I can’t find much info on core parking in opensuse.
Does anyone know if it can do it, and if so, how to set it up? I’ve monitored my CPU for hours, and it throttles back to 1.6ghz, but never parks.
My current computer is far too powerful 90% of the time. It would make a lot of sense to turn off a few cores to save some power.
Thanks
I can’t find much info on core parking in opensuse.
Does anyone know if it can do it, and if so, how to set it up? I’ve monitored my CPU for hours, and it throttles back to 1.6ghz, but never parks.
My current computer is far too powerful 90% of the time. It would make a lot of sense to turn off a few cores to save some power.
Thanks
Not sure about parking a CPU, but here a couple of blogs that might be helpful…
YaST Power Management - Control Your CPU Energy Usage How To & FAQ - Blogs - openSUSE Forums
Thank You,
On 10/08/2011 11:06 PM, wicked1 wrote:
> I can’t find much info on core parking in opensuse.
where do you find a lot of information on “core parking”?
*
the Intel or AMD site, maybe?
–
DD
openSUSE®, the “German Automobiles” of operating systems
*
On 2011-10-08 23:06, wicked1 wrote:
>
> I can’t find much info on core parking in opensuse.
> Does anyone know if it can do it, and if so, how to set it up? I’ve
> monitored my CPU for hours, and it throttles back to 1.6ghz, but never
> parks.
Never heard of it.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” at Telcontar)
Really? All the new intel core processors can do it. (I say new, but I got my first i7 3 years ago so that’s actually quite old in comp terms)
not much to explain about what it does… turns unused cores off. And the OS controls it. (like turbo-boost, which does work well in linux (Most utils still report the wrong speed, but “turbostat” showed me exactly what the cores are really doing))
Just google “i7 core parking” and you’ll see thousands of posts on the topic w/ windows… Usually about how to DISable it, so their games don’t lag or whatever.
Really? All the new intel core processors can do it. I’ve got an i7.
not much to explain about what it does… turns unused cores off. And the OS controls it. (like turbo-boost, which does work well in linux (Most utils still report the wrong speed, but “turbostat” showed me exactly what the cores are really doing))
Just google “i7 core parking” and you’ll see thousands of posts on the topic w/ windows… Usually about how to DISable it, so their games don’t lag or whatever.
On 2011-10-09 16:36, wicked1 wrote:
>
> Really? All the new intel core processors can do it. (I say new, but I
> got my first i7 3 years ago so that’s actually quite old in comp terms)
>
> not much to explain about what it does… turns unused cores off. And
> the OS controls it. (like turbo-boost, which does work well in linux
> (Most utils still report the wrong speed, but “turbostat” showed me
> exactly what the cores are really doing))
Well, it is news to me. No idea if the linux kernel controls it or not, and
then whether the admin can influence it or not.
I would have a look at kernel config options (source code config) and try
to find about it.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” at Telcontar)
Really? All the new intel core processors can do it. (I say new, but I got my first i7 3 years ago so that’s actually quite old in comp terms)
not much to explain about what it does… turns unused cores off. And the OS controls it. (like turbo-boost, which does work well in linux (Most utils still report the wrong speed, but “turbostat” showed me exactly what the cores are really doing))
Just google “i7 core parking” and you’ll see thousands of posts on the topic w/ windows… Usually about how to DISable it, so their games don’t lag or whatever.
So you did not indicate if you looked at the cfu bash script which does allow you to set the dynamic frequency used by your CPU to save energy and reduce heat.
