No-Mad

Go to the power calculator down in the bottom left hand corner of this page at UD http://www.ud.com/rescenter/
Plunk in the values 13839 2.6Ghz P4 PC's running at a 90% donation to WCG

I think you'll find that it is not.

My other thread http://www.worldcommunitygrid.org/forums/wcg/viewthread?thread=839
acknowledges that there is redundancy in grid computing and explains that this is a measure of the raw computing power that our collective PC's are contributing to the cause.

Regards

Dave

Just double checking my figures and seeing if the UD calculator is correct.

Just benchmarked my PC with Sisoft Sandra and my PC alone produces 3.447 Gflops.

My PC is a 3200+ athlon scoring 29.2 points per CPU hour. The average score for us all for points per CPU Hour is 24.96 so the average PC is a 24.96/29.2 * 3200 = 2735 Mhz or roughly equal to 2.7 Ghz P4. My average utilisation of my procesor is about 2% so the remaining 98% is going to WCG.

The closes figure available in the UD power calculator is a p4 running at 2.6Ghz (on the low side) selecting what I think is on the low side for spare capacity of 90% donated . Saturday we did 37 years 334 days of processing that is the equivalent of 37*365+334 = 13839 days of processing meaning a WCG average computer on for 13839 days or as in our case 13839 2.6GHz P4's on for a day. Enter those figures into the power calculator and yes a paltry 64.77 Tereflops.
91.5% the raw number crunching power of Blue Gene just 19 days after launch.

I was trying to cheer the forum up with this news.

Regards

Dave

Here is an article titled This Supercomputer's No Teraflop at http://www.business2.com/b2/web/workingtech

It is generally possible to speed up a computer 2 or 3 orders of magnitude using special architecture if the problem can be vectorized. However, there are a lot of problems that cannot be defined this way. Or at least, there is not an obvious way. In which case, the stodgy PC is not out of the running if a LOT of PCs can be used.

Looking at the cost of maintaining a supercomputer mentioned in the article, you can see how easy it must be to restrict their use to a few favored projects / institutions / ‘Name’ scientists. I remember reading an article by a senior scientist back in the 1980s in which he said that when he was younger he worried about the horde of scientists being turned out by the Soviet Union’s educational system. But as the decades passed he saw the same pattern followed again and again. A new institution would be built, and turn out a burst of new discoveries. Then results would trickle off as the power hierarchy ossified and all scarce resources became ‘owned’ by a few names who dedicated them to their pet projects, working over the same old territory.

So clusters of computers like the World Community Grid make a lot of sense in the first half of the 21st century. Each such organization can look at proposals without a lot of built-in preconceptions. And the PC contributors can vote by moving from one cluster to another with a simple Grid Agent download if that seems called for. It ought to help keep scientific research free and unbounded. When the Manhattan Project was made public, some members of Congress wondered if that represented the future of scientific research. The famous reply was that narrowly directed research projects of that sort were the least efficient way to use research funds. The most efficient way to advance is to direct resources where the bounds of science are being pushed out. This is constantly changing and requires extreme flexibility. Flexibility practically defines organized clusters of computers such as the World Community Grid.

A) If clusters are worthless, why are you wasting your time on these forums?

B) Why should we trust you, you have provided nothing more than a statement, where is your verfication, proofs?

C) You are saying that the cluster projects are a bunch of junk, as a general statement, or specific applications in specific?

D) Let's assume your calculations are correct, 70 teraflops is closer to 5. Even then, x teraflops to the researchers is x more teraflops of computer power than they had before, and at low/no (not sure of agreement between WCG and researchers) than they had before, do they not?