CPU multiplier
From Wikipedia, the free encyclopedia
The clock multiplier (or CPU multiplier or bus/core ratio) is the ratio of the internal CPU clock rate to the frequency of its external address/data bus, today often termed front side bus (FSB). A system with a CPU multiplier of 10x will have its CPU execute 10 complete cycles for every cycle of its FSB. For example, a system with an FSB running at 133 MHz and a clock multiplier of 10x, the CPU will run at 1.33 GHz.
Modern computers has several interconnected devices (CPU, RAM, peripherals, etc - see diagram) that typically run at different speeds. Therefore internal buffers (or caches) are used when communicating with each other via the shared buses in the system. The CPU's external address and data buses connects the CPU to the rest of the system via the "northbridge". Nearly all desktop CPUs produced since the 486DX2 has employed a clock multiplier to run its internal logic at a higher frequency than its external bus, but still syncronous with it. This improves the CPU performance by relying on internal cache memories and/or wide buses, often also capable of more than one transfer per clock cycle, to make up for the frequency difference.
Some CPUs, such as Athlon 64 and Opteron processors, handle main memory using a separate and dedicated low-level memory bus. These processors communicate with other devices in the system (including other CPUs) using a somewhat higher level HyperTransport link (or several links). On these systems, the clock multiplier refers to the ratio of the CPU clock speed to the HyperTransport clock speed (typically 800 MHz or 1 GHz, as of 2007)
Most systems allow the clock multiplier to be changed in the BIOS menu. Increasing the clock multiplier will increase the CPU clock speed without affecting the clock speed of other components. Increasing the FSB speed will affect the CPU as well as RAM and other components. These are the two main methods of overclocking and underclocking a system; careless overclocking can cause damage to a CPU or other component due to overheating.
[edit] Clock doubling
Clock doubling refers to a clock multiplier of 2.
The most famous example of a clock-doubled CPU is the Intel 80486DX2, which ran at 50 or 66 MHz on a 25 or 33 MHz bus. Another example was the Weitek SPARC POWER µP, a clock-doubled 40 MHz version of the SPARC processor that could be dropped into the otherwise 20 MHz SPARCStation 2. In both cases the overall speed of the systems increased by about 75%.[citation needed]
Today almost all processors run at a different speed than the underlying bus, and the term clock doubling is no longer needed.
For CPU-bound applications, clock doubling will theoretically improve the overall performance of the machine substantially, if fetching data from memory is not the limiting factor. In more modern processors where the multiplier is not 2, but much higher, the fetching data indeed typically becomes the limiting factor, and the processor's performance is limited by both the latency and bandwidth of the memory bus.
