[opensuse] apparently Suse-Linux can't properly use quadcore processor???
Hello, I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it! That's not good, since I rely on doing computational experiments with my laptop, and, of course, I purchased the quadcore laptop to do exactly that. More precisely: A month ago, after having installed Suse 11.3, I already experienced that Linux hesitates to use all four cores, say, it puts the first four processes on the first core etc., never used all. But after running experiments for one day or so, it "warmed up", and from then on it worked. I didn't turn it off for a month, and it worked as expected: for 1, 2, 3, 4 running processes it used 1, 2, 3, 4 cores, and more processes then led to equal distribution. The day before yesterday I made a complete update, and I noticed something about a "kernel optimised for laptops". That made me already worry (what could this mean other than energy saving and such stuff --- while I need full power), but I hoped the best. However since since it keeps at least one core idle, apparently independent of how many processes you are running (that is, long time processes --- shorter ones apparently get scheduled on the idle processor). And actually most of the time 2 cores (from the four) are kept idle, so that for example 8 long-time processes are cramped on 2 cores! Another thing which stopped working after the update is the temperate measurement of "Bubblemoon": In order for it to work, before I had to restart it several times, with several settings, but at some time it started working. Now the temperature measurements are just complete nonsense. Could there be a relation? Hope somebody can help. I can not imagine that Linux can't handle a quadcore processor! Thanks for your attention. Oliver -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sun, Oct 31, 2010 at 20:59, Oliver Kullmann <O.Kullmann@swansea.ac.uk> wrote:
Hello,
I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it! That's not good, since I rely on doing computational experiments with my laptop, and, of course, I purchased the quadcore laptop to do exactly that.
What is the exact kernel you're running? Is it 32 bit, 64 bit? PAE version... etc. C -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
I tried Yast to find out information about that, but couldn't find anything ("systems" has only "kernel settings", but apparently there no information can be found). So all I can say that it is a 64bit system, and that I've updated Suse 11.3 2 days ago (where a new kernel version has been installed). The first lines of hwinfo are libhd version 17.3 (x86-64) using /var/lib/hardware kernel version is 2.6 ----- /proc/cmdline ----- root=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A60V9698-part6 resume=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A 60V9698-part5 splash=silent quiet vga=0x317 ----- /proc/cmdline end ----- The kernel version:
uname -r 2.6.34.7-0.5-desktop
Oliver On Sun, Oct 31, 2010 at 09:04:58PM +0100, C wrote:
On Sun, Oct 31, 2010 at 20:59, Oliver Kullmann <O.Kullmann@swansea.ac.uk> wrote:
Hello,
I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it! That's not good, since I rely on doing computational experiments with my laptop, and, of course, I purchased the quadcore laptop to do exactly that.
What is the exact kernel you're running? Is it 32 bit, 64 bit? PAE version... etc.
C -- 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
* Oliver Kullmann <O.Kullmann@swansea.ac.uk> [10-31-10 16:30]:
I tried Yast to find out information about that, but couldn't find anything ("systems" has only "kernel settings", but apparently there no information can be found).
So all I can say that it is a 64bit system, and that I've updated Suse 11.3 2 days ago (where a new kernel version has been installed). The first lines of hwinfo are
libhd version 17.3 (x86-64) using /var/lib/hardware kernel version is 2.6 ----- /proc/cmdline ----- root=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A60V9698-part6 resume=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A 60V9698-part5 splash=silent quiet vga=0x317 ----- /proc/cmdline end -----
The kernel version:
uname -r 2.6.34.7-0.5-desktop
rpm -qa kernel* will provide the version and arch ps: Please trim and don't top post. tks -- Patrick Shanahan Plainfield, Indiana, USA HOG # US1244711 http://wahoo.no-ip.org Photo Album: http://wahoo.no-ip.org/gallery2 Registered Linux User #207535 @ http://counter.li.org -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
I tried Yast to find out information about that, but couldn't find anything ("systems" has only "kernel settings", but apparently there no information can be found).
So all I can say that it is a 64bit system, and that I've updated Suse 11.3 2 days ago (where a new kernel version has been installed). The first lines of hwinfo are
libhd version 17.3 (x86-64) using /var/lib/hardware kernel version is 2.6 ----- /proc/cmdline ----- root=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A60V9698-part6 resume=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A 60V9698-part5 splash=silent quiet vga=0x317 ----- /proc/cmdline end -----
The kernel version:
uname -r 2.6.34.7-0.5-desktop
1. YaST/Hardware/Hardware Information/CPU 2. >cat /proc/cpuinfo 3. #hwinfo --cpu You should see a separate cpu record for each core, and each record should indicate nbr of cores (or units) to be 4. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sun, Oct 31, 2010 at 5:16 PM, dwgallien <dwgallien@gmail.com> wrote:
I tried Yast to find out information about that, but couldn't find anything ("systems" has only "kernel settings", but apparently there no information can be found).
