Odd recent avg. credit situation
14 Nov 2007 7:54:41 UTC
Topic 193314
(moderation:
Perhaps one of you folks could explain to me why this machine http://einsteinathome.org/host/952086 has a rac that is half of this machine http://einsteinathome.org/host/728417 Both of them run 24x7 and do not go into standby mode. It is odd that the slower machine has the higher rac. The preferences are set very similarly, the slow one with the @work pref. and the faster one with the @home prefs.
Thank you
Anton
Odd recent avg. credit situation
)
What else is either machine doing? Attached to additional other projects?
And even then, do both machines crunch the same (or comparable) dataset(s)?
RE: Perhaps one of you
)
The [[RAC]] is an average of your credits over the last few weeks (see the WIKI link) and can take a few weeks to settle (see the charts in the WIKI article).
The two computers you mentioned didn't start simultaneously, in a week or two the RAC should reflect the real performance.
CU
Bikeman
As mentioned, the RAC
)
As mentioned, the RAC responds to changes over a long period
boincstats
This shows that there is a clear difference in output over the past month between the hosts concerned.
The machines are not running
)
The machines are not running any other projects. The faster one (via benchmark) has been running Einstein at home for over two years, the slower one, for about one year. I did have some trouble when I upgraded to a newer version of boinc. I installed the newer version into a different directory (default) by mistake. The stats show a credit for tyanworkstation of 36,470 when it should be 204990.
My total credit thus far is 600726. Could that have anything to do with it?
I like that stats link.
PS
I still have the old directory on disk.
RE: The machines are not
)
You said earlier that you run both machines 24/7. When you look at your results lists for each machine, your dual tualatin is clearly crunching a result faster than your dual coppermine - which is as it should be. However there are about twice as many completed results in the list of the slower machine over the last 7 days or so. I would think there are basically two possible explanations for this.
Either -
1. The dual tualatin is not crunching 24/7
Or -
2. There is something else running on the dual tualatin which is consuming large numbers of CPU cycles.
You need to investigate more closely what is happening on the dual tualatin box.
Cheers,
Gary.
RE: The machines are not
)
Mystery solved, click here.
In this view, you see all hosts, not just those active in the past 30 days. Your hosts were assigned new IDs so there exist now different copies of the PCs. The newer copies have not yet reached their maximum RAC because the averaging over time effect. You could try to merge the hosts using the links in the respective details view.
CU
Bikeman
RE: Mystery solved ....
)
No, not fully :).
Huff can find his total credit by adding up the individual totals for the two instances of his dual tualatin but this doesn't at all explain why the RAC for the current instance is so low. After all, the old instance last communicated back in May (6 months ago) so there has been plenty of time for the current instance to have achieved a stable RAC if it was indeed crunching 24/7 with no other projects onboard.
If you look at the credit received since May 24, it averages about 200 per day. The RAC is now around 130 which is significantly below the long term average. There is something else consuming cpu cycles on that machine.
Cheers,
Gary.
My tip: Both computer
)
My tip: Both computer summaries show 2 CPUs, but only one of them runs Einstein@Home on two cores.
RE: My tip: Both computer
)
I also had considered this. The OP said that both machines were crunching 24/7 with no other projects on board. He also said that he was using two different venues. Both machines are identified as having two processors. There is a general preference for how many processors to use on multi-processor machines. If that preference were set to 1 for the first venue but to 2 for the second venue, wouldn't that fact be shown on the respective summary pages??
I'm not sure of the answer to that question. I had thought that the summary page showed what BOINC was actually using, but maybe I'm wrong ....
Cheers,
Gary.
RE: RE: My tip: Both
)
The number of cpu's is determined by the global_prefs_override.xml file if the menu advanced/preferences has been used.