Einstein started Work

kloeniksen
kloeniksen
Joined: 28 Jul 05
Posts: 7
Credit: 2067071
RAC: 0
Topic 194432

Huzzay, after 2 months of "consider aborting" countdown is normal, for me anyway :
Elapsed 37:45
Progress 21,4%
To completion 58:00
Report deadline 13.7

Just another regular guy: Einstein 600/SETI 400
XP Pro Sp3 1,5G
6.6.36
SETI since 99
Einstein since 05

I don't know what I don't know

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5883
Credit: 119065788000
RAC: 24450697

Einstein started Work

Quote:
Huzzay, after 2 months of "consider aborting" countdown is normal, for me anyway :


Care to tell us a little more about what problem you've been having that would cause you to "consider aborting"?

Quote:
Elapsed 37:45
Progress 21,4%
To completion 58:00


This doesn't seem right for a 1.5GHz machine, although it's rather hard to judge when you don't give any other detail about your hardware and your computer is hidden. Also, are you talking about the normal GW search or the Arecibo binary pulsar search (ABP1)?

As an example for you to consider, I've just checked a PIII 1.0GHz host of mine and it is doing the GW tasks in around 24 hours. At your current rate of progress, the task you mention is going to take around 176 hours to complete and this seems far too excessive to be "normal".

If you care to give us a few more details, it should be possible to work out why your machine's performance seems to be so poor - is it some sort of low power (laptop/netbook) processor?

Cheers,
Gary.

kloeniksen
kloeniksen
Joined: 28 Jul 05
Posts: 7
Credit: 2067071
RAC: 0

Boinc, not me, considers

Message 93583 in response to (parent removed)

Boinc, not me, considers aborting-because time limit is not met. Accordingly I abort. This is the 1st instance in 2 months that make anything nearing a "normal" progress so far, although I agree it's far from expectations.
The task running now is an S5R5A, the tasks I have aborted were, if I remember right ABP1's. Part of the problem could lie there, I know there is a setting but have not experimented with that yet.
Details:
IBM Thinkcentre DESKTOP 8191
Intel Celeron 2.00, IBM Bios 01/10/05
Some other Points: I am running NIS 09 which classified 1st download of ABP1 as malware. I cleared the file (directory) in NIS after reading up, should not be an issue anyway?
I only run SETI and Einstein, while testing I have set all SETI suspended, +no new work from either on this computer.
Also set to run "always" on this test, normally some lean preferences.
Let me also repeat that Einstein has run "normally", that is slow but within limits up to sometime in May.

Details as I speak:
Elapsed 55:25
Progress 31,5
To completion 67,15

Thanks for taking an interest, looking forward to more info.
BRGDS
K

Quote:
Quote:
Quote:
Huzzay, after 2 months of "consider aborting" countdown is normal, for me anyway :

Care to tell us a little more about what problem you've been having that would cause you to "consider aborting"?

Quote:
Elapsed 37:45
Progress 21,4%
To completion 58:00

This doesn't seem right for a 1.5GHz machine, although it's rather hard to judge when you don't give any other detail about your hardware and your computer is hidden. Also, are you talking about the normal GW search or the Arecibo binary pulsar search (ABP1)?

As an example for you to consider, I've just checked a PIII 1.0GHz host of mine and it is doing the GW tasks in around 24 hours. At your current rate of progress, the task you mention is going to take around 176 hours to complete and this seems far too excessive to be "normal".

If you care to give us a few more details, it should be possible to work out why your machine's performance seems to be so poor - is it some sort of low power (laptop/netbook) processor?



I don't know what I don't know

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6592
Credit: 331745632
RAC: 315279

RE: Boinc, not me

Message 93584 in response to message 93583

Quote:
Boinc, not me ............
Details:
IBM Thinkcentre DESKTOP 8191
Intel Celeron 2.00, IBM Bios 01/10/05
.........


Hmmmm. What are your power saving settings? Is 'sleep' or 'hibernation' selected for anywhere?

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter ...

... and my other CPU is a Ryzen 5950X :-) Blaise Pascal

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5883
Credit: 119065788000
RAC: 24450697

RE: Boinc, not me,

Message 93585 in response to message 93583

Quote:
Boinc, not me, considers aborting-because time limit is not met.


OK, I understand now.

Quote:
Accordingly I abort.


Actually, you don't necessarily need to abort. BOINC is simply giving advice based on past performance and without any knowledge of what will happen in the future. You can easily choose to ignore that advice if you know you can change operating conditions in such a way that the task will complete in time. Even if you think you may go slightly over the limit - even up to a day over - there would be a very good chance of the result being accepted.

Quote:
This is the 1st instance in 2 months that make anything nearing a "normal" progress so far, although I agree it's far from expectations.
The task running now is an S5R5A, the tasks I have aborted were, if I remember right ABP1's.