So all I can say that it is a 64bit system, and that I've updated Suse 11.3 2 days ago (where a new kernel version has been installed). The first lines of hwinfo are
libhd version 17.3 (x86-64) using /var/lib/hardware kernel version is 2.6 ----- /proc/cmdline ----- root=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A60V9698-part6 resume=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A 60V9698-part5 splash=silent quiet vga=0x317 ----- /proc/cmdline end -----
The kernel version:
uname -r 2.6.34.7-0.5-desktop cat /etc/SuSE-release openSUSE 11.1 (x86_64) VERSION = 11.1
1. YaST/Hardware/Hardware Information/CPU
2. >cat /proc/cpuinfo
3. #hwinfo --cpu
You should see a separate cpu record for each core, and each record should indicate nbr of cores (or units) to be 4. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
To see what kernel you are current, just do this. uname -a You will either see i386/i686 which is the the 32bit, if you see something like this x86_64 then you are running the 64-bits, this is from my box running 11.1. Linux tatl0wlluna 2.6.27.48-0.2-default #1 SMP 2010-07-29 20:06:52 +0200 x86_64 x86_64 x86_64 GNU/Linux Also, take a look at /etc/SuSE-release to see what was installed... cat /etc/SuSE-release openSUSE 11.1 (x86_64) VERSION = 11.1 -- ----------------------------------------- Discover it! Enjoy it! Share it! openSUSE Linux. ----------------------------------------- openSUSE -- en.opensuse.org/User:Terrorpup openSUSE Ambassador/openSUSE Member skype,twiiter,identica,friendfeed -- terrorpup freenode(irc) --terrorpup/lupinstein Have you tried SUSE Studio? Need to create a Live CD, an app you want to package and distribute , or create your own linux distro. Give SUSE Studio a try. www.susestudio.com. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
my system uses quad core cpu's fine. i'm running 64bit linux(2.6.36). here is 'uname -a' on my system: Linux linux 2.6.36 #2 SMP Fri Oct 22 23:42:00 CDT 2010 x86_64 x86_64 x86_64 GNU/Linux make sure you are not passing the 'nosmp' flag to the kernel during bootup(means no symmetric multi processing). and i've noticed that you are running a desktop kernel and i don't think it should but it may play a factor in the issue here. i personally run a self-compiled kernel so i can't say whether running a desktop kernel matters or not. and here is the output of ls /sys/devices/system/cpu : cpu0 cpu1 cpu2 cpu3 cpufreq cpuidle kernel_max offline online perf_events possible present probe release your system should also show something like that(listing cpu's 0 to 3). there is a great utility called htop that will show you cpu usage in real time. i've run 'make -j8' or 'make -j4' on building something like the kernel or some other source tree that'll take time to compile and i've seen the cpu usage be 100% on all cpu's when make is using 8 threads and around 80 something with 4 threads. and don't forget that the only way all your 4 cores will be used is if the program you are running is running at least 4 threads of execution. --- On Sun, 10/31/10, Chuck Payne <terrorpup@gmail.com> wrote:
I tried Yast to find out information about that, but couldn't find anything ("systems" has only "kernel settings", but apparently there no information can be found).
So all I can say that it is a 64bit system, and
From: Chuck Payne <terrorpup@gmail.com> Subject: Re: [opensuse] apparently Suse-Linux can't properly use quadcore processor??? To: "dwgallien" <dwgallien@gmail.com> Cc: opensuse@opensuse.org Date: Sunday, October 31, 2010, 4:24 PM On Sun, Oct 31, 2010 at 5:16 PM, dwgallien <dwgallien@gmail.com> wrote: that
I've updated Suse 11.3 2 days ago (where a new kernel version has been installed). The first lines of hwinfo are
libhd version 17.3 (x86-64) using /var/lib/hardware kernel version is 2.6 ----- /proc/cmdline ----- root=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A60V9698-part6 resume=/dev/disk/by-id/ata-WDC_WD5000BEVT-75A0RT0_WD-WXB1A 60V9698-part5 splash=silent quiet vga=0x317 ----- /proc/cmdline end -----
The kernel version:
uname -r 2.6.34.7-0.5-desktop cat /etc/SuSE-release openSUSE 11.1 (x86_64) VERSION = 11.1
1. YaST/Hardware/Hardware Information/CPU
2. >cat /proc/cpuinfo
3. #hwinfo --cpu
You should see a separate cpu record for each core,
and each record should indicate nbr of cores (or units) to be 4.
-- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
To see what kernel you are current, just do this.
uname -a
You will either see i386/i686 which is the the 32bit, if you see something like this x86_64 then you are running the 64-bits, this is from my box running 11.1.
Linux tatl0wlluna 2.6.27.48-0.2-default #1 SMP 2010-07-29 20:06:52 +0200 x86_64 x86_64 x86_64 GNU/Linux
Also, take a look at /etc/SuSE-release to see what was installed...
cat /etc/SuSE-release openSUSE 11.1 (x86_64) VERSION = 11.1
-- ----------------------------------------- Discover it! Enjoy it! Share it! openSUSE Linux. ----------------------------------------- openSUSE -- en.opensuse.org/User:Terrorpup openSUSE Ambassador/openSUSE Member skype,twiiter,identica,friendfeed -- terrorpup freenode(irc) --terrorpup/lupinstein
Have you tried SUSE Studio? Need to create a Live CD, an app you want to package and distribute , or create your own linux distro. Give SUSE Studio a try. www.susestudio.com. -- 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
Oliver Kullmann wrote:
A month ago, after having installed Suse 11.3, I already experienced that Linux hesitates to use all four cores, say, it puts the first four processes on the first core etc., never used all. The day before yesterday I made a complete update, and I noticed something about a "kernel optimised for laptops". That made me already worry (what could this mean other than energy saving and such stuff --- while I need full power), but I hoped the best. However since since it keeps at least one core idle, apparently independent of how many processes you are running (that is, long time processes --- shorter ones apparently get scheduled on the idle processor). And actually most of the time 2 cores (from the four) are kept idle, so that for example 8 long-time processes are cramped on 2 cores!
Are they also CPU-bound?
Hope somebody can help. I can not imagine that Linux can't handle a quadcore processor!
No, it works very well here, I have several multi-CPU and multi-core boxes running 11.3 -- Per Jessen, Zürich (8.4°C) -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On 31.10.2010, Oliver Kullmann wrote:
I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it!
You maybe could be affected by this phenomenon here: https://bugzilla.kernel.org/show_bug.cgi?id=19702 -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Hello, thanks for the inputs. On Mon, Nov 01, 2010 at 08:58:28AM +0100, Heinz Diehl wrote:
On 31.10.2010, Oliver Kullmann wrote:
I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it!
You maybe could be affected by this phenomenon here: https://bugzilla.kernel.org/show_bug.cgi?id=19702
Actually, frequency doesn't seem to be a (big) problem. Actually, yesterday for the first time it got stuck at 1.2 GHz (scheme "performance" regarding cpu-frequency), but switching to "on demand" got it back on track (switching for each core between 1.2 GHz and 2.4 GHz). So there is something wrong, but by just toggling the modes it seems fixable. Regarding the usage of all four cores: NOW, as I had it a week ago, it runs fine again: Using all four cores as required (independent of the frequency). And, again, Bubblemoon shows correct (that is, sensible) cpu temperature. Could it be that if the temperature sensor doesn't work properly, then the system "hesitates" to use all cores, and after some time the system "warms up", the temperature sensor starts working, and all cores are being used? Here is the data for the various outputs (the frequency shown is currently 1.2GHz, since I run on battery): ------------------------------------------------------
rpm -qa kernel* kernel-default-2.6.34.7-0.5.1.x86_64 kernel-desktop-2.6.34.7-0.5.1.x86_64 kernel-firmware-20100617-2.2.noarch
cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 0 cpu cores : 2 apicid : 0 initial apicid : 0 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.01 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management:
processor : 1 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 2 cpu cores : 2 apicid : 4 initial apicid : 4 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.05 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management: processor : 2 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 0 cpu cores : 2 apicid : 1 initial apicid : 1 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.04 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management: processor : 3 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 2 cpu cores : 2 apicid : 5 initial apicid : 5 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.02 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management:
cat /etc/SuSE-release openSUSE 11.3 (x86_64) VERSION = 11.3
kullmann-0:~> uname --all Linux linux-ubd8.site 2.6.34.7-0.5-desktop #1 SMP PREEMPT 2010-10-25 08:40:12 +0200 x86_64 x86_64 x86_64 GNU/Linux ----------------------------------------------------------- thanks again Oliver -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
Hello,
thanks for the inputs.
