Let's look at just one project on WCG, the Help Fight Childhood Cancer project. At project launch in March 2009, project scientists said it would take them 100 years to do the computations on the computers available to them. (Also posted on their main page.) They estimated it would take us 2 years. In fact, it only took about 1 year and 3 months! This allowed them to identify some new targets and toss them our way, nearly tripling the scope of the project. So in the course of a few years, we've done way more than their resources could do in 100 years. Granted, if they had a shiny supercomputer hiding in the closet, they could pound out results too. They don't, so that's where we come in.

The way Boinc is setup is to run one workunit on each cpu, not one workunit on many cpu's. A Super Computer from the top500.org list is setup to run a single problem on many cpu's all at once bringing in effect a sledge hammer or a 10 ton hammer to the problem. For WCG, or any other Boinc Project, to run on a Super Computer would necessitate a whole rewrite of the software and would ONLY then work on Super Computers. Time on a Super Computer is EXTREMELY expensive and not easily obtained. In many cases people have to submit their problems several times before it is approved and even then it may not run for several months down the road. It is kind of like time on a top notch telescope, limited and very expensive! Also alot of the newer Super Computers are using gpu's to bump up their numbers, so while they may have 3,000 processors, 2,500 of those may be on gpu's as opposed to straight cpu's. My AMD 5770 gpu's has 800 'shaders' each, or as Super Computers call them 'processors', I have 6 of those running here in my pc's now! In short you and I supply FREE resources and the program is already written to take advantage of our hardware.

Latest Top 500 Super Computer round, where Linux rulez at > 85%. WCG still rates fairly well with, who would you believe, 448Teraflops per BOINCStats [around spot 150], but 881TFL [around spot 80] according to moi , as we are effectively still on credit old [though it says credit_new on the tin].

http://www.zdnet.com/article/linux-still-rules-supercomputing/

The list: http://top500.org/

Once upon a time we were in the top 10, sustained performance, but with expanding 'smartphone' computing, can't compete with Xeons. Still, not at all shabby.

using GRID computing is CHEAP, right now it's FREE / donated...supercomputers COST money, even if u get the time on it as academic...
here in Croatia we have one several supercomputers...most of them r v 50% capacity or under...not many request 4 them!

putting BOINC on supercomputer, depends on OS...if it's (usually is) UNIX/Linux, than BOINC can allocate the core per WU...most probably it will do same thing on Windows based...but haven't tried it, maybe someone operating some SC can tell more?

projects will be done within a months...'cause these beasts have so much cores that's incredible! but SC costs money to maintain, so scientists don't get so much access to them...& a lot of paperology is written to have it allocated...

GRID is donated & as such is so much cheaper than SC...