There is something drastically sub-optimal with the performance of your machine. It should be able to complete GW tasks in around 20-30 hours rather than 55.42/0.315 = 176 hours. a possible cause of the problem could be some form of CPU throttling or bus frequency reduction, or some other rogue process that is consuming most of the available resources. One simple check you could make is to run a utility like CPU-Z and confirm what speed your processor is running at. Another thing to do is compare wall clock time with the elapsed time reported by BOINC Manager. Does 1 hour of "real" time correspond pretty closely with 1 hour of CPU time as reported by BOINC?

Quote:
Part of the problem could lie there, I know there is a setting but have not experimented with that yet.


ABP1 tasks will take longer on your machine but that's not the issue. If your machine were running at its proper potential, you would be having no issues with either type of task.

Quote:
I am running NIS 09 which classified 1st download of ABP1 as malware. I cleared the file (directory) in NIS after reading up, should not be an issue anyway?


Security programs/virus scanners do quite often find "false positives" in the downloaded files from many projects. If you are attached to a reputable project like E@H you can safely exclude the BOINC tree from being scanned by these programs. This is probably the easiest thing to do to mitigate this problem.

Quote:
I only run SETI and Einstein, while testing I have set all SETI suspended, +no new work from either on this computer.
Also set to run "always" on this test, normally some lean preferences.


If you are currently running on "generous" preferences, I'd hate to experience what it would be like if you put it back to "lean". :-).

Quote:
Let me also repeat that Einstein has run "normally", that is slow but within limits up to sometime in May.


And let me repeat that there is something drastically wrong with what you consider to be "normal". Because your machine is "hidden" it's rather hard to be specific about what the problem might be.

Cheers,
Gary.

kloeniksen
kloeniksen
Joined: 28 Jul 05
Posts: 7
Credit: 2067071
RAC: 0

OK I should now be

Message 93586 in response to message 93585

OK I should now be unhidden/updated. Ref ID 381946. Comments I hope.
BRGDS

Quote:
Quote:
Boinc, not me, considers aborting-because time limit is not met.

OK, I understand now.

Quote:
Accordingly I abort.

Actually, you don't necessarily need to abort. BOINC is simply giving advice based on past performance and without any knowledge of what will happen in the future. You can easily choose to ignore that advice if you know you can change operating conditions in such a way that the task will complete in time. Even if you think you may go slightly over the limit - even up to a day over - there would be a very good chance of the result being accepted.

Quote:
This is the 1st instance in 2 months that make anything nearing a "normal" progress so far, although I agree it's far from expectations.
The task running now is an S5R5A, the tasks I have aborted were, if I remember right ABP1's.

There is something drastically sub-optimal with the performance of your machine. It should be able to complete GW tasks in around 20-30 hours rather than 55.42/0.315 = 176 hours. a possible cause of the problem could be some form of CPU throttling or bus frequency reduction, or some other rogue process that is consuming most of the available resources. One simple check you could make is to run a utility like CPU-Z and confirm what speed your processor is running at. Another thing to do is compare wall clock time with the elapsed time reported by BOINC Manager. Does 1 hour of "real" time correspond pretty closely with 1 hour of CPU time as reported by BOINC?

Quote:
Part of the problem could lie there, I know there is a setting but have not experimented with that yet.

ABP1 tasks will take longer on your machine but that's not the issue. If your machine were running at its proper potential, you would be having no issues with either type of task.

Quote:
I am running NIS 09 which classified 1st download of ABP1 as malware. I cleared the file (directory) in NIS after reading up, should not be an issue anyway?

Security programs/virus scanners do quite often find "false positives" in the downloaded files from many projects. If you are attached to a reputable project like E@H you can safely exclude the BOINC tree from being scanned by these programs. This is probably the easiest thing to do to mitigate this problem.

Quote:
I only run SETI and Einstein, while testing I have set all SETI suspended, +no new work from either on this computer.
Also set to run "always" on this test, normally some lean preferences.

If you are currently running on "generous" preferences, I'd hate to experience what it would be like if you put it back to "lean". :-).

Quote:
Let me also repeat that Einstein has run "normally", that is slow but within limits up to sometime in May.

And let me repeat that there is something drastically wrong with what you consider to be "normal". Because your machine is "hidden" it's rather hard to be specific about what the problem might be.


I don't know what I don't know

kloeniksen
kloeniksen
Joined: 28 Jul 05
Posts: 7
Credit: 2067071
RAC: 0

RE: RE: Boinc, not me

Message 93587 in response to message 93584

Quote:
Quote:
Boinc, not me ............
Details:
IBM Thinkcentre DESKTOP 8191
Intel Celeron 2.00, IBM Bios 01/10/05
.........

Hmmmm. What are your power saving settings? Is 'sleep' or 'hibernation' selected for anywhere?

No sleeping on my accounts-:)

Cheers, Mike.


