Different makes and models of CPU have different characteristics, but it is not easy to find the right criteria for comparing them.
Early CPUs had machine languages with very simple instructions. They might not even be able to multiply two numbers as a single instruction. Instead, many "add" instructions would have to be executed. Over time, as CPUs became more complicated, so did their machine languages. Single instructions were added to do the work of multiple instructions in previous CPUs. For example, to retrieve two numbers from memory and multiply them could be one instruction.
These CPUs with large numbers of more powerful instructions become known as CISCs, or Complex Instruction Set Computers. Having more powerful instructions would seem an advantage, and it is, but it has strong disadvantages as well. The main problem is that the logic circuits for executing an instruction must be more complicated as well, which means they take up more space on the chip. This in turn leaves less space on the chip for other things, like registers and cache, which improve the CPU's performance.
Because of these problems, CPU designers started making instructions simple again, as simple as they could make them. With the saved space, they added large register sets and caches. This meant that programs had to execute more instructions to accomplish the same tasks as CISCs, but because each instruction could execute faster, the overall performance was better. CPUs that use a small set of fast-executing instructions are called RISCs, or Reduced Instruction Set Computers.
The RISC idea can be said to have won for now, and most processors can be said to be more RISC-like than CISC-like. Those that retain CISC features do so mainly for backwards compatibility. Why is RISC faster? Because the design executes instructions faster, the only way for CISC to keep up is if the number of instructions per program can be dramatically reduced.
This isn't possible because the most common instructions used in programs are the simplest ones. The most complicated instructions on a CISC machine—the ones that do the work of many RISC instructions—are so rarely used in actual programs that they make little difference.
When a computer is advertised, a number like 2.8 GHz is often displayed prominently at the top of the feature list. This number is the clock speed, which is the number of CPU cycles per second. The Hz indicates "Hertz," which means cycles per second. The G means "Giga," which means 1 billion. Thus, a 2.8 GHz processor has an internal clock that pulses nearly three billion times a second, and the actions of the CPU are triggered by those pulses.
If a CPU executed one machine language instruction per cycle, the clock speed would directly give the number of instructions per second, but because of pipelining and other complications, these numbers are never the same. The clock speed is most useful in comparing two CPUs of the same make and model. An Intel Pentium IV executing at 2.8 GHz is a little faster than the same Pentium IV executing at 2.6 GHz.
Clock speed is not reliable, however, when comparing different model CPUs. AMD, another chip maker, has a CPU called the Athlon XP. When this chip is clocked at 2.2 GHz, it executes about the same number of instructions per second as a 2.8 GHz Pentium IV.
AMD chooses not to identify their processors by clock speed because their processors would appear to the consumer as slower than they really are. Intel is planning to follow suit, because it is offering many variants of the same chip now. Having multiple Pentiums with the same clock speed available won't help consumers pick the right one for their needs. Instead, both companies will use artificial model numbers that have nothing to do with clock speed.
Still, the question remains open: How does one compare one CPU's performance against another if clock speed isn't the answer? The best solution is benchmarking, which in computer science means running the same program on multiple computer systems and timing the results. If both computers are as identical as possible except for the different CPUs, then the difference in overall performance of the system should be attributable to the difference in processors.
Unfortunately, even benchmarking is not an absolute guide. CPUs have different strengths and weaknesses. Just because CPU A is ten percent faster than CPU B using one particular program doesn't mean it will be ten percent faster on all programs. CPU A may even be slower on other programs. Still, benchmarking is the most reliable tool for comparisons of real-world performance, especially if the benchmarks are performed using the kind of software the user is interested in running.
Our website is not responsible for the information contained by this article. Articleinput.com is a free articles resource thus practically any visitor can submit an article. However if you notice any copyrighted material, please contact us and we will remove the article(s) in discussion right away.
Note: This article was sent to us by: Ryan Welsh at 02122011
1. Interactive shape deformation
All articles are property of their respective authors. Please read our Privacy Policy!
© 2009 ArticleInput.com.
Partners: Damenmode