My most recent SETI@Home task, work unit 140472475, seems to be behaving a little strangely. Since it started, the "time to completion" has gone up by roughly one second for every second of CPU used. It now sits at 29:51:02 CPU time used, 30:02:01 time to completion. It has been stuck at 24.668% complete for some hours now.

I also noticed this task ran continuously for at least 6 hours (maybe longer, I don't check my BOINC statistics regularly). My machine normally switches between SETI and Climateprediction at roughly one hour intervals. I rebooted my machine, and it came back with Climateprediction running.

This isn't very scientific, but based on viewing the graphics for this task, it appeared to be proceeding painfully sloooooowly.

Should I just kill this task, or can it be salvaged?

---

My most recent SETI@Home task, work unit 140472475, seems to be behaving a little strangely. Since it started, the "time to completion" has gone up by roughly one second for every second of CPU used. It now sits at 29:51:02 CPU time used, 30:02:01 time to completion. It has been stuck at 24.668% complete for some hours now.

I also noticed this task ran continuously for at least 6 hours (maybe longer, I don't check my BOINC statistics regularly). My machine normally switches between SETI and Climateprediction at roughly one hour intervals. I rebooted my machine, and it came back with Climateprediction running.

This isn't very scientific, but based on viewing the graphics for this task, it appeared to be proceeding painfully sloooooowly.

Should I just kill this task, or can it be salvaged?

Try restarting BOINC (reboot if necessary). If it doesn't make any progress after trying that, then abort it.

You say it's still progressing, so it's not stuck, but something seems amiss. perhaps something is stopping boinc from running. Something like a preference setting (do work while in use NO, comes to mind). Have you tried a restart of the puter to free resources? Have you tried suspending and resuming the task? If else fails, just abort it and move on.

You say it's still progressing, so it's not stuck, but something seems amiss. perhaps something is stopping boinc from running. Something like a preference setting (do work while in use NO, comes to mind). Have you tried a restart of the puter to free resources? Have you tried suspending and resuming the task? If else fails, just abort it and move on.

Thanks all for the suggestions. I did reboot this morning, and Climatepredication has been running for about 3 hours straight now. Presumably this is to make up for the time SETI was "stuck". (both SETI and Climateprediction have equal resource shares on my machine). I'll let you know what happens when SETI restarts.

This isn't very scientific, but based on viewing the graphics for this task, it appeared to be proceeding painfully sloooooowly.

Should I just kill this task, or can it be salvaged?

Bill:

I noticed that you are running the standard Seti application. You may want to download and install one of Simon's Seti apps from the Chicken Coop. It should double your speed.

Franz

...
It has been stuck at 24.668% complete for some hours now.
...

It was truly stuck and making no progress. The application recalculates its progress very frequently when running normally, even on a very slow system the fractional part should show changes each time BOINC Manager updates its display.

If ever I saw unchanging progress for 30 seconds I'd take action to kill the application and restart it from the last checkpoint. The advice to reboot of course accomplishes that for any combination of preferences and any platform.

                                                                Joe

[quoteIt was truly stuck and making no progress. The application recalculates its progress very frequently when running normally, even on a very slow system the fractional part should show changes each time BOINC Manager updates its display.

If ever I saw unchanging progress for 30 seconds I'd take action to kill the application and restart it from the last checkpoint. The advice to reboot of course accomplishes that for any combination of preferences and any platform.

                                                                Joe

[/quote]

I finally suspended Climateprediction, just to see what would happen. SETI restarted, ran normally for a few seconds (time to complete going down as CPU time went up), but then exited with an error:

7/15/2007 2:10:07 PM|SETI@home|Reason: Unrecoverable error for result 07my00aa.14716.3026.461070.3.229_2 ( - exit code -1073741819 (0xc0000005))

So, was the file I received corrupt in some way, or is there anything I need to do to keep SETI running on my machine?

[quoteIt was truly stuck and making no progress. The application recalculates its progress very frequently when running normally, even on a very slow system the fractional part should show changes each time BOINC Manager updates its display.

If ever I saw unchanging progress for 30 seconds I'd take action to kill the application and restart it from the last checkpoint. The advice to reboot of course accomplishes that for any combination of preferences and any platform.

                                                                Joe

I finally suspended Climateprediction, just to see what would happen. SETI restarted, ran normally for a few seconds (time to complete going down as CPU time went up), but then exited with an error:

7/15/2007 2:10:07 PM|SETI@home|Reason: Unrecoverable error for result 07my00aa.14716.3026.461070.3.229_2 ( - exit code -1073741819 (0xc0000005))

So, was the file I received corrupt in some way, or is there anything I need to do to keep SETI running on my machine?
[/quote]
Don't know whether it applies in this case, but I think I remember seeing somewhere that this is related to Windows handling of the -9 overflow error. Many times (if I am correct) Windows hangs until you restart and then gives this error when you get a noisy work unit instead of exiting cleanly and giving the -9 error.

---

Jim

Some people plan their life out and look back at the wealth they've had.
Others live life day by day and look back at the wealth of experiences and enjoyment they've had.

