so fast. Higher voltages and looser timing can allow you to run at greater frequencies. In short, higher frequencies are better.
Timing, latencies, or CAS values are all measures of how often a memory cell can be written. The value are often called "tight" or "loose" based upon values. Smaller values are better.
Finally, modern systems are more interesting. With the inclusion of the North Bridge into the CPU there is no discrete memory control module. As such, systems generally have a memory frequency that they operate at best, with diminishing returns as frequency increases. The current Intel systems need 1600 MHz, the APUs need 1866 MHz, and BD is an interesting beast that I've never been able to get my mind around. The lowest timings that meet these minimum frequencies are generally all most users need.
Assuming that you don't need special RAM (some high performance computing systems and programs might show preference for either tight timings or high frequencies), just shoot for what pricing you can afford. Forget all the crazy RAM overclocking, because most common users won't see a difference between similarly timed 1600 MHz and 1866 MHz.
