Harnessing the Power

Duel-cores, triple-cores, quad cores, sex-cores (that's six for those who don't know their prefixes), it's hard to keep up with processing power these days.  I remember when my Dad bought our first family PC.  It was a Gateway 2000 and it was a serious toss-up as to what I liked more; solitaire on the computer or the cow painted box.  And I remember getting my first major PC game, NCAA Football '98 and it couldn't run on our computer.  It required the old Pentium II chip.  Well now I have upgraded my personal PC to the AMD Phenom X4.  Granted it is still on the old 65nm architecture as compared to the new Phenom II's that have 45nm, but it is quite fast.  And now both Intel and AMD have the "sex-core" processors on the market between the Phenom II X6 and Core i7.  "But Matthew, why do I care?"  You probably don't but I wanted to mark a reference for you when I explain the "computer" I am working on in my Computational Chemistry course.

Computational Chemistry is exactly what it sounds like.  It uses computer algorithms to calculate how particles act.  More specifically their energy in my current lab.  Now, I was frustrated that the lab consisted of a Linux Shell program not because it was Linux, but because we were given no formal background on coding in the system.  I think Linux is simpler and I hope to utilize it some day in the future but I began ranting and raving to a few friends about how I don't understand why we can't pay Computer Scientists to make programs for Windows/MAC that I can just input data into and read the output without being a coding expert.  This is a carry over from my statistical courses as they employ the same thing into programs such as SAS.

In my frustration I found myself searching for a "help section" for this shell program.  The program/project is called Steele and it is a compute cluster that can be used remotely on campus for calculations.  It is basically access to a computer that links many processors to together in order to make elaborate calculations housed on campus at Purdue.  This is where I found my surprise.  The Steele compute cluster (named after an old professor) houses an astonishing 902 8-core Dell processors.  902! That is 7216 cores for those of you keeping score at home!  Each processor consists of two Quad-cores that are linked together.  To a pion like myself, that is almost unfathomable.  Now I understand why we use this program.  The cores are separated into further clusters, the largest one being able to compute 46.53 teraflops (basically a calculation per second, it is more complicated than that but I'm not an expert and will probably mess it up).  A Phenom processor similar to mine was clocked at 6.7 gigaflops.  This is about a 7,000,000% increase in computations.  Wow!

Now, here's where I stopped thinking about these types of calculations...I ran a simple program that uses a random number generator to calculate the energy of a particle 10,000 times and it took the computer about 20 seconds to run this program. The computer with 7216 cores needed time to think about my simple little calculation. Now there are scientists running much, much larger calculations on much larger computers.  That is when I had to stop thinking about the magnitude of this computer and starting writing down answers.  So sorry Dad for complaining about not having that Pentium II processor.