Chapter 4 Processor: Faster and Faster Most of the computers, no matter how it looks, can be cut into five parts: Input/Output brings things in and, once done, sends out the result; a memory remembers things such as what I am doing, what I have got, etc.; it is the processor that gets things done; and finally, we need a controller to let the other parts work together. This is what we called von Nuemann architecture in the ENIAC module, as shown in the following figure: 1
Have you heard of CPU? We typically combine a processor and the controller into one piece, called CPU, or Central Processing Unit, which, simply put, is the brain of a computer. CPU used to be pretty bucky. The following picture shows the CPU, some memory and the wires that hook them up, of an aged machine, PDP-8i. 2
It has come a long way With the development of personal computers, starting with 1980 s, the form, design and implementation of CPUs have changed dramatically, while their fundamental operation remains essentially the same. As a result, CPU, based on the integrated circuit technology, gets much smaller but faster. This combination of miniaturization and standardization of CPU have increased the presence of these digital devices in modern life, far beyond the original application in dedicated computing machines. 3
CPU is now used everywhere, in every cell phone, GPS, Nintendo Wii, children s toys,... Some of them have also been put into tires to make sure that they are filled with a proper level of air. The following figure shows a rather new central processing unit, the Intel 90486D2, with its size being 12 * 6.75 mm, about half by quarter inch. The above image is thus four times as big as its actual size. 4
What does a processor do? The job of a CPU is to execute programs. As we mentioned in the ENIAC chapter, in a computer built in the von Neumann architecture, at the very bottom, a computer program is kept in the computer memory together with the data, and it consists of a sequence of steps, or instructions. When a program is executed, CPU, starting from the very first instruction, executes one instruction after another, until the last one is done. For each and every instruction, CPU will always go through the following four steps: fetch, decode, execute, and store. 5
What are they? 1. In the fetch phase, the control unit, part of the CPU, will get the instruction from the memory into CPU. 2. In the decode phase, the control unit will decide the meaning of this instruction, addition, subtraction, comparison, or anything else, and fetch some additional data from the memory. 3. In the execution phase, CPU will execute the corresponding arithmetic, e.g., what is 3+4, 3-4, 3*4, and 3/4; or logical operation, e.g., between 3 and 4, which is bigger? 4. Finally, the result will be stored back in the memory. 6
Why is it called digital? We call a digital wrist watch digital since it can t tell nothing between 1:25 p.m. and 1:26 p.m., while an analog clock can, which is continuous, like that flowing stream. In other words, in a digital device such as a digital clock, it can only represent a finite number of states, while an analog device such as a wall clock can represent infinite number of states. As another example, a dripping faucet shows us a discrete sequence of water droplets; while a creek shows a continuous flow. 7
The fundamental weakness Computer is also a digital device. As we will talk about in the Memory module, everything and anything is represented in terms of two things: 0 and 1. As a digital device, a CPU is thus limited to a set of discrete states, and requires some kind of switching elements to differentiate between and change states between 0 and 1. That is why a computer is finite, no matter how big its memory is, there are only a finite number of states, although huge, that we can represent in a computer. This finiteness leads to the fundamental weakness of any computer. 8
What to use? As long as it does the switching thing, the question is then what technology we should use to build up such a switch element. The basic requirement is that it has to be tiny and reliable. A light switch does switch a light bulb between on and off, and is pretty reliable, but too big, thus can t be used. 9
A small step forward The earlier processors consisted of lots of vacuum tubes as shown in the followoing figure. Question: Do we remember how many vacuum tubes that ENIAC used? Answer: 18,000. It was relatively small, but generates lots of heat, thus not reliable. In fact, it lasted about five and half hours on average between repairs. 10
A big one The design complexity of CPUs increased as various technologies facilitated building smaller and more reliable electronic devices. The first such improvement came with transistor, which can also be used as a switching element, but much smaller, reliable and faster, as compared with vacuum tube. Indeed, it is considered by many to be the greatest invention of the twentieth-century. With this technological progress, more complex and reliable CPUs were built onto one or several printed circuit boards containing discrete (individual) components. 11
Trickle, flow and flood Starting in the 1950 s and 1960 s, more complex and reliable CPUs were built onto one or several printed circuit boards, a way to build transistors. The rest is history. Essentially, people put more and more transistors on such circuit boards, going from SLI(small scale integration, MLI (medium scale integration, to LSI(large scale integration, when we put ten, hundreds, and thousands, transistors on a printed circuit board. At present, with the VLSI, very large scale integration, technology, as of early 2008, billiontransistor processors are commercially available, such as Intel s Montecito Itanium chip. 12
Moore s law... Since the 1970 s, the number of transistors that we can put in a computer ship has been doubled roughly every two years, the so-called Moore s law. As a result, the speed of computer has also been doubled about every two years. As an example, ASCI Red, a computer that Intel built for the Sandia National Lab in 1997, can carry out 2.1 trillion operations per second, a 21 million factor speed up over UN- VAC, the very first commercial computer built in 1951. On the other hand, the world record for 1 mile run was set in 1999 at 3.43.13, just a 7% improvement over the old record set in 1944, which was 3.59.4. 13
... and its end By doubling every eighteen months,, it essentially means that we have to make the wires 2 thinner every eighteen months, which cannot be done infinitely. Although every transistor produces only a tiny bit of heat, it does add up to reach a power limit. Remember the video on how hot it could be? We also have essentially dug out all the benefits of a complicated single processor architecture. It is expected that Moore s law will hold for another decade or two, while multi-processor architecture is emerging. 14
Activities Read some of the references as given in the book and look for others in the Internet and other books. Write about the following topics and we will discuss them in a group setting. 1. Find out at least five application of CPU which are not related to computers: Where are they used? How? 2. Justify the statement that Transistor is considered by many as the greatest invention of the twentieth-century. Who was the first to make this claim, where, and why? 3. What do you think are the ten most important inventions of the twentieth century in the sense that they fundamentally changed our lives? Why? 15