On Mon, Nov 01, 2010 at 08:58:28AM +0100, Heinz Diehl wrote:
On 31.10.2010, Oliver Kullmann wrote:
I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it!
You maybe could be affected by this phenomenon here: https://bugzilla.kernel.org/show_bug.cgi?id=19702
Actually, frequency doesn't seem to be a (big) problem. Actually, yesterday for the first time it got stuck at 1.2 GHz (scheme "performance" regarding cpu-frequency), but switching to "on demand" got it back on track (switching for each core between 1.2 GHz and 2.4 GHz). So there is something wrong, but by just toggling the modes it seems fixable.
Regarding the usage of all four cores: NOW, as I had it a week ago, it runs fine again: Using all four cores as required (independent of the frequency). And, again, Bubblemoon shows correct (that is, sensible) cpu temperature. Could it be that if the temperature sensor doesn't work properly, then the system "hesitates" to use all cores, and after some time the system "warms up", the temperature sensor starts working, and all cores are being used?
Here is the data for the various outputs (the frequency shown is currently 1.2GHz, since I run on battery):
------------------------------------------------------
rpm -qa kernel* kernel-default-2.6.34.7-0.5.1.x86_64 kernel-desktop-2.6.34.7-0.5.1.x86_64 kernel-firmware-20100617-2.2.noarch
cat /proc/cpuinfo processor : 0 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 0 cpu cores : 2 apicid : 0 initial apicid : 0 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.01 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management:
processor : 1 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 2 cpu cores : 2 apicid : 4 initial apicid : 4 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.05 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management:
processor : 2 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 0 cpu cores : 2 apicid : 1 initial apicid : 1 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.04 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management:
processor : 3 vendor_id : GenuineIntel cpu family : 6 model : 37 model name : Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz stepping : 5 cpu MHz : 1199.000 cache size : 3072 KB physical id : 0 siblings : 4 core id : 2 cpu cores : 2 apicid : 5 initial apicid : 5 fpu : yes fpu_exception : yes cpuid level : 11 wp : yes flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology nonstop_tsc aperfmperf pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 popcnt aes lahf_lm ida arat tpr_shadow vnmi flexpriority ept vpid bogomips : 4788.02 clflush size : 64 cache_alignment : 64 address sizes : 36 bits physical, 48 bits virtual power management:
cat /etc/SuSE-release openSUSE 11.3 (x86_64) VERSION = 11.3
kullmann-0:~> uname --all Linux linux-ubd8.site 2.6.34.7-0.5-desktop #1 SMP PREEMPT 2010-10-25 08:40:12 +0200 x86_64 x86_64 x86_64 GNU/Linux
-----------------------------------------------------------
thanks again
Oliver
You may be talking about different (albeit inter-related) things - throttling, temperature, and core threading. You may be seeing the effects of either the kernel's interaction with the cpu's throttling mechanism, the bios ACPI, or KDE's power management application. With throttling enabled, the cpu will respond to demand by increasing the voltage which will in turn increase the temperature. You indicated in your first post that you "need full power"; if so but you also want to conserve battery, then "ondemand" is the correct scaling policy. You should expect to see the processors throttled down unless the cpu needs more power to perform the requested task. Throttling can move very fast, although of course under full load the processor will be pegged at its maximum. In the still unresolved bug report ref'd in Heinz reply, there are some tools listed with which you can see the actual governor settings and detailed throttling results as it occurs. From that bug report there is suspicion that there may be a kernel bug but there is also the possibility that the problem is related to how the bios is handling ACPI; unfortunately bios bugs particularly re ACPI are not that uncommon so it wouldn't hurt the check for a bios update for your machine. If throttling is not working properly for you and there is no bios update, I would suggest following that bug thread (if it's in the kernel, it is serious) until a solution if found. If you can manage w/o throttling for a time, you can disable it altogether in the kernel by adding "cpufreq=no" to the kernel boot line - judging from a couple of the posts in the bug thread, you may need to do that to get above an incorrectly imposed limit. Re temperature, that is driven by core load not the other way around. As load is put on the processor, voltage is increased to raise clock which in turn generates the greater heat. I don't know Bubblemoon so I can't comment on its temperature reading method. It is conceivable that it is getting its temperature from ACPI (which if itself is a problem as hypothesized in the above bug report, I suppose could also affect the temp readings), from a cpu socket sensor, or from a thermal sensor embedded in the cpu itself; when the sensors package is being used for readings sometimes the algorithm needs to be adjusted (in the config file) for a particular machine. I use AMD processors, with which there have been hardware issues in the past with the internal thermal sensor as well as the sensor driver in the sensors package (since resolved). I don't know if there are any such issues with Intel. So in other words, temp variations could be due to the throttling issue, or not; could also be due to where/how Bubblemoon gets its readings. I would compare to what the sensors package reports. Re distribution of processes on the cores, my reply would be the same as Larry's. Note in your cpuinfo the number of cores is 2, not 4. Your i5 is essentially 2 dual-cores bolted together; that is a different architecture than a true quad. As already noted, the kernel will allocate load differently on the latter vs the former. Nothing above is a definitive answer, but hopefully helps a bit. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
