Credits are unreasonable

jamesdavis
jamesdavis
Joined: 9 Feb 05
Posts: 14
Credit: 822,778
RAC: 0
Topic 193500

My system, a quad cpu, running Windows 2003 server 64 bit OS, with 5 gig of ram, is taking anywhere from 86,000 to 185,000 cpu seconds to complete these work units and yet EVERY result has been awarded from 236.47 credits to 236.71 credits. I towuld seem that I should be aborting long process work units or doing work for other BOINC projects.

Why?

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 2,140
Credit: 2,767,809,187
RAC: 988,976

Credits are unreasonable

Quote:

My system, a quad cpu, running Windows 2003 server 64 bit OS, with 5 gig of ram, is taking anywhere from 86,000 to 185,000 cpu seconds to complete these work units and yet EVERY result has been awarded from 236.47 credits to 236.71 credits. I towuld seem that I should be aborting long process work units or doing work for other BOINC projects.

Why?


Hello James.

With two posts, I would normally say 'welcome to the message boards'. But since your other post was 1087 days ago, perhaps I'd better say 'welcome back to the message boards'.

There is a regular, cyclical, and by now well understood variation in crunching times from WU to WU - see this thread for details. But that only accounts for a variation of +/- 10% or so. You say you are seeing a variation of more than 100%.

Your computers are hidden, so we have no easy way of analysing what has gone wrong. If you could unhide the computers, or provide a link to one of the long run-time tasks, we could look at it with you, and try and provide a diagnosis and hopefully some help. My suspicion is that for some reason your computer is failing to "checkpoint" (store a record of the work done so far), and is starting again from the beginning of the WU - thus wasting all the work previously done. Without further data, it is difficult to be more precise than that - but there is certainly something wrong with the way the tasks are running on your machine: the work as sent out by the Einstein servers usually shows nothing like the variation you've reported.

jamesdavis
jamesdavis
Joined: 9 Feb 05
Posts: 14
Credit: 822,778
RAC: 0

I will gladly show you

Message 78559 in response to message 78558

I will gladly show you whatever you wish to see to try to resolve this issue. Here is a link from Einstein that shows the 186,000 second run time computer information: http://einsteinathome.org/host/1096521
and here is a link to the einstein page which shows the work unit and the two computers (one of them mine) that established the credit - the other computer took only 52,000 seconds to complete: http://einstein.phys.uwm.edu/workunit.php?
wuid=36807983

Is that enough information for you or do I really need to unmask my computers?

jamesdavis
jamesdavis
Joined: 9 Feb 05
Posts: 14
Credit: 822,778
RAC: 0

RE: I will gladly show you

Message 78560 in response to message 78559

Quote:

I will gladly show you whatever you wish to see to try to resolve this issue. Here is a link from Einstein that shows the 186,000 second run time computer information: http://einsteinathome.org/host/1096521
and here is a link to the einstein page which shows the work unit and the two computers (one of them mine) that established the credit - the other computer took only 52,000 seconds to complete: http://einstein.phys.uwm.edu/workunit.php?
wuid=36807983

Is that enough information for you or do I really need to unmask my computers?

Well it's Windows 2008 instead of 2003, but it is the 64 bit version.

Now I was not of the opinion that one or more of my slower computers should be ignored - it's just that this particular computer is my most powerful one with the most resources yet it took the longest time to complete the work unit. And, of course, reviewing all my completed work units (since I returned to this project) I note that they ALL received essentially the same number of credits (within .5 points), regardless of how long they took.

Here is the einstein page for my dual processor windows 2000 server with 1 gig of ram. It completed a work unit in about 85,000 cpu seconds and received the same (essentially) credit for doing so as my 'big' system: http://einsteinathome.org/host/1096519

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282,700
RAC: 0

RE: I will gladly show you

Message 78561 in response to message 78559

Quote:

I will gladly show you whatever you wish to see to try to resolve this issue. Here is a link from Einstein that shows the 186,000 second run time computer information: http://einsteinathome.org/host/1096521
and here is a link to the einstein page which shows the work unit and the two computers (one of them mine) that established the credit - the other computer took only 52,000 seconds to complete: http://einstein.phys.uwm.edu/workunit.php?
wuid=36807983

Is that enough information for you or do I really need to unmask my computers?

