WU will run for ~40 hours?

Piet
Piet
Joined: 13 Aug 10
Posts: 11
Credit: 93591
RAC: 0
Topic 195565

Today I got a new Binary Radio Pulsar Search 1.04 WU that will run for ~40 hours. Is that a joke or what?

See http://einsteinathome.org/task/214750205

Gundolf Jahn
Gundolf Jahn
Joined: 1 Mar 05
Posts: 1079
Credit: 341280
RAC: 0

WU will run for ~40 hours?

Quote:

Today I got a new Binary Radio Pulsar Search 1.04 WU that will run for ~40 hours. Is that a joke or what?

See http://einsteinathome.org/task/214750205


I made the link active.

That tasks seems to be your first BRPS one (in a longer time). So, the Duration Correction Factor might be out of whack for that application. After all, the times given in BOINC manager are only estimates.

Gruß,
Gundolf

Computer sind nicht alles im Leben. (Kleiner Scherz)

Piet
Piet
Joined: 13 Aug 10
Posts: 11
Credit: 93591
RAC: 0

The Boing Manager gave 58

The Boing Manager gave 58 hours, I let the WU run for some time to calculate the time by myself, result is ~40 hours :-(

Piet

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 756102829
RAC: 1157509

RE: The Boing Manager gave

Quote:

The Boing Manager gave 58 hours, I let the WU run for some time to calculate the time by myself, result is ~40 hours :-(

Piet

Hmm...your other tasks seem to underperform as well in comparison to similar hardware, see this one , for example:

http://einsteinathome.org/workunit/90850716

This is a notebook we are talking about? I suspect that it is not processing the tasks as fast as it could. Possible reasons for this include:

* Windows energy profile settings command the PC not to run at the max possible clockrate

* The cooling is poor and the PC throttles back automatically to prevent overheating.

I would suggest to use a tool like CPU-Z to monitor the CPU clock rate while the E@H tasks are executing. In theory, the runtime for a BRP3 unit on your kind of hardware should be more like 24 hrs, 30 hrs at worst.

CU
HB

Piet
Piet
Joined: 13 Aug 10
Posts: 11
Credit: 93591
RAC: 0

Yeah, a notebook. As

Yeah, a notebook. As everyone: I didn't change anything! :-)

But I will look for the reason of underperformance.

Thanks Piet

Piet
Piet
Joined: 13 Aug 10
Posts: 11
Credit: 93591
RAC: 0

RE: Hmm...your other tasks

Quote:


Hmm...your other tasks seem to underperform as well in comparison to similar hardware, see this one , for example:

http://einsteinathome.org/workunit/90850716

This is a notebook we are talking about? I suspect that it is not processing the tasks as fast as it could. Possible reasons for this include:

* Windows energy profile settings command the PC not to run at the max possible clockrate


The energy profile is set to full power if requested.

Quote:


* The cooling is poor and the PC throttles back automatically to prevent overheating.

I would suggest to use a tool like CPU-Z to monitor the CPU clock rate while the E@H tasks are executing. In theory, the runtime for a BRP3 unit on your kind of hardware should be more like 24 hrs, 30 hrs at worst.

CU
HB


This little box IS getting hot, but the fan is not getting crazy, just hearable.
CPU-Z shows a core speed fluctuating between 990-1815MHz running 2 Einstein WUs. Interesting is Einstein WUs show more often 990MHz then lets say Spinhenge or Poem. Any more ideas?

Piet

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1364
Credit: 3562358667
RAC: 0

RE: This little box IS

Quote:


This little box IS getting hot, but the fan is not getting crazy, just hearable.
CPU-Z shows a core speed fluctuating between 990-1815MHz running 2 Einstein WUs. Interesting is Einstein WUs show more often 990MHz then lets say Spinhenge or Poem. Any more ideas?

Piet

Thermal throttling. I'm not sure how tweaked the BRP WU's are at the moment, but the GC1 WU's are highly optimized to run the CPU at maximum performance, if Spinhenge and Poem aren't as heavily tweaked they wouldn't put as much of a load on the CPU causing it to run cooler.

Probably your heatsink is clogged up with dust. Turn your computer off and remove the battery. Then unscrew the access plate over the fan, hold the fan in place with a finger tip, and blow it out with a can of air. You can try blowing without opening it up, but generally will only dislodge a fraction of the dust. If you do this stick something into the fan holes to stop it from spinning, I've heard of some fans failing after being spun too fast when blown out.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 756102829
RAC: 1157509

RE: Then unscrew the access

Quote:
Then unscrew the access plate over the fan,

IF you have an access plate....I can remember my last attempt to get to the heatsink of my previous Samsung notebook...I was really relieved when I finally managed to get it all back together again :-(. No fun.

CU
HBE

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1364
Credit: 3562358667
RAC: 0

RE: RE: Then unscrew the

Quote:
Quote:
Then unscrew the access plate over the fan,

IF you have an access plate....I can remember my last attempt to get to the heatsink of my previous Samsung notebook...I was really relieved when I finally managed to get it all back together again :-(. No fun.

CU
HBE

There is that, which is why I mentioned blowing through the fan holes as an alternative. The worst I've ever had to deal with was a 5(?) year old Toshiba that interlocked the access plates so I had to remove 4 others before I could release the one over the fan.

BarryAZ
BarryAZ
Joined: 8 May 05
Posts: 190
Credit: 325566843
RAC: 16999

Notwithstanding the

Notwithstanding the supposition that the very long run work units are user workstation configuration specific, I too have see some REALLY long run work units of late -- a number in excess of 25 hours -- on a variety of hardware.

I suspect that at least part of this is in fact project specific.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 756102829
RAC: 1157509

Hi, The (relatively) new

Hi,

The (relatively) new BRP3 tasks which are now crunching Parkes radio telescope data can indeed take around 25 hrs. Needless to say the credits per task are also higher than for the other tasks (S5GC1HF LIGO data crunching jobs).

Happy crunching
HB

Comment viewing options

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