Windows vs. Linux

Message boards : Number crunching : Windows vs. Linux
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · Next

AuthorMessage
Profile Toby
Volunteer tester
Avatar

Send message
Joined: 26 Oct 00
Posts: 1005
Credit: 6,366,949
RAC: 0
United States
Message 725163 - Posted: 12 Mar 2008, 19:00:39 UTC
Last modified: 12 Mar 2008, 19:03:16 UTC

I see the same behavior on my laptop running Ubuntu. Idle processes do not cause the kernel to initiate CPU power level change. You can change this behavior with the cpufreq-selector command. To make the CPU run at full speed all the time, just type this command into a terminal:
sudo cpufreq-selector -g performance

This causes the kernel to use the "performance" governor to determine when speedstep/powernow features are enabled. The other governors that are available are powersave, conservative, ondemand and userspace. I believe the default one is "ondemand" which, as the name suggests, keeps the CPU at low speed unless more power is needed. You can also lock the CPU in at a specific frequency with this command:
sudo cpufreq-selector -f 1670000

The 1670000 means 1.67 GHz (my CPU can run at 1, 1.67 or 2 GHz).

I'm sure there is some GNOME/KDE interface to this feature that doesn't require breaking out a terminal but thats the quick and dirty way :)

You can view the speed your CPU is currently running at with this command:
cat /proc/cpuinfo

and looking for the line that starts with "cpu MHz." There is also a GNOME panel widget (or whatever they're called) under the "System & Hardware" section entitled "CPU Frequency Scaling Monitor" that will graphically tell you at what speed your CPU is running.
A member of The Knights Who Say NI!
For rankings, history graphs and more, check out:
My BOINC stats site
ID: 725163 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 12055
Credit: 122,007,377
RAC: 68,975
United Kingdom
Message 725213 - Posted: 12 Mar 2008, 21:26:28 UTC - in response to Message 724592.  

I got "'systeminfo' is not recognized as an internal or external command, operable program or batch file." when I tried it. According to this write up, it's in Windows Professional only. I'll blag a copy and see if it's transferrable.

Do I need to slap you around and shake you like they did to that woman in "Airplane"??? (happens at 2:36 in the trailer video) :-P What are ya doing running XP Home?

Snaffled a copy of systeminfo.exe from c:\\windows\\system32 on an XP Pro I was setting up for a client, and dropped it into the same place on XP Home. Works fine, so there are no DLL dependencies to worry about. This machine says:

System Up Time:            15 Days, 3 Hours, 21 Minutes, 24 Seconds

- and that's after doing my 'Update Wednesday' chores this morning.

To answer Shirley,

1) I run XP Home SP2 on my daily driver, just as I ran Windows 98 SE on the previous daily driver, and WfW 3.11 on the one before that (or was there a 95b in between?), and DOS 6.2, and CP/M 1.4, and..... because they came pre-installed at OEM prices, and I've patched/updated them enough to get the reliability I need without adding features I don't. Although I run a file/print server at home, it doesn't need to be an active directory domain controller, so I don't need the domain membership Pro is needed for, and I don't need the advanced file and directory security. I specced Pro for the octo (didn't have much choice, really) to get dual socket support, but there's no other reason. And I'm not planning to rip the guts out of the daily driver (P4 2.0GHz) to put a dual-socket motherboard in there....

2) Some time ago, I put in an order for some MS Office software for a client. Included in the package I received (but not on the order I made, or the invoice I paid) were three copies of XP Home OEM, and three copies of XP Pro OEM. Under UK Law, I waited six months for them to be collected, but no show. So when I came to build my two quaddies last year, I had a use for my free gifts. I've kept the Pro in reserve, as I sometimes come across business clients who didn't know they should have ordered Pro to attach to their servers, and I used Home for the home crunchers for reasons as above.
ID: 725213 · Report as offensive
Profile ML1
Volunteer tester

Send message
Joined: 25 Nov 01
Posts: 9420
Credit: 7,345,054
RAC: 271
United Kingdom
Message 725483 - Posted: 13 Mar 2008, 9:47:07 UTC
Last modified: 13 Mar 2008, 9:52:10 UTC

Good sleuthing folks!

So it was the speedstep/powernow/cool'n'quiet throttling back the CPU clock frequency after all.


Out of interest, what is the comparison at the moment between the various versions of s@h science apps for the various OSes? Any tables anywhere?!

And what's with the latest optimisations? Is everything fairly static for the time being?