I don't know what I don't know

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5883
Credit: 119065788000
RAC: 24450697

RE: OK I should now be

Message 93588 in response to message 93586

Quote:
OK I should now be unhidden/updated. Ref ID 381946. Comments I hope.


You have two machines, the celeron 2.0GHz and a P4 2.66GHz. I would guess that the P4 should take around 14-20 hours per task and the celeron around 22-30 hours. There is quite a bit of variability between different tasks so it's hard to be precise.

Neither machine currently has a completed task to look at but do you recall approximately how long the P4 has been taking? I'm interested to know if perhaps both machines are performing well below average.

I asked previously and didn't get an answer - is the increment in CPU time (elapsed time) for a crunching task as shown in BOINC Manager pretty much in accord with the increment in wall clock time?

I checked the benchmarks of both your machines - the celeron shows 1048/1890 and the P4 1302/2499. These values are pretty much in line with those of similar machines that I run and so there seems to be no problem with your hosts performing as they should, at least as far as benchmarks are concerned.

At first, I was tempted to think that perhaps you had severely throttled your celeron through your preferences. Can you please check your computing preferences on the website to see if you have "use at most 100% of CPU time" set and not some drastically lower value. I can't see that it can be this because in that case the elapsed CPU time would still be normal even though the wall clock time would be very much larger. The figures you have previously listed show abnormally long CPU time which is not consistent with throttling.

There are two other things to check that I can think of - and hopefully more that others might come up with.

Firstly, is the celeron running in a hot environment or does it have a heat sink that is clogged with fluff. Is it possible that the machine is throttling itself due to thermal conditions?

Secondly, does the machine have a keyboard attached? I'm currently running a very large number of machines, most of which are running Linux and a few run Windows. Virtually all machines are headless and have no keyboard and mouse attached. Those that run Windows are afflicted by an issue that appears after a month or two of continuous running where the hard disk shows continuous activity and network activity becomes just about impossible. The CPU is still crunching but slows to a crawl. The fix is to plug in a keyboard and instantly things return to normal. The linux machines don't seem to suffer from this problem.

One final point. I notice over at Seti about 2.5 years ago you wrote a message about having to investigate why Seti crunching had slowed to a crawl. Did you ever work out why that was happening? Is your current Seti crunching performing properly or is it vastly slowed down as well? How long is it taking for your celeron to crunch a Seti task?

Cheers,
Gary.

Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5893653
RAC: 0

RE: I asked previously and

Message 93589 in response to message 93588

Quote:
I asked previously and didn't get an answer - is the increment in CPU time (elapsed time) for a crunching task as shown in BOINC Manager pretty much in accord with the increment in wall clock time?


The Elapsed Time column in the Tasks tab, is wall-clock time. To see CPU Time (the time that only the CPU actually works on anything), you should click on the running task, click Properties on the left menu and scroll down for the frozen-at-the-moment-you-clicked-Properties moment of CPU Time.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5883
Credit: 119065788000
RAC: 24450697

RE: The Elapsed Time column

Message 93590 in response to message 93589

Quote:
The Elapsed Time column in the Tasks tab, is wall-clock time. To see CPU Time (the time that only the CPU actually works on anything), you should click on the running task, click Properties on the left menu and scroll down for the frozen-at-the-moment-you-clicked-Properties moment of CPU Time.


OK, this is obviously a "feature" of recent versions of BOINC. The most recent version I use is 6.2.15 where it is still CPU time. I seem to remember Richard Haslegrove pointing out to me at one stage that CPU efficiency had gone so I guess removing CPU time from the main display was also part of the change made.

This is useful to know, thank you, since it now removes the "difficulty" I was having with my thinking that the slow crunching wasn't due to CPU throttling. If the OP checks his task properties as suggested above and finds a very much shorter CPU time than is shown for elapsed time on the main tasks tab, then the poor performance most likely is due to CPU throttling. Funnily enough, the 2.5 year old thread over at Seti, that the OP contributed to, was to do with CPU throttling as a way of reducing the impact of BOINC on everyday computing tasks so maybe the OP introduced severe throttling at that point into his preferences and he's been saddled with poor performance ever since.

Cheers,
Gary.

tullio
tullio
Joined: 22 Jan 05
Posts: 2118
Credit: 61407735
RAC: 0

On the properties tab I

On the properties tab I find:
CPU time 0:59:40
Elapsed time 1:01:46
This for a hierarchical search unit now running. I am using BOINC 6.6.31 (which declares itself to be 6.6.29) after switching from 5.10.45 on my Linux box. I tried briefly 6.6.36 but it did not work on my SuSE 10.3, so I switched back, having conserved the installation file .sh.
Tullio
Edit. One unpleasant result from my BOINC switch is substantially lower benchmark value for floating point ops, from ~1300 to ~1100, Integer ops remained constant. This on an Opteron 1210 at 1.8 GHz.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.