-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 2011-04-03 15:29, Oliver Kullmann wrote:

Hello,

I can not replicate your findings, because I don't have such long running heavy tasks. However, after hibernation the scheduling seems right on mine. I use the same session perhaps for weeks, with daily hibernation, and I don't get a slow desktop. I use gnome, though. And frequency adjustment works fine, as far as I can see. I'm on 11.2 still. It might happen that the desktop thinks that if the mouse and keyboard are idling, the computer is idling and slows things down. It is wrong if heavy task are running, of course. Or, it can take the chance to trigger background search jobs, which are disk intensive. You have to disable that, or wait till they finish. - -- Cheers / Saludos, Carlos E. R. (from 11.2 x86_64 "Emerald" at Telcontar) -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.0.12 (GNU/Linux) Comment: Using GnuPG with SUSE - http://enigmail.mozdev.org/ iEYEARECAAYFAk2YeqkACgkQtTMYHG2NR9W2tQCfXFSS+g7YQv0OCbTurwn28Uit F2gAn3E/J3lnn5aAUGfHo9WRC9ikupF3 =MzSL -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

I suggest upgrade if applicable. 2011/4/3 Oliver Kullmann :

Hello,

finally, the performance of Suse 11.3 (all updates applied) on my laptop (i5 quadcore, 4GB) is so awful, one needs to find out the reasons (or change the distribution, or whatever). I hope I find some help here. According to my experience, it's best to start trying to establish some real facts (which, unfortunately, seems to be already a non-trivial endeavour).

One basic problem, which I already encountered with previous Suse versions, is that the frequency setting doesn't work: after a few days of running, at least one of the four cores will be set to a lower frequency. To eliminate this (but see below) I set in the BIOS "no frequency scaling". The frequency shown in Ksysguard since then is constantly "2,394 MHz" (but this can't be; see below).

Another basic problem is that (reliably) after "sleeping" or "hibernation" the process scheduling doesn't function properly anymore, and typically all processes (whatever the numbers --- I tried hundreds) are scheduled only on two or sometimes even only on one core, while the other two or three cores are idle. Of course, would be great if this malfunction would be treated (finally, I thought Linux would be about speed; apparently these problems don't occur with the windows 7 still on that Laptop (though I don't really run that)). But to ignore that, and for the following I have always rebooted the Laptop  (as if it would be windows).

Direct after reboot everything is exactly as it should be. I have various computers, on which I perform (a lot of) scientific computations, where I can measure the times reliably, and the times are as they should be.

But after a few days of running (and I have tasks, which necessarily run over weeks) performance becomes very weak. It doesn't show up on Ksysquard (or top), there the percentages and frequencies are the same, but for example Kde sticks like glue (reaction time for any button press at least a second, and if you wait for another week, it's  5 seconds; scroll bars become unusable).

More reliably, the computations I run become a factor of 5 to 10 (in words, five to ten(!)) slower. Just running one process, memory usage about 50%. So nothing goes on, except that performance is awful.

To make these numbers more trustworthy to the outside world, is there some standard (simple!) Linux tool/command to measure some (very) basic performance? I don't mean an elaborated benchmark set, but just a very simple common program, where people know what the running times should be? The only explanation I have for the slow performance is that the frequency goes down dramatically (without Linux noting it), and that can be demonstrated with something very simple. Of course, I have my own programs to run, and I can write a simple C program, but I guess it would make communication much easier if one would have a standard program.

Hopefully we can shed some light on that.

Thanks for your consideration.

Oliver

-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org

Hello,

. . .

Hopefully we can shed some light on that.

Thanks for your consideration.

Oliver

Doubtful this is of any help, but fwiw . . . First, besides disabling scaling in the BIOS, also instruct the kernel to do the same. Add the "cpufreq=no" argument to your boot line. My cpu's are overclocked, and this works at least as far as scaling is concerned. On my Phenom II wkstn, I have also found performance degradation over time. Tools like top, htop, and ksysguard will report much higher cpu consumption than the detail processes sum up to. I have noticed that X will momentarily but repeatedly spike the cpu, and that the pct creeps up. While the only little evidence I've been able to find has been X, with possibly Plasma or Firefox also somehow related, I can't pin it down with certainty. Usually re-starting X resolves, but on occasion I need to re-boot (I suspect to flush memory). Of note is that this behavior continues with different kernels. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org