Happy crunchin',
Martin
See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)
ID: 725483 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 12055
Credit: 122,007,377
RAC: 68,975
United Kingdom
Message 725493 - Posted: 13 Mar 2008, 10:27:39 UTC - in response to Message 725483.  

Good sleuthing folks!

So it was the speedstep/powernow/cool'n'quiet throttling back the CPU clock frequency after all.

I don't think we've quite reached that point yet. It certainly seems to be the best working hypothesis so far, but it would be really cool confirmation (sic!) if one of the two posters who brought the matter to our attention - DeMus or Gerry Green - could run a couple of 72-credit WUs under Linux, with whatever it takes to get the CPU running at full speed on their specific system - Toby's suggestion sounds the most generic. Let us know the Task ID for the one(s) you do at full speed, and we can all see the CPU time for a proper comparison. That would nail it once and for all, and start to answer Martin's next question, as well.
ID: 725493 · Report as offensive
Gerry Green

Send message
Joined: 15 May 99
Posts: 13
Credit: 3,034,952
RAC: 0
United States
Message 725497 - Posted: 13 Mar 2008, 10:58:27 UTC - in response to Message 725493.  

Good sleuthing folks!

So it was the speedstep/powernow/cool'n'quiet throttling back the CPU clock frequency after all.

I don't think we've quite reached that point yet. It certainly seems to be the best working hypothesis so far, but it would be really cool confirmation (sic!) if one of the two posters who brought the matter to our attention - DeMus or Gerry Green - could run a couple of 72-credit WUs under Linux, with whatever it takes to get the CPU running at full speed on their specific system - Toby's suggestion sounds the most generic. Let us know the Task ID for the one(s) you do at full speed, and we can all see the CPU time for a proper comparison. That would nail it once and for all, and start to answer Martin's next question, as well.


I rebooted into the linux side this morning. cat /proc/cpuinfo confirmed that all 4 cores were running at 1.6GHz not 2.4. Even though BOINC is getting 100% of the CPU it is at the lower speed... That's pretty close to confirmation right there -- but I agree it will be interesting to see how long a 72 credit work unit takes.

I switched it to the "performance" option as opposed to ondemand, per Toby's suggestion and it is now showing 2.4 on all 4 cores.

I'm still running the stock client on the linux side at least till this test is done and will report back as soon as I see some 72 credit work units pop out that have been fully processed at the right speed.

ID: 725497 · Report as offensive
Nick: ID 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 11976
Credit: 32,160,047
RAC: 2,617
United Kingdom
Message 725503 - Posted: 13 Mar 2008, 11:23:22 UTC

Good work on the cpu throttling.
Just wondered if this subject should be a separate thread. So that one, it is short and two, a new title might bring it to the attention of the casual Linux visitor.
ID: 725503 · Report as offensive
Richard Haselgrove Project Donor
Volunteer tester

Send message
Joined: 4 Jul 99
Posts: 12055
Credit: 122,007,377
RAC: 68,975
United Kingdom
Message 725504 - Posted: 13 Mar 2008, 11:40:42 UTC - in response to Message 725503.  

Good work on the cpu throttling.
Just wondered if this subject should be a separate thread. So that one, it is short and two, a new title might bring it to the attention of the casual Linux visitor.

I was just thinking the same thing. Once the results of Gerry's test are in, we really ought to preserve the conclusion (but not WW3!) for the benefit of future Linux experimenters.

I don't think another thread is necessarily the answer, though. Unless it was sticky (and we've got enough of those already), it would soon fall off the front page. Perhaps a candidate for Jim-R's FAQs?

And of course, anyone who recommends Linux in the future without mentioning this little gotcha is going to end up shooting themselves in the foot - potentially some very dis-satified customers out there.
ID: 725504 · Report as offensive
Gerry Green

Send message
Joined: 15 May 99
Posts: 13
Credit: 3,034,952
RAC: 0
United States
Message 725566 - Posted: 13 Mar 2008, 15:44:44 UTC - in response to Message 725504.  

Here's a result that should have been processed entirely with the higher CPU speed this morning. I picked an older unit from the Vista side that was processed when I was still using the stock client for comparison. ARs are nearly equal.

