The Road to More Processing Power

For years, we mistakenly judged the performance of a processor by one thing: the clock speed. As Intel and AMD leapfrogged their way to 1 GHz chips and beyond, this clock speed was the benchmark that gave us bragging rights on the playground, or the office watercooler. As AMD went for overall performance over pure clock speed, they struggled for equivalency in the minds of consumers, and used their artificial PR system to simplify it for the masses.

Along the way, the 2 and 3 GHz barriers got smashed, and 4 GHz was directly in the crosshairs. However, with so many transistors smooshed onto one little piece of silicon, stability in the face of heat became too much of a problem, and speeds topped out officially at 3.8 GHz (of course unofficially overclockers have hit way beyond 4 gigahertz).

Why no one really cares today that there was never an official 4 GHz part is that the market has shifted to dual cores. In the realm of two processor chips, the amount of on chip memory and its flexibility in being used by the busier processor, as well as the efficient communication between the cores, play as much (or more) of a role in overall performance than clock speed does. What's the proof of this? Look at the clock speed of the lowest Core 2 Duo chip and we see that its clock speed is under 2 GHz, but it still will outperform any single core chip, hands down.

Dual core chips are clearly here to stay, and their single core predecessors will be relegated to budget duty only. Meanwhile, the fun part is to look at where the market goes from here. In all likelihood, the next stop will be four processor chips.

AMD is working on their "4 x 4" project which is their answer to the Core 2 Duo. The plan is to create a motherboard that can fit two dual core processors. Aside from the added cost of buying 2 processors (then again, gamers happily buy 2 graphics cards) this will be less than ideal. While I'm sure there will be a performance boost, and it's anyone's guess if it will be faster than the current Core 2 Duo, the problem will be communicating between the cores, and sharing the workload. If a piece of data has to be sent through the motherboard to get to the other processor, it's going to be faster for it to stay on the original processor. In the original Intel dual cores chips, the cores didn't connect directly, which is why their performance was less than the AMD X2 chips with their bridge between the cores.

While AMD's "4 x 4" project makes a good news item, for true quad core performance we'll need the four cores all together, in direct communication with each other, with a large cache that can be dynamically adjusted to each core's requirements. If this sounds too ambitious, or the stuff of science fiction, rest assured that even with delays, we should have this type of processing technology available in 2007. My partly tongue-in-cheek guesses for their names of the new chips would be the Athlon X4 (or the Athlon 64 4 x 4), and the Core 4 Duo, (with alternates of the Core 2 Duo Duo, or Core 4 Quado).

While two and four cores have gotten the attention thus far, there is no reason that it has to be an even number of cores. Perhaps we'll see a "Celeron Core Three," or a five core "Pentium Core Pentium" down the road. The possibilites are quite endless here.

Multiple cores represent a great leap forward in processing technology. With the current power of the latest dual core chips, and the promise of quad cores, I can get over the fact that there won't be a 4 GHz processor.

--Jonas

Digg This
Del.icio.us