Yes, Jim, I looks like one of those 1 in 10,000 wus bugs where it goofs up and then on restart goes to error. Not much to do, but move on. Thankfully they aren't very frequent.

---

What do we do when all cached units are showing excessive times to complete?

Starting about 3 days ago, the times to complete for the WUs in my cache went from 2-6 hours to 47-160 hours. The actual time to complete has gone up but not that much (2-4 hours went to 2.5-5.0 hours). The main problem with this situation is that my cache will drain down to 1-2 WUs (only one WU if it has an ETC of 160 hours).

This is ocurring on a Pentium M 1.8GHz with WinXP that gets re-booted daily.

TIA
QS

---

What do we do when all cached units are showing excessive times to complete?

Starting about 3 days ago, the times to complete for the WUs in my cache went from 2-6 hours to 47-160 hours. The actual time to complete has gone up but not that much (2-4 hours went to 2.5-5.0 hours). The main problem with this situation is that my cache will drain down to 1-2 WUs (only one WU if it has an ETC of 160 hours).

This is ocurring on a Pentium M 1.8GHz with WinXP that gets re-booted daily.

TIA
QS

Stop BOINC and - with care and a text editor - find this line in your 'client_state.xml' file:

<project>
    <master_url>http://setiathome.berkeley.edu/</master_url>
...
    [b]<duration_correction_factor>[color=red]0.483604[/color]</duration_correction_factor>[/b]

You'll probably find the figure in red is in the tens, or even hundreds.

If so, change it. The nominal value is 1.000000 (six decimal places): it will often be lower, particularly if you run an optimised application. If you do, somewhere round 0.400000 or even 0.300000 will be appropriate.

Re-start BOINC. You may over- or under-fetch the first time, but the <duration_correction_factor> is self-correcting over time.

You'll probably find the figure in red is in the tens, or even hundreds.

If so, change it. The nominal value is 1.000000 (six decimal places): it will often be lower, particularly if you run an optimised application. If you do, somewhere round 0.400000 or even 0.300000 will be appropriate.

Re-start BOINC. You may over- or under-fetch the first time, but the <duration_correction_factor> is self-correcting over time.

My DCF seems about right for using Chicken Soup 2.2B:

    <duration_correction_factor>0.251810</duration_correction_factor>

QS

---

Then I'm stumped. Very odd.

QSilver, what are your current benchmarks, and what did they used to be.

[edit]Wholey Crimini batman, look at these benchmarks from the website::::

Measured floating point speed 47.48 million ops/sec
Measured integer speed 92.71 million ops/sec


somethings gone whacky, Try manually rerunning benchmarks. your benchmarks should be roughly 1550/3200

First, check temps/cooling
then check ram

[second edit] the only time I've seen lower benchmarks is with my Pentium 60, heck my Celeron 500 is 409/673

---

Just ran benchmarks, and they came in at 702/1849.

To be honest, I've never really paid attention to the results when benchmarks run on their own. However, my actual times to complete (while a little longer lately than historical) are definitely in what I consider to be their typical range. The Pentium-M's RAC is also at a typical value (i.e., has not taken a hit over the last 3-4 days since these long ETCs started showing up).

QS

---

Still might wanna check temps. Here's some benchmarks from other Intel Pentium M 1.8Ghz machines:

1564.84 3204.84
1607.44 3298.38
1510.68 3160.84
1533.10 3139.66
1551.44 3217.64
1571.77 3238.79
1586.59 3242.01
1586.59 3242.01
1579.53 3292.57

Traditionally, when a thermal diode (or coolnquiet, etc) limits the freq to 1/2 it's normal value, you end up with benchmarks that are nearly 1/2 that of others.

tony

---

Just ran benchmarks, and they came in at 702/1849.

To be honest, I've never really paid attention to the results when benchmarks run on their own. However, my actual times to complete (while a little longer lately than historical) are definitely in what I consider to be their typical range. The Pentium-M's RAC is also at a typical value (i.e., has not taken a hit over the last 3-4 days since these long ETCs started showing up).

QS

Yes, with the new apps and clients using FPOPS for establishing credit you won't notice any variation regardless of the benchmarks unless you happen to get paired with someone using an old client with low benchmarks! Now the benchmarks are only used to estimate the time to completion and thus the amount of work you can download at a time. It doesn't affect your RAC and credit.

---

Jim

Some people plan their life out and look back at the wealth they've had.
Others live life day by day and look back at the wealth of experiences and enjoyment they've had.

just noticed it was a pentium M 1.86. I have no samples for that processor.

Checking temps...

I installed SpeedFan 4.32. In the first 15 minutes, it showed the CPU temp stepping slowing up to 91C from 68, stepping slowly down to 69. It now heading back up again (currently at 84).

Again, performance doesn't appear to be impacted; it's only the ETCs that show excessive times.

QS

[edited to add missing word]

---

My AMD's go into alarm at like 69C and Failure at 124C. I'm not sure what Pent M 1.86's are, or where to look. Anyone got a link?