Message boards : Number crunching : Why difficult Points for the same time
Message board moderation

To post messages, you must log in.

AuthorMessage
tom

Send message
Joined: 12 Nov 06
Posts: 4
Credit: 501,420
RAC: 2
Message 25146 - Posted: 25 Jan 2013, 19:00:51 UTC
Last modified: 25 Jan 2013, 19:01:28 UTC

Hello,
i have any WU's,
ex:
sse3 (5825354; 21,287.58 sec ; 165.92 points )
sse2 (5852212; 21,204.30 sec; 104.85 points )
pni (5887555; 21,290.52 sec; 166.24 points )
for the same Workime i became different Points.

WHY?????

Tom
ID: 25146 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 810
Credit: 654,554,944
RAC: 253,993
Message 25160 - Posted: 26 Jan 2013, 17:54:43 UTC - in response to Message 25146.  

I'd say the SSE3 & PNI ones are within experiment error.

sse2 one is lower because your wingman wash shorter CPU time.

the credit awarded is an average of the 2 computers times so it can vary due to the averaging process.
ID: 25160 · Report as offensive     Reply Quote
tom

Send message
Joined: 12 Nov 06
Posts: 4
Credit: 501,420
RAC: 2
Message 25174 - Posted: 27 Jan 2013, 15:39:22 UTC - in response to Message 25160.  

Hello,
the WU 5852212
my Wingman have 19,457.10 sec.
I have 21,204.30 sec.
Ok.
But why make thats a different from 60 Points ?

Tom
ID: 25174 · Report as offensive     Reply Quote
Toby Broom
Volunteer moderator

Send message
Joined: 27 Sep 08
Posts: 810
Credit: 654,554,944
RAC: 253,993
Message 25191 - Posted: 28 Jan 2013, 3:16:02 UTC

It does seem to be quite a few point difference for not much time difference.

I don't have a good explanation of why though.

I have one here that's a much bigger spread but not so much on points:

http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=6140477

Here's a huge one!

http://lhcathomeclassic.cern.ch/sixtrack/workunit.php?wuid=6146677

Maybe it's an SSE2 vs SSE3 thing?
ID: 25191 · Report as offensive     Reply Quote

Message boards : Number crunching : Why difficult Points for the same time


©2024 CERN