Linux - does changing 'nice' make a big diff??

Cochise
Cochise
Joined: 11 Feb 05
Posts: 38
Credit: 3717
RAC: 0
Topic 188102

Has anyone seen if changing the nice level on linux makes a big diff in computational time? Right now the default is 19 (I think) and I was wondering if changing it to say 5 or so would make a big diff. I would just change it and see myself but the box is in a closet just running with out a monitor or keyboard or anything, so it's kind of a hassel so I thought I'd ask first if it's worth it.

Thanks!

Seti-Cruncher
Seti-Cruncher
Joined: 9 Feb 05
Posts: 70
Credit: 7114
RAC: 0

Linux - does changing 'nice' make a big diff??

> Has anyone seen if changing the nice level on linux makes a big diff in
> computational time? Right now the default is 19 (I think) and I was wondering
> if changing it to say 5 or so would make a big diff. I would just change it
> and see myself but the box is in a closet just running with out a monitor or
> keyboard or anything, so it's kind of a hassel so I thought I'd ask first if
> it's worth it.

If you don't have anything else taking significant CPU time in the machine, no, it will make no difference whatsoever. If you are using the machine interactively, making BOINC less nice will make your machine less responsive. At the extreme, it could virtually bring it to a standstill on all other tasks.

Be lucky,

Neil

Cochise
Cochise
Joined: 11 Feb 05
Posts: 38
Credit: 3717
RAC: 0

well i guess in this case it

well i guess in this case it would make zero diff then because that's all the machine is doing, crunching einstein

Thanks!

Chiana
Chiana
Joined: 20 Feb 05
Posts: 7
Credit: 177270
RAC: 0

Actually it seems so, I

Actually it seems so, I lowered it to 1 on one of my comps in order to beat the deadline of a wu, and it worked, but I had to forcingly (pressing RESET button) to gain control of it again...

MortenLan
MortenLan
Joined: 1 Mar 05
Posts: 1
Credit: 95
RAC: 0

> Actually it seems so, I

Message 6035 in response to message 6034

> Actually it seems so, I lowered it to 1 on one of my comps in order to beat
> the deadline of a wu, and it worked, but I had to forcingly (pressing RESET
> button) to gain control of it again...
>

I tried lowerning the nice-level too, with good results. However I was

A) running ClimaPrediction under BOINC, not Einstein@Home.

Also I did this on a

B) laptop computer with frequency management of the CPU.

When running at a nice level of 19, my CPU was at about 500 MHz, however when decreasing the nice level to 1 (or was it 0 ?) the CPU used its full power at 1500 MHz, and the cooler fan ran at top speed. I did not notice any annoying sluggisness, though. I could keep on browsing the web, while having the CPU full throttle.

Needless to say I had adjusted my preferences to "run-always" for BOINC, otherwise it would of course step down whenever used keyboard or mouse.

Best Regards,
Morten

mhe
mhe
Joined: 24 Feb 05
Posts: 11
Credit: 7576
RAC: 0

> B) laptop computer with

Message 6036 in response to message 6035


> B) laptop computer with frequency management of the CPU.
>
> When running at a nice level of 19, my CPU was at about 500 MHz,

Hi,

the frequency management is configured to ignore niced processes. So it will not increase the frequency and core voltages if a process has 100% cpu and a nice level > 0. You can edit the config file which belongs to your management software to count niced processes as well.

regards

Comment viewing options

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