CompuServe Messages

> I was under the impression that clock speed, in it's basic form, > doesn't necessarily make a machine fast. Now, let me qualify > that with "100mhz DOES make a fast machine", but more via brute > force than actually architectural integrity… but just because > a system has a higher clock rate than another system, doesn't > always make it faster. High clocks also imply that the CPU is "simple", this can mean it's cheap to make. In the end all that matters is raw speed. There are two basic ways to achieve this. The "Super Pipelining" camp says make it super simple and push the clock very high. This is what MIPS does. The other camp says go "Super Scalar", add in execution units to let more things happen in parallel. This makes it slightly harder to push the clock high. PowerPC is in this camp. The CISC chips are generally limited to "Super Scalar", but get fewer benifits because the overall complexity don't allow the execution units to do as much as they should. In the end, it's easier to throw execution units at something than to raise the clock. 200Mhz clocks are expensive to control and distribute. Also, a clever "Super Scalar" design will also allow high clocks. For whatever reason, Moto seems to have had trouble getting the clocks up on the 68K family. The '060 will barely reach 66Mhz by the end of the year. By any definition it's a "medium" speed chip. It will help people run exisiting binaries faster, or make a simple "plug-in", but it won't compete in raw speed with any other modern CPU. —— Andrew