dwgallien wrote:
Re distribution of processes on the cores, my reply would be the same as Larry's. Note in your cpuinfo the number of cores is 2, not 4. Your i5 is essentially 2 dual-cores bolted together;
2 x dual-core = 4 cores. No, an Intel Core i5 is a dual-core processor with hyper-threading. I don't know if the Linux kernel take hyper-threading into account when scheduling. Google will no doubt be able to help. -- Per Jessen, Zürich (9.2°C) -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
dwgallien wrote:
Re distribution of processes on the cores, my reply would be the same as Larry's. Note in your cpuinfo the number of cores is 2, not 4. Your i5 is essentially 2 dual-cores bolted together;
2 x dual-core = 4 cores. No, an Intel Core i5 is a dual-core processor with hyper-threading.
I don't know if the Linux kernel take hyper-threading into account when scheduling. Google will no doubt be able to help.
Per, thanks for the follow-up. My wording was poor. The i5 520 series is dual-core with HT, so the OS sees 4 "logical" cores. The SMP kernel of course supports HT. I also an unfamiliar with how the current scheduler manages the threads, although I recall reading an Intel paper indicating that doing so in an optimized fashion was a kernel priority. Oliver, your cpuinfo shows that the kernel sees hyper-threading enabled. I may be out-of-date on this point, but IIRC how an application is written will have an impact on the load distribution. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
You may be talking about different (albeit inter-related) things - throttling, temperature, and core threading.
You may be seeing the effects of either the kernel's interaction with the cpu's throttling mechanism, the bios ACPI, or KDE's power management application. With throttling enabled, the cpu will respond to demand by increasing the voltage which will in turn increase the temperature. You indicated in your first post that you "need full power"; if so but you also want to conserve battery, then "ondemand" is the correct scaling policy. You should expect to see the processors throttled down unless the cpu needs more power to perform the requested task. Throttling can move very fast, although of course under full load the processor will be pegged at its maximum. In the still unresolved bug report ref'd in Heinz reply, there are some tools listed with which you can see the actual governor settings and detailed throttling results as it occurs. From that bug report there is suspicion that there may be a kernel bug but there is also the possibility that the problem is related to how the bios is handling ACPI; unfortunately bios bugs particularly re ACPI are not that uncommon so it wouldn't hurt the check for a bios update for your machine. If throttling is not working properly for you and there is no bios update, I would suggest following that bug thread (if it's in the kernel, it is serious) until a solution if found. If you can manage w/o throttling for a time, you can disable it altogether in the kernel by adding "cpufreq=no" to the kernel boot line - judging from a couple of the posts in the bug thread, you may need to do that to get above an incorrectly imposed limit.
I used to use "performance", which used to stay on 2.4 GHz, but then I experienced a bug, where it stayed on 1.2 GHz, and switching to "ondemand" just repaired that. So it seems that there is a bug, but by just switching back and forth it can be resolved.
Re temperature, that is driven by core load not the other way around.
I just made a hypothesis, which could explain the experimental fact that for the first day or so tasks are only distributed to 2 processors (mostly), while then, "suddenly", it starts working as expected, and this alwayed happened at the same time when temperature monitoring started working.
As load is put on the processor, voltage is increased to raise clock which in turn generates the greater heat. I don't know Bubblemoon so I can't comment on its temperature reading method. It is conceivable that it is getting its temperature from ACPI (which if itself is a problem as hypothesized in the above bug report, I suppose could also affect the temp readings), from a cpu socket sensor, or from a thermal sensor embedded in the cpu itself; when the sensors package is being used for readings sometimes the algorithm needs to be adjusted (in the config file) for a particular machine. I use AMD processors, with which there have been hardware issues in the past with the internal thermal sensor as well as the sensor driver in the sensors package (since resolved). I don't know if there are any such issues with Intel. So in other words, temp variations could be due to the throttling issue, or not; could also be due to where/how Bubblemoon gets its readings. I would compare to what the sensors package reports.