Thank You,
On 10/09/2011 04:36 PM, wicked1 wrote:
>
> Just google “i7 core parking” and you’ll see thousands of posts on the
> topic w/ windows
then, google
linux "i7 core parking"
and you exactly four
hits (one of which is yours)…i therefore leap to the conclusion that
WinTel worked together to make i7 and Win7 to work and play nicely
together (they need to do that you know, because the software is so
bloated and games so resource demanding. . .)
so, i recommend you keep the games on your WinTel and your real work on
something else…and, if you bought way too much game machine and wish
you it were an energy sipping machine for word processing and normal
browsing i can highly recommend the hardware and software i have (linked
in sig)…
btw, it would have been nice had you mentioned i7 in your initial post…
–
DD
Caveat-Hardware-Software-
openSUSE®, the “German Automobiles” of operating systems
I sense some attitude here… I did search first and didn’t find anything so I decided to ask. I never said anything about me playing games, and who cares if I do? But I haven’t since about 98… And, I do have other boxes, some of which use Atom processors, so I do know about saving power. My SBS/Exchange server runs on an atom! (Sure, it’s only got a few users) And, in MS’s defense, 7 runs beautifully on low powered machines. (Unless the idiot using it can’t figure out how to click NO)
But on the other hand, I prefer linux for my personal daily workstation. It’s definitely not “too much machine” when I’m working. But, at night I slowly browse the web and don’t need it all. BUT, whatever! People have a lot of reasons for needing one machine to have power sometimes, and then not waste it at others. The ability is there. Oh, and since you have so much “wintel” attitude, osx can park cores, too.
And finally more than just i7’s can have their cores disabled, so I didn’t want to narrow my question parameters too much.
jdmcdaniel3, yeah, as I mentioned, they throttle back to 1.6ghz perfectly… That’s unfortunately the low end limit for the processor, though.
Hi wicked1.
You can disable cores at boot time by using the ‘maxcpus’ boot parameter. For example to limit use to a single core, you’d use
maxcpus=1
You can test this behaviour, by typing this at the grub menu prompt, (or to make it permanent, add it to the appropriate boot entry in /boot/grub/menu.lst)
I guess you could present several different options at boot time, so that depending on the intended use, you could specify the number of cores accordingly.
I’ve never played with this kernel parameter before. FWIW, I have a Intel(R) Core™2 Duo CPU T8300 @ 2.40GHz processor (as reported by cat /proc/cpuinfo). After booting with ‘maxcpus=1’, I then checked the number of active cores with
grep -c cores /proc/cpuinfo
which confirmed that only 1 core was active. (Rebooting without the parameter results in both being active, as expected).
Hi
You can offline a cpu, but it’s not an automated process;
#To display the current cpu state
cat /sys/devices/system/cpu/cpuX/online
#To offline a cpu
echo 0 > /sys/devices/system/cpu/cpuX/online
#To online a cpu
echo 1 > /sys/devices/system/cpu/cpuX/online
More info can be found in the kernel documentation search
for the file called cpu-hotplug.txt.
–
Cheers Malcolm °¿° (Linux Counter #276890)
openSUSE 11.4 (x86_64) Kernel 2.6.37.6-0.7-desktop
up 3:09, 3 users, load average: 0.04, 0.24, 0.30
GPU GeForce 8600 GTS Silent - Driver Version: 280.13
More on the dynamic approach here:
http://www.kernel.org/doc/Documentation/cpu-hotplug.txt
While playing with this, I found that I couldn’t disable cpu0 (although cpu1 can be controlled at will). This is mentioned in the cpu-hotplug documentation
Q: Why can’t i remove CPU0 on some systems?
A: Some architectures may have some special dependency on a certain CPU.
On 10/10/2011 05:26 AM, wicked1 wrote:
>
> I sense some attitude here… I did search first and didn’t find
> anything so I decided to ask. I never said anything about me playing
> games, and who cares if I do?
sorry, i allowed myself to be mislead by wrongly parsing (reading ‘the’
for ‘their’) those parts of your posts: “My current computer is far too
powerful 90% of the time. . . DISable it, so their games don’t lag or
whatever.”
still, like some others here i’d not heard or seen the “core park”
term…sometimes folks post questions here trying to replicate what
they are used to, or what is the latest on how to tweak their (other)
system…(for example) two or three years ago there was a flurry of
questions about which “download accelerator” should be used in Linux…i
wrongly thought “core parking” was a similar new thing only useful to
game systems . . .
sorry.
