Loyd's reply:

Keeping up with all the jargon surrounding PC specifications seems to get harder and harder every day. Just as you start to figure out the old specs, manufacturers throw new ones at you! But I digress. You're asking a bunch of questions. Let me try to tackle them for you.

First of all, memory speeds are sort of redundant. In short, there is no PC2700 memory that doesn't run at 333 MHz. That's because the "2700" in PC2700 is the rated maximum bandwidth that 333-MHz memory can deliver. And as you point out, multiply the memory speed by 8 and you'll get its bandwidth. That's a maximum rating, however; the front-side bus (FSB) in your CPU dictates the real speed of your RAM.

For a long time, the FSB speed equaled the RAM speed. Recent innovations let the FSB pass memory data at a much faster rate, however. Memory designers were able to squeeze more than one data item into a clock cycle. So DDR (double-data-rate) memory can move two data items per clock cycle. This is made even more confusing by the fact that the memory clock and the I/O clock in memory are different. For example, DDR2-800 actually runs at 200 MHz, with an I/O clock of 400 MHz. Each I/O clock cycle can carry two data items, which is where we get DDR2-800 from. You may see DDR2-800 also referred to as PC6400--which is the maximum throughput, in megabytes per second.

You have to separate this idea a bit from the memory controller. In the case of Intel CPUs, the memory controller is part of the motherboard chipset. The CPU communicates with the memory controller via the front-side bus (FSB), and it can run at very high data rates. Current Intel FSBs can move four data items per clock cycle. So when you see an Intel CPU rated as a 1,333-MHz FSB, the actual clock rate is 333 MHz. Sometimes the throughput number is called the "effective clock speed."

AMD CPUs, starting with the Athlon 64 and moving forward through the quad-core Phenom CPUs, are different, since the memory controller is built into the CPU die itself--there is no front-side bus. The memory controller may be running at the same speed as the CPU, but they often run more slowly. There's also a bus clock, which is how the CPU communicates with the outside world. Most current AMD processors communicate via a bus clock at either 200 or 266 MHz.

Now on to your main question. In an ideal world, you want to buy or build a balanced system. In other words, you'd like to have a PC whose CPU isn't always waiting for memory, and whose graphics card isn't always waiting for the CPU to finish some task. That's a tall order. Perhaps the easiest way to accomplish it is to have the memory clock match or be an even multiple of the FSB clock; for example, if you have a system that uses DDR3 memory and a CPU with a 1,333-MHz FSB, match it up with 1,333-MHz DDR3 memory. In some cases, a balanced system will actually outperform a system with faster memory.