I dunno about others, but I would say that the times for a "Core2 Quad Q6600" should be significantly less, even with the contention that has been found. Are you sure you don't have something like Speedstep enabled? If you right-click on My Computer and then select Properties, what is the listed clock speed on the General tab?

jamesdavis
jamesdavis
Joined: 9 Feb 05
Posts: 14
Credit: 822,778
RAC: 0

RE: I dunno about others,

Message 78562 in response to message 78561

Quote:

I dunno about others, but I would say that the times for a "Core2 Quad Q6600" should be significantly less, even with the contention that has been found. Are you sure you don't have something like Speedstep enabled? If you right-click on My Computer and then select Properties, what is the listed clock speed on the General tab?

Each processor is running at 2.40 Ghrz

Jord
Joined: 26 Jan 05
Posts: 2,952
Credit: 5,779,100
RAC: 0

RE: Well it's Windows

Message 78563 in response to message 78560

Quote:
Well it's Windows 2008


Which could be a problem in itself, as BOINC 5.10 might not be completely compatible with this OS.

What kind of messages do you get when running Einstein tasks? Any indications of errors?

archae86
archae86
Joined: 6 Dec 05
Posts: 3,145
Credit: 7,052,154,931
RAC: 1,642,357

RE: I would say that the

Message 78564 in response to message 78561

Quote:
I would say that the times for a "Core2 Quad Q6600" should be significantly less, even with the contention that has been found.


I run a Core2 Quad Q6600, and even accounting for the differences due to:

1. ap version 4.15 vs. 4.26
2. my overclock to 3.006 GHz vs. your ??? but 2.4 GHz if stock
3. expected variation by sequence number over the cycle.
4. less well understood but definite impact from other tasks running at the same time.

Still, your times are far higher than I'd expect. Mine varied over the cycle from a low of about 27,000 to a high of about 34,000 CPU seconds.

It appears that the size of the "extra bad" effect on your run times also varied among the four results currently on display, as the variation in time with sequence number is not as expected.

I also notice that the benchmark reported measured floating point speed for this host as:
1143.13 million ops/sec

vs. my Q6600's current value of:
2882.76 million ops/sec

If you are running stock CPU clock of 2.4 GHz, then I'd scale my result to suggest you would expect to report about double what you do for Floating point speed. That is just a benchmark, and not the ratio of impairment to be expected on the actual ap, but still a strong hint that something is wrong.

Something is quite sick on that machine. The benchmark result points firmly away from the Einstein ap itself.

jamesdavis
jamesdavis
Joined: 9 Feb 05
Posts: 14
Credit: 822,778
RAC: 0

RE: RE: Well it's Windows

Message 78565 in response to message 78563

Quote:
Quote:
Well it's Windows 2008

Which could be a problem in itself, as BOINC 5.10 might not be completely compatible with this OS.

What kind of messages do you get when running Einstein tasks? Any indications of errors?

There are no error messages of any kind. I run SETI, LHC, QMC, and Einstein projects 24/7. All projects successfully complete work units.

I will not overclock my CPU's - they run at rated speed.

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,364
Credit: 3,562,358,667
RAC: 241

Do you get the expected

Do you get the expected credit rates on your other projects with your QC system? If so, I suspect that something in the current E@H app is effectively incompatable with win2k8. I haven't seen mention of anyone else trying to use this OS version so you might be the first to discover an issue.

archae86
archae86
Joined: 6 Dec 05
Posts: 3,145
Credit: 7,052,154,931
RAC: 1,642,357

RE: Is that enough

Message 78567 in response to message 78559

Quote:
Is that enough information for you or do I really need to unmask my computers?


If you unmasked, I could probably take a look at your Q6600's SETI result completion time for a few representative AR's and say whether there is a major discrepancy there as well.

If you think you are protecting some more important secret by masking, then we probably should not be having this interaction.

Comment viewing options

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