For this one unit, linux took 3.5% longer. Compiler differences and/or other processes running on the linux side could easily explain that.

openSUSE 64bit at 2.4GHz

CPU time 11172.718251
stderr out <core_client_version>5.10.28</core_client_version>
<![CDATA[
<stderr_txt>
setiathome_enhanced 5.28 Revision: 26 g++ (GCC) 4.1.2 (Ubuntu 4.1.2-0ubuntu4)
libboinc: BOINC 6.1.0

Work Unit Info:
...............
WU true angle range is : 0.395519
Optimal function choices:




Vista 64

CPU time 10792.897985
stderr out <core_client_version>5.10.30</core_client_version>
<![CDATA[
<stderr_txt>
setiathome_enhanced 5.27 DevC++/MinGW

Work Unit Info:
...............
WU true angle range is : 0.395459
Optimal function choices:

ID: 725566 · Report as offensive
kykkeaton

Send message
Joined: 2 Jun 06
Posts: 1
Credit: 326,481
RAC: 0
United States
Message 725618 - Posted: 13 Mar 2008, 18:17:37 UTC - in response to Message 725566.  

This thread has me completely fascinated! And now I'll ask the very question I had intended to post before I got caught up into skimming the whole thread:

I run an Intel Celeron 2ghz w/ Debian stable.
That is not to say I am a Linux guru by no means ( read a complete novice ).
To see exact stats on this box see AllenF machines ( he's my team-mate in JenKen Group http://setiathome.berkeley.edu/team_display.php?teamid=126603 )

The machine is actually an effort between the both of us. I supplied the hardware and he maintains the software most of the time. I do dig a bit into it when I have time however, and would be very interested in disabling processes that 'nerf' the box. We have a few procs we use occasionally ( TeamSpeak, X ) that take some CPU to run but my thinking is I could in theory batch up something that would disable procs that we don't need to run S@h at full throttle. IE without power saving etc.
And yes I too have noticed that among all the computers I have in this room ( 4 desktops and 2 laptops in a 12ft x 12ft room ) the Debian box is by far the quietest. Heck these laptop fans occasionally 'scream'.... probably why I just had to replace the cooling fan unit on the IBM laptop, and the power jack & adapter on the Compaq. My expertise is in hardware,.. software has never been my thing. Both my laptops were recently DOA is why their stats are so low,.. took me a while til I had the time to disassemble them and break out my soldering iron. Anyway,.. too much off topic info I'm sure.. but if you linux gurus out there have any suggestions I'm ready and willing to read through them and discuss them with Allen. Thanks a bunch,.. and Happy Crunchin.

kykkeaton

P.S. In case your wondering,.. I haven't had my heating vents opened in this room all winter,.. and have even had to open the window to drop the temp down to a bearable 75f.


ID: 725618 · Report as offensive
UncleVom

Send message
Joined: 25 Dec 99
Posts: 123
Credit: 5,734,294
RAC: 0
Canada
Message 725650 - Posted: 13 Mar 2008, 19:33:54 UTC

I run mostly sidux, which is Debian sid, with added smoothing for the rough times.

With Lunatics/Crunch3r SETI apps I find it produces pretty much the same result times, give or take, as Win XP and slightly worse than Win98, sorry no figures to back this up, but in the past I have taken the time to convince myself.

I turn of power saving stuff in the BIOS setup and disable the fancontrol and powersave deamons in the distro that are sometime on by default.

It seems to me that a lot of user friendly, one size fits all, Linux distros pride themselves on "green"-ness and laptop compatibility in their default set-up. While this may make some sense for typical computer use if you are trying to run the machine at max it is not required. My only nod to power saving is shutting down the monitor.


UncleVom



ID: 725650 · Report as offensive
Brian Silvers

Send message
Joined: 11 Jun 99
Posts: 1681
Credit: 492,052
RAC: 0
United States
Message 725658 - Posted: 13 Mar 2008, 19:47:25 UTC - in response to Message 725504.  

Good work on the cpu throttling.
Just wondered if this subject should be a separate thread. So that one, it is short and two, a new title might bring it to the attention of the casual Linux visitor.

I was just thinking the same thing. Once the results of Gerry's test are in, we really ought to preserve the conclusion (but not WW3!) for the benefit of future Linux experimenters.

I don't think another thread is necessarily the answer, though. Unless it was sticky (and we've got enough of those already), it would soon fall off the front page. Perhaps a candidate for Jim-R's FAQs?

And of course, anyone who recommends Linux in the future without mentioning this little gotcha is going to end up shooting themselves in the foot - potentially some very dis-satified customers out there.


Almost makes me wonder if I wasn't fighting the same thing with my VM experimentation over at Einstein... I never quite understood why the VM had that much overhead.

