The initial estimate may be quite wrong. It will be corrected over time as BOINC learns how long the tasks actually take.
You have two computers listed, an old single core Pentium 4 and a Pentium dual core. The P4 machine shows a completed task which took about two days. To me, that seems about right for that CPU running the new Arecibo, large tasks. For more information about this search, check the announcement thread in Technical news.
The dual core machine doesn't have any completed tasks. Looks like you aborted several after they had done some processing - perhaps an hour or so. You should ignore the estimate and allow one to complete. It would probably take between 1.5 - 2 days as well since it has a lower clock speed than the older P4.
The Arecibo, large tasks are probably your best choice for those two machines since the other CPU search (gamma-ray pulsar) has been announced to be finishing soon. Each task is designed to run for quite a long time since it's actually a bundle of 8 of the previous tasks. Check the announcement thread for what has been stated about all of this.
The Arecibo, large search has a 7 day deadline which should not be a problem if you set a moderate work cache size - eg. 0.1 day until the first task completes (to prevent initial overfetch) - followed by a change to a bit higher (say 1 day) once BOINC sees exactly how long these tasks take and adjusts any initial low estimate.
Your initial message contained very little information so I hope I'm guessing your actual concerns correctly.
The Arecibo, large tasks are probably your best choice for those two machines since the other CPU search (gamma-ray pulsar) has been announced to be finishing soon. Each task is designed to run for quite a long time since it's actually a bundle of 8 of the previous tasks. Check the announcement thread for what has been stated about all of this.
Gary do you have an idea of when this might happen, I could use some stats in this that might take more than 2 weeks or so to get.
BOINC has always needed 11 valid task to develop the correct APR index for an application on the hardware that is used to crunch it. Will have valid estimated runtimes after that.
But if the host also runs other applications, that throws a monkey wrench into the estimated runtimes as this project does not use the standard BOINC server software and does not use the APR index. It uses the older DCF index and only allows one DCF value that gets applied to all applications and tasks. So running a bunch of short work followed by long work will have the new long work with invalid and extremely long estimated runtimes.
Best is to just ignore the runtimes and just crunch the work.
Gary do you have an idea of when this might happen ...
When what might happen??
If you're talking about the CPU search called Arecibo, large - that's been running for a week or two now.
If you're talking about the proposed "BRP7" for GPUs, all I'm aware of is what was mentioned in the announcement thread. No firm time was given so it's anybody's guess.
Gary do you have an idea of when this might happen ...
When what might happen??
If you're talking about the CPU search called Arecibo, large - that's been running for a week or two now.
If you're talking about the proposed "BRP7" for GPUs, all I'm aware of is what was mentioned in the announcement thread. No firm time was given so it's anybody's guess.
Sorry I should have included this "since the other CPU search (gamma-ray pulsar) has been announced to be finishing soon."
Sorry I should have included this "since the other CPU search (gamma-ray pulsar) has been announced to be finishing soon."
No need to be sorry! If I hadn't been in such a hurry and read it a bit more carefully, I could have worked out you were probably most concerned about the "finishing up" bombshell that Bernd dropped for the long running GRP searches :-).
What was posted in TN doesn't really make it clear if both FRGP5 and FGRPB1G are finishing at the same time. I assumed both and I also assumed that the GPU search might end when the 3011L04 data file was done. However, yesterday a new file 3012L08 landed so I guess that will last at least for several more days. Who knows after that. I assume they won't actually drop FGRPB1G until "BRP7" is ready. I wonder if there is an 'EAH Time' just like 'Elon Time' where 'soon' could mean anything :-).
I'm not running the CPU search (FGRP5) so I have no idea how long that might continue. With the "Arecibo, large" now running for CPU users, they might drop FGRP5 fairly quickly.
Neal Watkins wrote:Download
)
The initial estimate may be quite wrong. It will be corrected over time as BOINC learns how long the tasks actually take.
You have two computers listed, an old single core Pentium 4 and a Pentium dual core. The P4 machine shows a completed task which took about two days. To me, that seems about right for that CPU running the new Arecibo, large tasks. For more information about this search, check the announcement thread in Technical news.
The dual core machine doesn't have any completed tasks. Looks like you aborted several after they had done some processing - perhaps an hour or so. You should ignore the estimate and allow one to complete. It would probably take between 1.5 - 2 days as well since it has a lower clock speed than the older P4.
The Arecibo, large tasks are probably your best choice for those two machines since the other CPU search (gamma-ray pulsar) has been announced to be finishing soon. Each task is designed to run for quite a long time since it's actually a bundle of 8 of the previous tasks. Check the announcement thread for what has been stated about all of this.
The Arecibo, large search has a 7 day deadline which should not be a problem if you set a moderate work cache size - eg. 0.1 day until the first task completes (to prevent initial overfetch) - followed by a change to a bit higher (say 1 day) once BOINC sees exactly how long these tasks take and adjusts any initial low estimate.
Your initial message contained very little information so I hope I'm guessing your actual concerns correctly.
Cheers,
Gary.
Gary Roberts wrote: . The
)
Gary do you have an idea of when this might happen, I could use some stats in this that might take more than 2 weeks or so to get.
BOINC has always needed 11
)
BOINC has always needed 11 valid task to develop the correct APR index for an application on the hardware that is used to crunch it. Will have valid estimated runtimes after that.
But if the host also runs other applications, that throws a monkey wrench into the estimated runtimes as this project does not use the standard BOINC server software and does not use the APR index. It uses the older DCF index and only allows one DCF value that gets applied to all applications and tasks. So running a bunch of short work followed by long work will have the new long work with invalid and extremely long estimated runtimes.
Best is to just ignore the runtimes and just crunch the work.
I'll try one. This is a new
)
I'll try one. This is a new problem (2 weeks or so). I have been crunching for a while now.
Over 1 million credits with E@H. They don't show in the stats....
You'll need to change a
)
You'll need to change a project setting that allows export of user stats to 3rd party aggregation sites.
Account >> Preferences >> Privacy >> Do you consent to exporting your data to BOINC statistics aggregation websites?:
Privacy Preferences
mikey wrote:Gary do you have
)
When what might happen??
If you're talking about the CPU search called Arecibo, large - that's been running for a week or two now.
If you're talking about the proposed "BRP7" for GPUs, all I'm aware of is what was mentioned in the announcement thread. No firm time was given so it's anybody's guess.
Cheers,
Gary.
Gary Roberts wrote: mikey
)
Sorry I should have included this "since the other CPU search (gamma-ray pulsar) has been announced to be finishing soon."
mikey wrote:Sorry I should
)
No need to be sorry! If I hadn't been in such a hurry and read it a bit more carefully, I could have worked out you were probably most concerned about the "finishing up" bombshell that Bernd dropped for the long running GRP searches :-).
What was posted in TN doesn't really make it clear if both FRGP5 and FGRPB1G are finishing at the same time. I assumed both and I also assumed that the GPU search might end when the 3011L04 data file was done. However, yesterday a new file 3012L08 landed so I guess that will last at least for several more days. Who knows after that. I assume they won't actually drop FGRPB1G until "BRP7" is ready. I wonder if there is an 'EAH Time' just like 'Elon Time' where 'soon' could mean anything :-).
I'm not running the CPU search (FGRP5) so I have no idea how long that might continue. With the "Arecibo, large" now running for CPU users, they might drop FGRP5 fairly quickly.
Cheers,
Gary.
6 day workunit completed in 2
)
6 day workunit completed in 2 days.
Neal Watkins wrote: 6 day
)
That's pretty good!!