Dr. Mike. GPU means Graphic Processing Unit. That is the processor that is designed to take information from inside the computer and present it on the display so the user can see it.

CPU means Central Processing Unit. That is the traditional brain of the computer. When people talk about dual-core or Quad-Cores this is the unit that they are talking about. In some computers, the GPU is just added on the motherboard and shares main memory with the other components. The CPU then does most of the work associated with presenting information to the user.

Now some years back they started putting these Graphic Processing Units to work doing the display thereby releasing the CPU for the true computational tasks. Of particular note were the workstations and the gaming PCs. It was taking so much of the CPU's time just trying to keep the screen up to date that the overall applications, such as CAD or Games ran slowly. So the idea was to relieve the CPU of the duties associated with display and let it compute while the GPU took care of all the display issues.

This progressed over the years to deal with the shading and shadows and other nuances commonplace in the newer games and other applications to the point that the Graphic Processing Units have become very powerful and, because they are specialized processors (not having to deal with all the variety of tasks that the CPU has to deal with) they have become very fast.

Next came the idea of employing these specialized GPUs in other ways (other than just doing the display). Folks, as folks always will, figured out a way to use the GPUs to do other things. In this case, they have figured a way to use these GPUs to do crunching. Now we have the GPUs crunching work units. These processors have a limited instruction set that they run very fast (as was pioneered by something called RISC* processors some years back) and how well they compete with the Generalized CPUs (sometimes called CISC) in getting work done/cures found, I do not know.

There is a lot of disagreement between the various people involved with these alternatives. These GPUs are very fast, but are they efficiently employed to do crunching? I don't really know but I can tell you that "it depends" It depends on how good the programming is on the two different processor types, it depends on how applicable the problem is to the instruction set offered by the processor types and perhaps the biggest variable of all....It depends on who you are talking to.

My final point is about points. Points are points. You cannot pay your electric bill or buy Twinkies with them. They are an arbitrarily assigned number based on some measure of work done at some level independent of whether or not they actually progress to an ultimate goal. Like games; 100 points is one game are as hard to get as 100,000,000 are in another.

Are points comparable from one project to another? That is highly debatable. My considered opinion is: Not really. It is just like the two previously mentioned games. One project assigns points one way and another another and there is no real way to associate a point with the progress it represents towards a cure/solution but rather some measure of instructions executed (effective or not).

Now points are fun to tease with and banter about but they really do not do much of anything towards enlightening us relative to the true contribution we are making to find the cure(s)/solution(s). Like a college football game, it is fun to get passionate about it and all that, but the team with the most points does not tell you which is the best school or produces the "best" graduates, or charges the most tuition or almost any other important attribute about the school.

So CPU and GPU are types of processors that are employed to do all kinds of computing, including crunching.
WCG and SETI and many others (sometimes just called GPU- which is confusing to most people) are different organizations that provide facilities (servers, software and projects/Work Units) to crunch.

This note may be controversial and precipitate comment relative to the whole set of issues that I have raised. I hope it and any subsequent discussion helps clear up any confusion.

*RISC – Reduced Instruction Set Computer, CISC – Complex Instruction Set Computer.

Like a college football game, it is fun to get passionate about it and all that, but the team with the most points does not tell you which is the best school or produces the "best" graduates, or charges the most tuition or almost any other important attribute about the school.

You'd better be careful RT. In some Texas counties, saying such a thing is a hangin' offense.
Now back to your regularly scheduled GPU discussion.

- D

Well - I don't know what to say. Sorry I brought up CREDITS at all.
We don't to get see what they do with the massive amounts of computing power they get for free, so all we have is CREDITS.
I'll never mention it again...

Edit - Man, GPUGrid sure hands them out, though!

Edit Again - How are CREDITS a motivator against yourself, anyway? I think most people use CREDITS to compare themselves against other people. And it is useful to compare CREDITS across projects. I know that the CREDITS are worthless, but at the very least they can be used as a value of appreciation. That's the way I take it, anyway...

Are CREDITS comparable from one project to another? That is highly debatable.

Thank you, Dataman - I stand humbly corrected.