Ever wonder what makes your computer work? What controls all the information? Without going into huge details about CPUs and the physics behind their workings, everything generally revolves around the Central Processing Unit (CPU). This month's column will be about CPUs and the various jargon associated with them. Heck, maybe I'll even take a jab or two at Intel's Pentium‚Ñ¢ series microprocessor. ;-)
The CPU (also referred to as a microprocessor, microchip, or processor) is the part of the computer that makes everything happen. In order to understand how the CPU serves all the purposes intended for it, it is helpful to know a bit about how programming languages are understood by the computer.
Programming-languages such as C++, Java, PASCAL, and so on, are known as high-level languages because they perform very complex and abstract processes on very complex and abstract data structures. However. the machine and all its components (including the CPU) are only capable of performing simple logical operations (such as AND and OR) and mathematical operations (such as addition and multiplication), and are only capable of performing these on very simple pieces of data represented in binary (strings of 0's and 1's). Thus, in order for high-level programming languages to be used, all of the instructions and data structures they represent must be broken down into these simple terms. To achieve this, the abstract language a programmer works with is compiled (converting all of its instructions into simple arithmetic and memory operations, and turning all its data into references to simple values) and then assembled (converting these simple instructions and data representations into sequences of binary code which the machine can work with directly).
 
RISC or CISC?
There are two main types of CPUs, Complex Instruction Set Computer (CISC) and Reduced Instruction Set Computer (RISC). A CISC CPU is exactly what its name describes — it has many specific instructions hard-coded into the chip so as to provide instructions for all operations that might be needed. CISC CPUs tend to be very complicated, power-hungry, and slow. A common example of a CISC CPU is the Pentium series designed by Intel.
RISC CPUs perform the same functions as CISC chips, but instead break all of the complex instructions down into sets of simpler ones. When John Cocke first thought up the concept of RISC, he proved that 20% of instructions in a CISC CPU did about 80% of the work. David Peterson, a professor teaching at the University of California at Berkeley, first coined the term RISC, and the basic idea behind it is that, by reducing the basic instructions used by the CPU into much smaller and simpler ones, programs can use just the simplest instructions they need; more complex instructions can be custom-made out of the parts provided by the simple instructions, eliminating the need to call upon very large, complex instructions which may often carry a lot of unneeded extra baggage. Thus efficiency is increased by making it easier for a program to ask the CPU to do exactly what it needs, and nothing more.
A RISC CPU is fairly uncomplicated, and therefore the space on the CPU can be used for other important items, such as an on-chip cache (see L1 cache below). (CISC chips like the Pentium II can (and some do) have L1 caches, but we all know how large those processors can be.) High-level programming languages can produce more efficient code due to the fact that they generally use the instructions already found on the RISC CPU. It is easier for programmers to program low level code on a reduced instruction set.
 
RISC systems are more popular in the server and high-end
workstation environment. IBM's PC/XT, released in the early 80s,
was the first computer to use a microprocessor designed using the RISC concept. IBM then later developed the RS/6000 (RISC System 6000). DEC's Alpha microprocessor is designed of the RISC concept. Sun Microsystems's SPARC stations also use RISC. The PowerPC microprocessor series, designed jointly by IBM, Motorola, and Apple, is RISC. It's found everywhere, from Apple's computers to IBM's Big Blue (which uses 5,856 PowerPC 604 microprocessors) and even the family car.
 
Cache is Good 
Cache is temporary memory that stores frequently and recently used
data and instructions. There are several different types of caches (Web browser cache, disk cache, etc) but the types I will be describing are the ones associated with the CPU.
Level 1 cache memory (L1 cache) is a very small amount of memory that is located on the actual CPU. The PowerPC 750 CPUs (more commonly referred to as the G3 series) have two separate caches: a 32 KB instruction cache and a 32 KB data cache.
Level 2 cache memory (L2 cache) is a larger amount of memory (relative to the L1 cache) that is generally located right next to the CPU (rather than actually on the CPU). L2 cache is slower than L1 cache, but it is still accessed much faster than regular RAM. L2 caches are generally upgradeable and can come in several different sizes. Generally, L2 caches can range widely from 256 KB to 2 MB.
Level 3 cache memory (L3 cache) is yet another, larger amount of memory. L3 cache isn't very common and is generally found on only really new systems and high-end workstations. L3 caches range from 2 MB all the way up to a huge 8 MB on some server machines.
 
How Fast is it Really?
A CPU's speed is labeled by Megahertz (MHz). A hertz (named after the German physicist, Heinrich Hertz) is a single "cycle" or oscillation of an electromagnetic signal per second. MHz stands for one million hertz, or one million cycles per second. All the old-school electrical and computer engineers will be more familiar with the term cycle per second (CPS), which predated MHz.
MHz isn't a terribly good speed indicator, though. To really measure how good a processor is you have to measure how many million instructions per second (MIPS) are done, as well as how much is accomplished by each instruction (this differs among various machines, and especially between RISC and CISC processors). Another excellent measure of how good a CPU is by measuring how many floating-point operations per second (FLOPS) it can do. IBM describes floating-point as "a method of encoding real numbers within the limits of finite precision available on computers." That is to say, floating-point is a method of taking an analog number and giving it a digital representation that is as accurate as possible.
 
Word to the CPU
You have probably heard the term "16-bit" or "32-bit." This is in reference to the most basic packet of information that can be manipulated by the computer — it is called the word size. The larger the word size, the more information the CPU can work with at once. The bit size is always a multiple of 8 and the byte size is always a perfect square (i.e.: a CPU is a 32 bits, and there are 8 bits in a byte, so the CPU is 4 bytes.)
 
 
overclock — If you have a craving for speed and today's fastest off-the-shelf processors don't satisfy your desires, then maybe it's time to turn on the nitro (figuratively speak, of course). There is a special clock on the board where the processor resides which determines how fast the processor cycles (a processor running at 100 MHz cycles 100 million times per second). CPUs are engineered with a safety margin between the maximum possible MHz and the speed at which it is shipped.
You can reset this clock to tell the processor to cycle faster. There is a catch though — as the processor's speed increases past its rated speed, the more likely errors are to happen, and the more wear and tear the CPU endures. Overclocking causes a processor to draw more power, which subsequently causes the processor to heat up and become stressed.
Depending on the make of your processor and the circuit board it is mounted to, there are different ways to overclock your system. Some boards have special jumpers (sets of small pins whose configuration in a series tells the motherboard how fast to run the CPU), and and other boards might have tiny little rheostats (think of an old-fashioned wheel thermostat) for you to turn.
Beware though! Overclocking voids most manufacture's warranties!
 
 
Q. I would like to know what "octet stream" means. Please help!
— Silver
A. Octet is derived from the Latin word "octo" which translates to "eight." An octet is 8 bits. "Octet stream" refers to a sequence of 8-bit packets that are being transferred between a server and a client.
"But why call it an octet and not a byte?!" you ask. The term was orginally created because on some computer systems and packet-switched networks, a byte isn't 8 bits.
