What is the fastest computer in the world and how does it relate to public computers like the pentium ,amd etc

Question

I think the fastest is the Bluegene but I cannot appreciate how fast it is until i can relate it to what we use.like i hav a 2.66GHZ computer.whats the bluegene?like 10000GHZ or something.I need figures like that to appreciate how fast it is.Teraflops mean nothing to me.

Answer 1

One thing you haven't yet learned is that raw gigahertz doesn't necessarily translate to fast computing. For example, many Athlon XP's and 64's from the last 3-4 years ran much slower than Pentium 4's that they were smoking the crap out of.

Efficiency of the "pipeline" and the alogorithms used to process data from cache matter just as much as raw speed.

Look at the newer dual-core CPUs for example. They run slower than the fastest single-core models, but in many situations outperform them.

Check Wikipedia for "supercomputer" and I'm sure you'll find out more about what you're looking for...

Answer 2

I suggest you get used to FLOPS because that's one of the primary measurements used to compare supercomputers. Supercomputers are built for scientific work, so they're designed for mathematically intensive work, and that kind of work is measured in FLOPS.

What are FLOPS? FLOPS measure FLoating point OPerations per Second. A single FLOP is one multiplication and one addition. This is a basic operation in matrix and vector math.

If you have two vectors (a,b,c) and (d,e,f) you take their dot product like this:

dotproduct = a*d + b*e + c*f

The "a*d +" part is a single FLoating point OPeration, or FLOP. Effectively, the whole dot product is three FLOPs (if you throw in an extra "+ 0" to round it out).

If you multiply two 3-by-3 matrices you're going to perform 27 operations. Each of the nine matrix elements will be the sum of three pairs of numbers added to each other (each element in the product matrix is the dot product of a row in the first matrix and a column in the second matrix). To learn more check out a book that talks about matrices or linear algebra.

Why not gigahertz? Modern processors, between pipelining, speculative execution, multiple cores, and other tricks do so many weird manipulations on the instructions it executes it's tough to determine exactly how many operations are being performed at any given point. Marketers and end users love clock speeds because it's a simple number which they can wrap their head around, but once you start learning about modern computer architecture you'll find that the clock speed is only the beginning of the story and a lot more goes into evaluating computer performance.

Check out the Top 500 website. It keeps track of the standings among the top scientific computers in the world. As you said in your question, IBM's Blue Gene is at the top of the most recent list, which was first reported at the Supercomputing '06 conference last year in Tampa, Florida.

To get a comparison you might want to do a little web research to find FLOPS measurements on the CPU in your computer, or get a numerical benchmark program and run it.

Answer 3

I saw something a couple of years ago that had multiple processors running in unison, sharing eachothers power. It was like five thousand dollars just in parts.