Message boards : ATLAS application : Credit for ATLAS
Message board moderation

To post messages, you must log in.

Previous · 1 · 2

AuthorMessage
Harri Liljeroos
Avatar

Send message
Joined: 28 Sep 04
Posts: 683
Credit: 44,046,417
RAC: 18,067
Message 46964 - Posted: 28 Jun 2022, 16:28:20 UTC - in response to Message 46944.  

I see huge increase of credits for the new 2.01 app. It will probably drop after a while as 'credit new' starts to 'learn' these new tasks.
ID: 46964 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 2129
Credit: 160,200,793
RAC: 30,762
Message 46965 - Posted: 28 Jun 2022, 16:41:47 UTC - in response to Message 46964.  

But we are back to 2.00. So, once again with the points.
ID: 46965 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 2129
Credit: 160,200,793
RAC: 30,762
Message 47323 - Posted: 28 Sep 2022, 9:46:55 UTC

ID: 47323 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 2129
Credit: 160,200,793
RAC: 30,762
Message 47572 - Posted: 3 Dec 2022, 13:33:11 UTC - in response to Message 47323.  

ID: 47572 · Report as offensive     Reply Quote
Erich56

Send message
Joined: 18 Dec 15
Posts: 1698
Credit: 105,311,619
RAC: 63,552
Message 47821 - Posted: 8 Mar 2023, 14:03:44 UTC
Last modified: 8 Mar 2023, 14:26:08 UTC

Interesting development of credits allocation between yesterday and today (crunched on the same machine, no changes whatsoever inbetween):

Upload date/time / total runtime / CPU time / credit points

7 Mar 2023, 9:10:07 UTC - 121,382.46 - 120,750.00 - 1,996.63

8 Mar 2023, 13:39:59 UTC - 107,433.44 - 106,845.40 - 211.93

can anyone offer a logical explanation for a such huge difference?

P.S.: I noticed just now:
In case of the 1.996,63 credit points, stderr under "max FLOPS of device" shows 37,11 GFLOPS,
in case of the 211.93 credit points, it shows 4,65 GFLOPS.

How can this come about? As said, we are talking about the same machine and the same environment.
Further, the indicated GFLOPS are never ever in alignment with the runtims / CPu times of the tasks.
So, something seems to be rather wrong with the various calculation in the ATLAS code :-(
ID: 47821 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 2129
Credit: 160,200,793
RAC: 30,762
Message 47822 - Posted: 8 Mar 2023, 17:11:49 UTC - in response to Message 47821.  

Seeing no difference (10 Cpu's) one hour runtime and about 1k points.
ID: 47822 · Report as offensive     Reply Quote
Erich56

Send message
Joined: 18 Dec 15
Posts: 1698
Credit: 105,311,619
RAC: 63,552
Message 47823 - Posted: 8 Mar 2023, 17:27:49 UTC - in response to Message 47822.  

Seeing no difference (10 Cpu's) one hour runtime and about 1k points.
???
ID: 47823 · Report as offensive     Reply Quote
maeax

Send message
Joined: 2 May 07
Posts: 2129
Credit: 160,200,793
RAC: 30,762
Message 47824 - Posted: 8 Mar 2023, 17:37:15 UTC - in response to Message 47823.  

350 Atlas-Tasks a day with 1k are 350k, sorry this is possible.
Number one and two of our Computer stats.
https://lhcathome.cern.ch/lhcathome/top_hosts.php
ID: 47824 · Report as offensive     Reply Quote
Profile Yeti
Volunteer moderator
Avatar

Send message
Joined: 2 Sep 04
Posts: 453
Credit: 193,576,736
RAC: 1,011
Message 47829 - Posted: 9 Mar 2023, 13:19:32 UTC - in response to Message 47821.  
Last modified: 9 Mar 2023, 13:20:18 UTC

In case of the 1.996,63 credit points, stderr under "max FLOPS of device" shows 37,11 GFLOPS,
in case of the 211.93 credit points, it shows 4,65 GFLOPS.

How can this come about? As said, we are talking about the same machine and the same environment.
Further, the indicated GFLOPS are never ever in alignment with the runtims / CPu times of the tasks.
So, something seems to be rather wrong with the various calculation in the ATLAS code :-(

Nope, it is not the Atlas code, but something on your machine.

The BOINC-Client makes Benchmarks from Time to Time, and the latest Benchmark was done, when the machine was not really idle or the older one was wrong.

Enshure that your machine is real idle and then initiate BOINC to benchmark again.

Maybe this fixes your problem


Supporting BOINC, a great concept !
ID: 47829 · Report as offensive     Reply Quote
computezrmle
Volunteer moderator
Volunteer developer
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 15 Jun 08
Posts: 2437
Credit: 229,375,778
RAC: 131,008
Message 47830 - Posted: 9 Mar 2023, 13:37:08 UTC - in response to Message 47821.  

It's (nearly) a factor of 8 between 37.11 and 4.65.
This makes me guess you modified the #cores at the prefs page:
8 (or unlimited) -> 1

The server will compensate this step by step if you keep the setting stable but it will take a while.
ID: 47830 · Report as offensive     Reply Quote
Erich56

Send message
Joined: 18 Dec 15
Posts: 1698
Credit: 105,311,619
RAC: 63,552
Message 47833 - Posted: 10 Mar 2023, 16:50:35 UTC - in response to Message 47830.  

It's (nearly) a factor of 8 between 37.11 and 4.65.
This makes me guess you modified the #cores at the prefs page:
8 (or unlimited) -> 1
Well, you are perfectly right :-) I found out about this lateron.
ID: 47833 · Report as offensive     Reply Quote
Previous · 1 · 2

Message boards : ATLAS application : Credit for ATLAS


©2024 CERN