–
DD
openSUSE®, the “German Automobiles” of operating systems
Hey guys,
came here because of a G. search for linux core parking (that is the 2nd hit for that).
Now in case someone missed the concept of core parking, let me brief that for you :
in newer processors (core i3/5/7 with Hyper Threading and AMD Bulldozer) you get a number of real processor cores and a number of virtual ones.
Say core i7 2500k - 4C/8T.
While that is a good thing and will speed up integer calculations significantly, for some FPU operations the virtual cores are not as effective as the idle one.
Example - dual threaded gzip process uses physical core 1, 2 threads. It runs not that good, because there are other 3cores with 6 threads available and the process of core parking is to always allocate to phycial cores first, then to virtual ones.
Linux will read the cores in a sequence say C1 owns T1 and T2, C2 owns T3 and T4 …
So when one does cat /proc/cpuinfo he will see 8 CPUs (which correspond to the number of threads).
When load is presented to the system in the typical scenario CPU 1 and CPU 2 will receive a fair amount of usage. But that isn’t exactly right.
CPU 1 and CPU 3 are on 2 different physical cores, whereas CPU1 and CPU2 are on the same physical core.
That is indeed the problem. So in gaming worlds (where almost everything is FPU calculations) this has a drastic effect so this is why gamers have found the issue at once.
While no one running linux will care about games, it would be indeed important to know what approach can we implement to better distribute the load of less threaded applications on a multithreaded Processor which uses SMT.
Any thoughts about that?
Cheers,
SV
On 04/05/2012 10:46 AM, sivFED wrote:
>
> Hey guys,
>
> came here because of a G. search for linux core parking (that is the
> 2nd hit for that).
>
> Now in case someone missed the concept of core parking, let me brief
> that for you :
> in newer processors (core i3/5/7 with Hyper Threading and AMD
> Bulldozer) you get a number of real processor cores and a number of
> virtual ones.
>
> Say core i7 2500k - 4C/8T.
>
> While that is a good thing and will speed up integer calculations
> significantly, for some FPU operations the virtual cores are not as
> effective as the idle one.
>
> Example - dual threaded gzip process uses physical core 1, 2 threads.
> It runs not that good, because there are other 3cores with 6 threads
> available and the process of core parking is to always allocate to
> phycial cores first, then to virtual ones.
>
> Linux will read the cores in a sequence say C1 owns T1 and T2, C2 owns
> T3 and T4 …
>
> So when one does cat /proc/cpuinfo he will see 8 CPUs (which correspond
> to the number of threads).
>
> When load is presented to the system in the typical scenario CPU 1 and
> CPU 2 will receive a fair amount of usage. But that isn’t exactly right.
>
> CPU 1 and CPU 3 are on 2 different physical cores, whereas CPU1 and
> CPU2 are on the same physical core.
>
> That is indeed the problem. So in gaming worlds (where almost
> everything is FPU calculations) this has a drastic effect so this is why
> gamers have found the issue at once.
> While no one running linux will care about games, it would be indeed
> important to know what approach can we implement to better distribute
> the load of less threaded applications on a multithreaded Processor
> which uses SMT.
>
> Any thoughts about that?
As very few developers hang out on this Forum, and likely none of them
understand scheduling in the detail required, I suggest you take this question
to linux-kernel@vger.kernel.org.
On 2012-04-05 19:13, Larry Finger wrote:
>
> As very few developers hang out on this Forum, and likely none of them
> understand scheduling in the detail required, I suggest you take this
> question to linux-kernel@vger.kernel.org.
Or try opensuse-kernel first.
–
Cheers / Saludos,
Carlos E. R.
(from 11.4 x86_64 “Celadon” at Telcontar)