Modern CPU's are what are called 'integrated chips'. The idea behind an integrated chip is that several types of components are integrated into a single piece of silicon (a single CPU), such as one or more execution core, arithmetic unit (ALU), logic unit, registers, instruction memory, cache memory and the input/output controller (bus controller).
Each transistor is a receives a set of inputs and produces output. When one or more of the inputs receive electricity, the combined charge changes the state of the transistor internally and you get a result out the other side. This simple effect of the transistor is what makes it possible for the computer to count and perform logical operations, all of which we call processing.
A modern computer's CPU usually contains an execution core with two or more instruction pipelines, a data and address bus, a dedicated arithmetic logic unit (ALU, also called the math co-processor), and in some cases special high-speed memory for caching program instructions from RAM.
The CPU's in most PC's and servers are general purpose integrated chips composed of several smaller dedicated-purpose components which together create the processing capabilities of the modern computer.
For example, Intel makes a Pentium, while AMD makes the Athlon, and Duron (no memory cache).
Generations
CPU manufacturers engineer new ways to do processing that requires some significant re-engineering of the current chip design. When they create this new design that changes the number of bits the chip can handle, or some other major way in which the chip performs its job, they are creating a new generation of processors. As of the time this tutorial was last updated (2008), there were seven generations of chips, with an eighth on the drawing board.
CPU Components
A lot of components go into building a modern computer processor and just what goes in changes with every generation as engineers and scientists find new, more efficient ways to do old tasks.
- Execution Core(s)
- Data Bus
- Address Bus
- Math Co-processor
- Instruction sets / Microcode
- Multimedia extensions
- Registers
- Flags
- Pipelining
- Memory Controller
- Cache Memory (L1, L2 and L3)
Measuring Speed: Bits, Cycles and Execution Cores
Bit Width
The first way of describing a processor is to say how many bits it processes in a single instruction or transports across the processor's internal bus in a single cycle (not exactly correct, but close enough). The number of bits used in the CPU's instructions and registers and how many bits the buses can transfer simultaneously is usually expressed in multiples of 8 bits. It is possible for the registers and the bus to have different sizes. Current chip designs are 64 bit chips (as of 2008).
More bits usually means more processing capability and more speed.
Clock Cycles
The second way of describing a processor is to say how many cycles per second the chip operates at. This is how many times per second a charge of electricity passes through the chip. The more cycles, the faster the processor. Currently, chips operate in the billions of cycles per second range. When you're talking about billions of anything in computer terms, you're talking about 'giga' something. When you're talking about how many cycles per second, your talking about 'hertz'. Putting the two together, you get gigahertz.
More clock cycles usually means more processing capability and more speed.
Execution Cores
The third way of describing a processor is to say how many execution cores are in the chip. The most advanced chips today have eight execution cores. More execution cores means you can get more work done at the same time, but it doesn't necessarily mean a single program will run faster. To put it another way, a processor with one execution core might be able to run your MP3 music, your web browser, a graphics program and that's about where it starts to slow down enough, it's not worth it running more programs. A system with a processor with 8 cores could run all that plus ten more applications without even seeming to slow down (of course, this assumes you have enough RAM to load all of this software at the same time).
More execution cores means more processing capability, but not necessarily more speed.
As of 2008, the most advanced processors available are 64-bit processors with 8 cores, running as fast as 3-4 gigahertz. Intel has released quad-core 64-bit chips as has AMD.
Multi-Processor Computers
And if you're still needing more processing power, some computers are designed to run more than one processor chip at the same time. Many companies that manufacture servers make models that accept two, four, eight, sixteen even thirty two processors in a single chassis. The biggest supercomputers are running hundreds of thousands of quad-core processors in parallel to do major calculations for such applications as thermonuclear weapons simulations, radioactive decay simulations, weather simulations, high energy physics calculations and more.
CPU Speed Measurements
The main measurement quoted by manufacturers as a supposed indication of processing speed, is the clock speed of the chip measured in hertz. The the theory goes that the higher the number of mega or gigahertz, the faster the processor.
However comparing raw speeds is not always a good comparison between chips. Counting how many instructions are processed per second (MIPS, BIPS, TIPS for millions, billions and trillions of instructions per second) is a better measurement. Still others use the number of mathematical calculations per second to rate the speed of a processor.
Of course, what measurement is most important and most helpful to you depends on what you use a computer for. If you primarily do intensive math calculations, measuring the number of calculations per second is most important. If you are measuring how fast the computer runs an application, then instructions per second are most important.
Processor Manufacturers
- American Micro Devices (AMD)
- Intel
- IBM
- Motorola
- Cyrix
- Texas Instruments
AMD and Intel have pretty much dominated the market. AMD and Intel are for IBM compatible machines. Motorola chips are made for MacIntoshes. Cyrix (another IBM compatible chip maker) runs a distant fourth place in terms of number of chips sold.
Today all chip manufacturers produce chips whose input and output are identical, though the internal architecture may be different. This means that though they may not be built the same way, they DO all run the same software.
The CPU is built using logic gates, and contains a small number of programs called 'microcode' built into the chip to perform certain basic processes (like reading data from the bus and writing to a device). Current chips use a 'reduced instruction set' or RISC architectures. Chips can also be measured in terms of instructions processed per second (MIPS).