However, my trial of VMWare Workstation 6.0 has long expired and I also uninstalled VM Player. I do still have one of the two VMs that I created, so I guess that sometime over the next few days I'll install VM Player again and check it out...
ID: 725658 · Report as offensive
Fred W
Volunteer tester

Send message
Joined: 13 Jun 99
Posts: 2524
Credit: 11,954,210
RAC: 0
United Kingdom
Message 725679 - Posted: 13 Mar 2008, 20:57:59 UTC - in response to Message 725566.  

For this one unit, linux took 3.5% longer. Compiler differences and/or other processes running on the linux side could easily explain that.


You don't have to invoke compliler differences for a 3.5% difference between crunch times on individual WU's. Here is a table showing the variance in crunch times at various AR's on my Q6600 running XP Home.

Angle Range	        No of results	Mean time (secs)	Min time (secs)	Max time (secs)	Variance (Max / Min) %
0.001 - 0.05	        915		5240.296		4795.91		6655.61		[b]38.78[/b]
0.05 - 0.12	        18		5216.752		4323.31		6687.7		[b]54.69[/b]
0.12 - 0.225485775        46		4280.681		3312.92		4788.03		[b]44.53[/b]
0.225485775 - 0.4         724		5942.495		5328.58		8259.05		[b]55.00[/b]
0.4 - 0.5	                 1558		5279.893		4480.91		7066.19		[b]57.70[/b]
0.5 - 0.75	        87		4753.137		4222.39		5363.61		[b]27.03[/b]
0.75 - 1.127428904        183		4079.774		3312.97		5029.06		[b]51.80[/b]
1.127428904 - 	        2487		1369.231		941.02		1944.2		[b]106.61[/b]
							
App: 	Q6600 @ 3150MHz+ Crunch3r V2.4 SSSE3		


At 3.5%, I would class the 2 OS's as effectively identical.

F.
ID: 725679 · Report as offensive
Profile Toby
Volunteer tester
Avatar

Send message
Joined: 26 Oct 00
Posts: 1005
Credit: 6,366,949
RAC: 0
United States
Message 725707 - Posted: 13 Mar 2008, 21:57:10 UTC

I would argue that the way linux handles idle processes as it relates to increasing CPU speed is probably better than windows. If a process is set to such a low priority then by definition it doesn't really NEED any CPU time - but it will take whatever the OS decides it can spare. For every application except distributed computing I would say leaving the CPU in a low power state would be preferable, especially considering all the energy conservation efforts being promoted these days. There are utilities you can get for Windows (like AMD's Power Monitor utility) that allow you to change the default behavior kind of like how cpufreq-selector works.
A member of The Knights Who Say NI!
For rankings, history graphs and more, check out:
My BOINC stats site
ID: 725707 · Report as offensive
DeMus
Avatar

Send message
Joined: 5 Jan 08
Posts: 238
Credit: 1,765,862
RAC: 0
Netherlands
Message 726310 - Posted: 15 Mar 2008, 0:11:04 UTC - in response to Message 725707.  

I would argue that the way linux handles idle processes as it relates to increasing CPU speed is probably better than windows. If a process is set to such a low priority then by definition it doesn't really NEED any CPU time - but it will take whatever the OS decides it can spare. For every application except distributed computing I would say leaving the CPU in a low power state would be preferable, especially considering all the energy conservation efforts being promoted these days. There are utilities you can get for Windows (like AMD's Power Monitor utility) that allow you to change the default behavior kind of like how cpufreq-selector works.


At the moment I am trying to intall another Linux distro, instead of Fedora. I have had Ubuntu, Mandriva,Gentoo, Debian, Suse but not one which I can install with everything working. Suse has no driver for my networkcard, the rest has either problems with Grub or with installing S@H.
I will keep on trying and when it's working I will crunch some with settings from replies in this thread to see the results.


______
DeMus


ID: 726310 · Report as offensive
DeMus
Avatar

Send message
Joined: 5 Jan 08
Posts: 238
Credit: 1,765,862
RAC: 0
Netherlands
Message 726536 - Posted: 15 Mar 2008, 14:37:11 UTC - in response to Message 725163.  

I see the same behavior on my laptop running Ubuntu. Idle processes do not cause the kernel to initiate CPU power level change. You can change this behavior with the cpufreq-selector command. To make the CPU run at full speed all the time, just type this command into a terminal:
sudo cpufreq-selector -g performance

This causes the kernel to use the "performance" governor to determine when speedstep/powernow features are enabled. The other governors that are available are powersave, conservative, ondemand and userspace. I believe the default one is "ondemand" which, as the name suggests, keeps the CPU at low speed unless more power is needed. You can also lock the CPU in at a specific frequency with this command:
sudo cpufreq-selector -f 1670000

The 1670000 means 1.67 GHz (my CPU can run at 1, 1.67 or 2 GHz).

