Average CPU efficiency?

ADDMP
ADDMP
Joined: 25 Feb 05
Posts: 104
Credit: 7,332,049
RAC: 0
Topic 193534

I have roughly comparable computers running both WIN & LINUX, & I am trying to compare these for E@H. I notice the BOINC statistic "average CPU efficiency" is about 75% under Linux & about 100% under WIN. BOINC is running close to 100% of the time in both cases. Should this statistic be taken seriously? Does the Linus OS impose some sort of overhead on the E@H calculation that WIN does not?

The reported cpu time to complete a unit looks roughly the same for the two OSes, but the WIN computer does more units/day anyway.

archae86
archae86
Joined: 6 Dec 05
Posts: 2,556
Credit: 1,852,720,426
RAC: 2,574,428

Average CPU efficiency?

Quote:

I have roughly comparable computers running both WIN & LINUX, & I am trying to compare these for E@H. I notice the BOINC statistic "average CPU efficiency" is about 75% under Linux & about 100% under WIN. BOINC is running close to 100% of the time in both cases. Should this statistic be taken seriously? Does the Linus OS impose some sort of overhead on the E@H calculation that WIN does not?

The reported cpu time to complete a unit looks roughly the same for the two OSes, but the WIN computer does more units/day anyway.


Taken at face value, this number reports that while you have BOINC turned on and the science aps authorized to run, that "something else" is using 25% of the CPU time on your Linux machine.

That sounds possibly consistent with your report that while reported CPU times are similar, the Windows box is completing more work.

As I have only Windows hosts, I don't know what is a typical reported cpu efficiency for a Linus host here. The Windows hosts which don't get substantial non-BOINC use are over 99%, while my main audio and browsing machine is just over 97% and the laptop on which my wife surfs the internet for lace deals and plays an inefficient Solitaire game is at 89%.

So 75% is really low--you might want to check what besides BOINC-related things is using time on that box.

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3,515
Credit: 408,946,101
RAC: 32,872

Same here. I've several Linux

Same here. I've several Linux hosts with a variety of Boinc versions and Linux distributions, and for those hosts that are exclusively crunching for E@H, the efficiency value is about 99%. For Hosts that do other stuff like Web-browsing or database operations, the value drops to around 90 %. So it's not that the efficiency value is generally lower on Linux.

You may want to use the top command to check what else is consuming CPU cycles on your CPU.

Are the two hosts you compared (Win and Linux) in the same "location" category ("home", "work", "school" ..) as visible in the web-interface? If not you might want to check location-specific settings, e.g. CPU throttling.

CU
Bikeman

ADDMP
ADDMP
Joined: 25 Feb 05
Posts: 104
Credit: 7,332,049
RAC: 0

Top was a good suggestion.

Message 79752 in response to message 79751

Top was a good suggestion. There is a "user" named "beaglein" that is running a program named "beagle-build-in" that is taking 25% of the cpu power.

These are home computers, & there is no one other than me that has any legitimate use of them.

"Beagle" is a search engine & "kerry" is a desktop interface for "beagle". These are programs I probably started up sometime to search without knowing anything about them. I had no idea that "beagle" has been running in the background eating up 25% of the cpu cycles for the past month. I have now uninstalled both of them from my computer.

Thanks.

ADDMP

Quote:

Same here. I've several Linux hosts with a variety of Boinc versions and Linux distributions, and for those hosts that are exclusively crunching for E@H, the efficiency value is about 99%. For Hosts that do other stuff like Web-browsing or database operations, the value drops to around 90 %. So it's not that the efficiency value is generally lower on Linux.

You may want to use the top command to check what else is consuming CPU cycles on your CPU.

Are the two hosts you compared (Win and Linux) in the same "location" category ("home", "work", "school" ..) as visible in the web-interface? If not you might want to check location-specific settings, e.g. CPU throttling.

CU
Bikeman


jowr
jowr
Joined: 19 Feb 05
Posts: 55
Credit: 1,947,636
RAC: 0

Hey bikeman - I'm tracking a

Hey bikeman - I'm tracking a rather inane issue with BOINC on linux. After awhile, the boinc process will suddenly come online and dominate my entire cluster of systems except for two, which confuses me. I figure I'll have to spend a half hour deleting everything and rebuilding, but I'm wondering what could cause the boinc process to run and dominate the system over the einstein process for days at a time.

I've never had this happen before, so it's weird.

Conan
Conan
Joined: 19 Jun 05
Posts: 161
Credit: 5,808,481
RAC: 1,029

RE: Top was a good

Message 79754 in response to message 79752

Quote:

Top was a good suggestion. There is a "user" named "beaglein" that is running a program named "beagle-build-in" that is taking 25% of the cpu power.

These are home computers, & there is no one other than me that has any legitimate use of them.

"Beagle" is a search engine & "kerry" is a desktop interface for "beagle". These are programs I probably started up sometime to search without knowing anything about them. I had no idea that "beagle" has been running in the background eating up 25% of the cpu cycles for the past month. I have now uninstalled both of them from my computer.

Thanks.

ADDMP

Quote:

Same here. I've several Linux hosts with a variety of Boinc versions and Linux distributions, and for those hosts that are exclusively crunching for E@H, the efficiency value is about 99%. For Hosts that do other stuff like Web-browsing or database operations, the value drops to around 90 %. So it's not that the efficiency value is generally lower on Linux.

You may want to use the top command to check what else is consuming CPU cycles on your CPU.

Are the two hosts you compared (Win and Linux) in the same "location" category ("home", "work", "school" ..) as visible in the web-interface? If not you might want to check location-specific settings, e.g. CPU throttling.

CU
Bikeman


I too have been 'dogged' by beagle when I installed Fedora Core 6 on a new (at the time) computer. I did not load it but it got loaded anyway.
It took up far more than 25% cpu time and was accumulating data (it traces what your computer is doing), at such a rate that in 2 weeks it almost filled a 250 GB hard drive, and I didn't know it either.

Killed it and got rid of it, deleted all the stuff it stored and the computer has been fine since.

ADDMP
ADDMP
Joined: 25 Feb 05
Posts: 104
Credit: 7,332,049
RAC: 0

I vote for classifying beagle

I vote for classifying beagle as a Linux worm & treating it accordingly.

ADDMP

Comment viewing options

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