WHY ARE COMPUTERS NOT SPEEDING UP?
Moore’s Law was coined in 1965, which said that the numbers of transistors on computer chips will double every year. So far, chip development has managed to stay on course, but recently, the relentless march has been stalled by a few walls. Computers are not getting any faster.
There was a time over the last ten years when certain friends of mine used to buy a new computer every two years. Each time they made a purchase they could bank on getting a new toy that is close to eight times faster than the one they already had. For instance, in the late nineties, desktop computers with 2 GB of hard disk storage space raised eyebrows and elicited whistles of envy. Back then, 32 MB of RAM was the ‘in thing’.
But recently, especially in the last two years, the phenomenon has changed a little. People are finding that they computer they have owned for two or three years is just as good (almost) as the high-end models that are available on the market today. Consequently, all my friends who had been replacing their computers every two years for the last decade decided to skip it this time. They like their old toys better.
So the speed at which computers are getting faster has been going down. Why?
There are three ‘walls’ that are stopping the onward march of CPUs: a memory bottleneck (the bandwidth of the channel between the CPU and a computer’s memory); the instruction level parallelism (ILP) wall (the availability of enough discrete parallel instructions for a multi-core chip) and the power wall (the chip’s overall temperature and power consumption).
Of the three, the last wall – the power wall – is proving to be the most difficult. As transistors become smaller and smaller, the leakage of current and building up of heat are becoming larger and larger, because the Silicon Dioxide that has been traditionally been used to insulate one transistor from another has been shaven down to such small thicknesses that more and more current is seeping through, generating unnecessary electricity and heat.
New materials are being researched, though. There are carbon-based nanotubes in the horizon which don’t have any heat loss problems, so they may replace the current materials and cause another computer boom. But until then, companies are betting on the much cheaper processes of parallelism by mounting multiple cores rather than increasing transistor density.
We may have to wait a little bit longer yet, therefore, to have robots walk in our midst.