I'm sure there is some GNOME/KDE interface to this feature that doesn't require breaking out a terminal but thats the quick and dirty way :)

You can view the speed your CPU is currently running at with this command:
cat /proc/cpuinfo

and looking for the line that starts with "cpu MHz." There is also a GNOME panel widget (or whatever they're called) under the "System & Hardware" section entitled "CPU Frequency Scaling Monitor" that will graphically tell you at what speed your CPU is running.


Hello again,
I finally installed another Linux Distro, Mandriva this time. With the help of a Boinc help page for Mandriva I also managed to install Boinc. It all looks good.

I did try the commands as written above but they don't seem to work here. I get error messages about a wrong governor, or missing arguments. I don't seem to get the right syntax for Mandriva.
The last one: Cat /proc/cpuinfo tells me I have 4 processors (as if I didn't know), all of them running at 2.4 GHz. That last remark is nice, since it used to be 1.6 using Fedora.
I am crunching 4 units now, and the final time will be around 2 1/2 hours each.
Will stay in touch.

______
DeMus


ID: 726536 · Report as offensive
Profile Toby
Volunteer tester
Avatar

Send message
Joined: 26 Oct 00
Posts: 1005
Credit: 6,366,949
RAC: 0
United States
Message 726677 - Posted: 15 Mar 2008, 21:58:12 UTC

Searching around a little it looks like Mandriva might use a program called "kpowersave" (part of KDE) to manage frequency scaling and other power control systems. I haven't used KDE in a while so I'm not sure how it works. But it would seem that your CPU(s) are now running at full speed so I guess you don't really need to worry about it :)
A member of The Knights Who Say NI!
For rankings, history graphs and more, check out:
My BOINC stats site
ID: 726677 · Report as offensive
Fred W
Volunteer tester

Send message
Joined: 13 Jun 99
Posts: 2524
Credit: 11,954,210
RAC: 0
United Kingdom
Message 726678 - Posted: 15 Mar 2008, 22:16:28 UTC

Very few results to go on yet, but this is starting to look interesting...


F.
ID: 726678 · Report as offensive
Profile Geek@Play
Volunteer tester
Avatar

Send message
Joined: 31 Jul 01
Posts: 2467
Credit: 86,146,931
RAC: 0
United States
Message 726684 - Posted: 15 Mar 2008, 22:49:48 UTC - in response to Message 726678.  

Very few results to go on yet, but this is starting to look interesting...


F.


Well thanks for that Fred............

Clicking on the photo opened some kind of girly window, full screen and was impossible to close. Thanks for the junk!

ID: 726684 · Report as offensive
DeMus
Avatar

Send message
Joined: 5 Jan 08
Posts: 238
Credit: 1,765,862
RAC: 0
Netherlands
Message 726685 - Posted: 15 Mar 2008, 22:53:30 UTC - in response to Message 726684.  

Very few results to go on yet, but this is starting to look interesting...


F.


Well thanks for that Fred............

Clicking on the photo opened some kind of girly window, full screen and was impossible to close. Thanks for the junk!



Excuse me Geek, I also opened the picture and it was a normal graph. No junk whatsoever. Don't know what went wrong in your PC but there is nothing wrong with the picture.

______
DeMus


ID: 726685 · Report as offensive
Profile Geek@Play
Volunteer tester
Avatar

Send message
Joined: 31 Jul 01
Posts: 2467
Credit: 86,146,931
RAC: 0
United States
Message 726688 - Posted: 15 Mar 2008, 23:12:31 UTC - in response to Message 726685.  

Very few results to go on yet, but this is starting to look interesting...


F.


Well thanks for that Fred............

Clicking on the photo opened some kind of girly window, full screen and was impossible to close. Thanks for the junk!



Excuse me Geek, I also opened the picture and it was a normal graph. No junk whatsoever. Don't know what went wrong in your PC but there is nothing wrong with the picture.


I clicked on his chart and the image opened up from image shack. BEHIND that window another window opened up, full screen and I could not close it. That window downloaded a bunch of crap that my virus detector fired up on most of it. I finally got task manager to close the offending window and cleaned up the registry and all the rest that **** that downloaded with it.

All I know is that it happened at the same exact instant I opened up his link to image shack. Probably some trash provided by the Image Shack and not from Fred's link.


ID: 726688 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · Next

Message boards : Number crunching : Windows vs. Linux


 
©2018 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.