Computing 101 (and a little 201)
In my last post I used common computing terms such as gigahertz, gigabytes, teraflops and terabytes. Most people understand that these terms represent “fast” and “lots” but don’t really grasp how much speed and information is really on the line. I decided to break down these terms into something (I hope) anyone should be able to understand.
Let’s start with memory and hard-drive space:
One bit (the term comes from binary digit) is equivalent to a 1 or a 0. Eight bits equals one byte, which is the amount of memory needed to store one character of text. The letter “a” is represented as 01100001.
Next comes kilobyte (KB), which is 1024 bytes, then megabyte (MB) which is 1024 KB, then gigabyte (GB) which is 1024 MB, then Terabyte which is 1024 GB. Petabit, exabit, zettabit, and yottabit follow in the same pattern.
Let’s say an average MP3 file is 5 MB. If you have one terabyte of harddrive space you could fit 209,715 songs on your hard drive.
The IU supercomputer, with it’s 8 TB of memory, could have over 1.5 million MP3′s loaded into RAM at one time. Seeing they use this computer for research, it’s much more likely they would have huge numbers and program resources stored in memory, but I think you get the point.
Now lets talk about speed.
For most of computing’s history, at least the history that includes the personal computer, computer speeds have been measured in megahertz (Mhz). My first PC ran at a blazing 8 Mhz in 1987.
To understand Mhz you first need to know the definition of the word “hertz” (and I’m not speaking of the rental car company.) Hertz simply means one per second. It is not a term exclusive to computing. Last night, my site was receiving hits at the rate of 45 Hz or 45 hits per second.
One megahertz is equavalient to 1000 Hz. One gigahertz is 1000 Mhz. Simply put, a modern PC running at 3.0 gigahertz is oscillating (going from high to low voltage) 3,000,000 times per second. Now you know why so much heat is created by your CPU!
In the supercomputing world, CPU power isn’t the only important factor. The floating-point unit (FPU) or math co-processor plays a decisive role in a supercomputers ability to work quickly. Typical operations performed by the FPU are addition, subtraction, multiplication, division, and square root. CPU speed, measured in Mhz or Ghz is important, but the number of floating-point operations the FPU execute per second matters a great deal when you are crunching large numbers, which is usually the case in the research arena. FPU speed is measured in FLOPS (floating-point operations per second)
20.4 Teraflops, the theoretical processing power of “Big Red”, means it is capable of 20,000,000,000+ (20 trillion) floating-point operations per second. A single Pentium D (dual-core) running at 2.8 Ghz is capable of approximately 1400 Megaflops or 1,400,000,000 floating-point operations per second.
That’s the lesson for today. Before I inevitably get a comment saying my math is wrong, take into consideration it is late and I am tired. If you feel I am in error, let me know. Also, any comments on the “understandability” of this post would be of great help for future reference.
