Think about the way you select bytes from a collection. If you have a piece of memory that holds 16 bytes, then to select one of the bytes you need to feed it an address. This address will be a number from 0 to 15, or from 0 base 2 to 1111 base 2. So the address space is 4 bits. If the piece of memory held 10 bytes, you would want to pass a number from 0 to 9, or 0 base 2 to 1010 base 2. So not only do you waste available address space, but also the piece of memory has to consider that what it gets might not be a valid address.
Now if you put two of these memory modules together and you wanted to address them, how would you do that? For two 16-byte modules all you need to do is to add another bit to the address which tells you in which module you should look. You have natural addressing with numbers 0-31, and in hardware you would have to add a handful of logic gates to handle this. Now if you combine two 10-byte modules you start getting problems. You still have to add another bit to the address. But you either use it as a way to tell form which module you want to select the bytes, Which means that only numbers 0-9 and 16-25 are valid addresses, which is weird, or you have to tack on some machinery that takes numbers 0-19 and selects the right byte from memory.
So you can see that there are at least two clear reasons:
- Addresses are numbers in a computer. The most natural way to use them is to have the memory have as many addresses as you can have numbers. A 16-bit integer will address every byte within a 64KB block of memory. If you want to double the memory, just add another bit.
- Memory is made by the way of doubling subassemblies. If you look at a ram chip you will see a bunch of smaller 512 or 256 meg chips. You did it yourself if you bought two memory sticks.
Log in to comment