Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 5
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 2090 times and has 4 replies Next Thread
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Computational performance of the grid

I'm a new member (came in with the /. boom biggrin ) and I can't find any information on the computational performance of the grid.

What I'm looking for is something that could be compared with supercomputers (eg. the ones found here: http://www.top500.org). It'd be nice to know where we would be on the list.

Or is calculating a rough mips/flops -figure even possible?
[Jan 2, 2005 3:51:51 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
cool Re: Computational performance of the grid

If you go over to http://www.ud.com/solutions/outsource/ you will find a link to The Power Calculator on that page for estimating the power of a grid. You can try entering in figures from the Global Statistics pages to make rough (very rough) estimates. Incidentally, you might want to look at the statistics for the Human Proteome Folding Project on http://www.grid.org/about/ while computing this.

Lawrence
[Jan 2, 2005 4:24:09 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Computational performance of the grid

2GHz Intel Pentium 4 has theoretical power of 2400mflops, so if every user here would have that kind of processor the total power of our machines would be 2400mflops * 55 000 = 130tflops. At this time the most "powerfull" supercomputer in the world is IBM's BlueGene with power of 70tflops. It only has about 33 000 processors laughing
[Jan 2, 2005 4:29:30 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: Computational performance of the grid

Last time I checked there were roughly 55k devices connected. I don't have any idea on what kind of machines people have nowadays, but this is what I inserted into the calculator:

10000 Intel P4 2GHz
10000 AMD Athlon 2GHz
10000 Intel Celeron 2GHz
10000 Intel P3 1.5GHz
10000 AMD Duron 1GHz

Also, utilization for every CPU was 50%.
This is what I got:

55 Teraflops

That's quite impressive, considering it is only second to IBM's BlueGene/L @ 70.72 Teraflops (http://www.top500.org)!
[Jan 2, 2005 4:45:33 PM]   Link   Report threatening or abusive post: please login first  Go to top 
David Autumns
Ace Cruncher
UK
Joined: Nov 16, 2004
Post Count: 11062
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: Computational performance of the grid

Just to back up arlehdon calculations

The "average" PC on the WCG is a Athlon XP 2700+ going on the average points per hour calculation on the global stats. Yesterday together we did 51 years 320 days of processing which is the equivalent of 18935 average PC's on for the full 24 hours. A 2700+ atlon will benchmark at around 2.9Gflops so with 18935 of them I too make that 55 Teraflops of raw number crunching power which only Blue Gene /L can beat at 70.72 Teraflops.

It has to be pointed out that due to nature of grid computing there is some redundancy in the task as the results have be verified as correct - this is medical science we are carrying out here - I believe that each work unit is allocated to a different PC on 5 occasions. So apart from the other work carried out by your PC which might take a max of ~5% and the time each PC takes to return the results when it is not crunching at that time. The real power of the grid - at the moment - is about 10Teraflops which is nothing to be sniffed at.

Keep on crunching it won't be long until we match BlueGene/L smile

Dave
----------------------------------------

[Jan 3, 2005 1:00:11 PM]   Link   Report threatening or abusive post: please login first  Go to top 
[ Jump to Last Post ]
Post new Thread