CPU Over-clocking

The basic processing unit consumes a stream of pre-formatted binary data “words”, which tell said processor which operation is being done, where the result is to be placed, where any arguments or input data is found, along with various status flags and data.  Computers only understand – if you could actually call it that – binary.  This is much like a bank of switches.  You flick switches on and off to indicate the operation you want to accomplish.  This is the process of setting up the instruction.

At the lowest level data is loaded into a “latch” or “register” which the “instruction” can be made to operate on.  A place to store results – also a latch or register.  A Modern Central Processing Unit, known as the CPU, stores it’s instruction word in a latch or register.   When everything is setup – the clock switch is triggered to process the instruction word, and post the result.  On a calculator you might call it the enter button.  In any case something is done to progress the operation.  The system clock does this for a computer.  There is a clock signal for loading and latching the next instruction, executing said instruction, and then latching the results until they are transferred to the system buss.  More complex instructions take more cycles, but the few mentioned above are somewhat minimal.  We know how long each of these operations optimally take by testing particular CPU technologies, and so design a clock scheme to accommodate the technology so as to get acceptable results.  CPU makers typically somewhat “under clock” this process so as to get something hopefully better than optimal results.  After all, six failures per million is really good for a person, but not so great for a computing machine doing 4 billion operations per second.

The practice of pushing the CPU clock speed up toward the absolute limit of the technology is called “over clocking”.  If you are willing to live with the occasional “random dynamic timing error”, or can blame failures on cosmic rays, or you can live with a couple hundred errors per minute, or deal with a heat damaged CPU, or have an otherwise fault tolerant system you may be able to significantly increase your CPU clock rate.  If you can not rationally accept those risks, just stay with the chip manafacturers recommendation.


Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: