CPU performance problem on old laptop

Konstantin Svist fry.kun at gmail.com
Mon Dec 17 05:00:25 UTC 2007


Konstantin Svist wrote:
> Hi all,
>
> I'm investigating slowness on my old laptop (HP ze4400, circa 2004; 
> see related thread "hdparm shows poor performance on cached reads for 
> PATA drive" if you're interested).
> Apparently the HD cached read speed is not the only thing that 
> suffered a performance hit.
>
> I ran nbench (http://www.tux.org/~mayer/linux/bmark.html) on the 
> laptop and got the results that showed that the performance is roughly 
> half of my newer laptop (Dell E1505) - which was very surprising, 
> since the new laptop is not THAT much more powerful than the old one 
> (plus, I'm pretty sure the test didn't use both cpu cores on the new 
> one).
>
> I decided to try nbench in knoppix environment - and to my surprise, 
> it performed nearly as well as (and in a few tests, better than) the 
> new laptop!!
> I've ran the same test on the new laptop in knoppix - and found that 
> the results weren't affected all that much.
>
> I've also tried nbench on the old laptop in runlevels 3 and 1: 3 gives 
> same performance as 5, but 1 gets somewhere in the middle between 
> knoppix and F8/runlevel 5.
>
> What should I investigate next, to find the cause?
> What do the results so far suggest? I'm guessing some part of the 
> problem is the services that start up in init 3/5 - and perhaps the 
> new kernel can be partially blamed, as well
> Has anyone else experienced a problem like this? What can be done 
> about it?
>

Answering my own question

After some more digging: apparently the resource hog was wdaemon (daemon 
that emulates a drawing tablet so that it can be hotplugged into X).
This was a surprise since my new laptop also has the same one installed 
- and doesn't exhibit the slowdown.
I've emailed the maintainer and asked him to check out the problem.

Another service that affected the speed was cpuspeed - but that's only 
when "performance" was not selected as the current profile.




P.S. Apparently this was the same issue as the HD performance indicated 
by hdparm - who knew :)






More information about the users mailing list