Unusually short crunch time

Donald A. Tevault
Donald A. Tevault
Joined: 17 Feb 06
Posts: 439
Credit: 73516529
RAC: 0
Topic 193641

Here's something that's strange. One of my old P-III machines got a workunit that only took about half as long as it should have to crunch. My wingman, running an Athlon64 3500+, took only about 5000 seconds less than I did with the P-III. Yet, the results validated okay.

Something just doesn't seem right with this.

Strange P-III results

Bernd Machenschalk
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4265
Credit: 244922893
RAC: 16808

Unusually short crunch time

If possible, dig through the client logs and see when this task was started and finished ("wall clock time"). Correctly counting the CPU time isn't that trivial on Linux, and problems on that are common.

BM

BM

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2139
Credit: 2752659155
RAC: 1473243

Bernd, Have a look at the

Bernd,

Have a look at the stderr for Donalds anomalous result.

He did the first 550 skypoints at one per app checkpoint, then the second 654 skypoints all within a single checkpoint interval.

His wingmate restarted at exactly the same skypoint 550. Coincidence?

peanut
peanut
Joined: 4 May 07
Posts: 162
Credit: 9644812
RAC: 0

I just looked at AKOSF's

I just looked at AKOSF's machine in the top 20 list and I saw he had some barely over 1000 times that did say valid.

Either he has a great new app in development or he has some short times as well.

Donald A. Tevault
Donald A. Tevault
Joined: 17 Feb 06
Posts: 439
Credit: 73516529
RAC: 0

RE: If possible, dig

Message 80827 in response to message 80824

Quote:

If possible, dig through the client logs and see when this task was started and finished ("wall clock time"). Correctly counting the CPU time isn't that trivial on Linux, and problems on that are common.

BM

Okay, I can try that. In the meantime, here's another one that looks abnormal. This time, my wingman finished first, and has an unusually short runtime for his 3500+ machine. (Mine in this case is the Xeon, which hasn't started crunching that one yet.)

A second strange one.

Donald A. Tevault
Donald A. Tevault
Joined: 17 Feb 06
Posts: 439
Credit: 73516529
RAC: 0

RE: If possible, dig

Message 80828 in response to message 80824

Quote:

If possible, dig through the client logs and see when this task was started and finished ("wall clock time"). Correctly counting the CPU time isn't that trivial on Linux, and problems on that are common.

BM

Okay, I found the information.

According to the log, this task started at 08:55 on 23 April, and finished computation at 00:44 on 24 April. That's a bit over 15 hours, which is about half what it should take on this P-III 667.

Donald A. Tevault
Donald A. Tevault
Joined: 17 Feb 06
Posts: 439
Credit: 73516529
RAC: 0

RE: I just looked at

Message 80829 in response to message 80826

Quote:

I just looked at AKOSF's machine in the top 20 list and I saw he had some barely over 1000 times that did say valid.

Either he has a great new app in development or he has some short times as well.

Yeah, I just saw that. It looks like he's kicked in the afterburners on that machine.

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

RE: RE: I just looked at

Message 80830 in response to message 80829

Quote:
Quote:

I just looked at AKOSF's machine in the top 20 list and I saw he had some barely over 1000 times that did say valid.

Either he has a great new app in development or he has some short times as well.

Yeah, I just saw that. It looks like he's kicked in the afterburners on that machine.

Not exactly. I tried to set up a PC with a Wolfdale CPU, but it doesn't want to run Einstein@Home...

Akos Fekete
Akos Fekete
Joined: 13 Nov 05
Posts: 561
Credit: 4527270
RAC: 0

I found some '1155.00 sec'

I found some '1155.00 sec' time in the results of my Kentsfield.
But the BOINC Manager showed about 2,5-3 hours for them.
I didn't find the reason of this behaviour yet.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 686042851
RAC: 592401

RE: RE: If possible, dig

Message 80832 in response to message 80827

Quote:
Quote:

If possible, dig through the client logs and see when this task was started and finished ("wall clock time"). Correctly counting the CPU time isn't that trivial on Linux, and problems on that are common.

BM

Okay, I can try that. In the meantime, here's another one that looks abnormal. This time, my wingman finished first, and has an unusually short runtime for his 3500+ machine. (Mine in this case is the Xeon, which hasn't started crunching that one yet.)

A second strange one.

I think this "second strange one" can be neglected, look at the other results from that host: most claim 0 seconds crunch time, and it's also using a very old BOINC client.

But that first result looks strange indeed.

CU
Bikeman

Donald A. Tevault
Donald A. Tevault
Joined: 17 Feb 06
Posts: 439
Credit: 73516529
RAC: 0

RE: I found some '1155.00

Message 80833 in response to message 80831

Quote:
I found some '1155.00 sec' time in the results of my Kentsfield.
But the BOINC Manager showed about 2,5-3 hours for them.
I didn't find the reason of this behaviour yet.

Darn, you had us hoping that you had found a miracle. ;)

Comment viewing options

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