Re distribution of processes on the cores, my reply would be the same as Larry's. Note in your cpuinfo the number of cores is 2, not 4. Your i5 is essentially 2 dual-cores bolted together; that is a different architecture than a true quad. As already noted, the kernel will allocate load differently on the latter vs the former.
As I explained in my other e-mail, for the jobs I'm running, after that initial phase it's over, I get the power of 4 independent processors (for what I'm running). So from my perspective I would just like to know how to skip that initial phase. But perhaps nobody knows.
Nothing above is a definitive answer, but hopefully helps a bit.
Yes, thanks! Oliver -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Sun, Oct 31, 2010 at 3:59 PM, Oliver Kullmann <O.Kullmann@swansea.ac.uk> wrote:
I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it! That's not good, since I rely on doing computational experiments with my laptop, and, of course, I purchased the quadcore laptop to do exactly that.
You are aware that a mobile Core i5 is actually a dual core processor with hyper threading? http://en.wikipedia.org/wiki/List_of_Intel_Core_i5_microprocessors Only 3 of the desktop Core i5's are true quad core, and they have hyperthreading disabled. The Linux kernel knows that the chip is actually a dual core with hyperthreading vs a true quad core, and will therefore schedule most tasks to work on the 2 actual cores, not the virtual cores. Win2k and XP had problems where they would use the virtual core and main core of the first chip of a dual proc system instead of the real 2 cores and performance would suffer. While the new Core ix series implementation of hyperthreading is better than what was on the P4s, it's still not a true additional core. Further, there are other issues: "Also, moving an application to two threads on a single SMT-enabled core will increase cache-thrashing by 42%, whereas it will decrease by 37% when moving to two cores." That's from this article: http://www.theinquirer.net/inquirer/news/1037948/arm-fan-hyperthreading Hyperthreading can be a nice boost, but a Core i7 with true quad core would be a better choice if you need 4 cores. -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
On Mon, Nov 01, 2010 at 11:16:16AM -0400, Larry Stotler wrote:
On Sun, Oct 31, 2010 at 3:59 PM, Oliver Kullmann <O.Kullmann@swansea.ac.uk> wrote:
I'm here experiencing an alarming problem (not just KDE or so, but the kernel): I have a laptop with a quadcore i5 processor, but the system doesn't want to use it! That's not good, since I rely on doing computational experiments with my laptop, and, of course, I purchased the quadcore laptop to do exactly that.
You are aware that a mobile Core i5 is actually a dual core processor with hyper threading?
As I said, after the (strange) "warming up phase" is over (for unknown reasons), it behaves exactly as having four independent processors (for the jobs I run, I can run 4 of them at the same speed as a single job).
http://en.wikipedia.org/wiki/List_of_Intel_Core_i5_microprocessors
Only 3 of the desktop Core i5's are true quad core, and they have hyperthreading disabled.
The Linux kernel knows that the chip is actually a dual core with hyperthreading vs a true quad core, and will therefore schedule most tasks to work on the 2 actual cores, not the virtual cores. Win2k and XP had problems where they would use the virtual core and main core of the first chip of a dual proc system instead of the real 2 cores and performance would suffer.
Perhaps that is the assumption at the beginning, and then it figures out differently?
While the new Core ix series implementation of hyperthreading is better than what was on the P4s, it's still not a true additional core. Further, there are other issues:
"Also, moving an application to two threads on a single SMT-enabled core will increase cache-thrashing by 42%, whereas it will decrease by 37% when moving to two cores."
That's from this article:
http://www.theinquirer.net/inquirer/news/1037948/arm-fan-hyperthreading
Hyperthreading can be a nice boost, but a Core i7 with true quad core would be a better choice if you need 4 cores.
I can't complain for what I'm running (some form of scientific computation, but not much numerics). Thanks Oliver -- To unsubscribe, e-mail: opensuse+unsubscribe@opensuse.org For additional commands, e-mail: opensuse+help@opensuse.org
participants (9)
-
C -
Chuck Payne -
dwgallien -
Heinz Diehl -
Larry Stotler -
michael getachew -
Oliver Kullmann -
Patrick Shanahan -
Per Jessen