Re: CPU temp's on core 2 duo, should they be significantly different?
On Mon, Jun 23, 2008 at 09:49:16PM -0700, George Hartzell wrote: > > Frank Shute writes: > > [...] > > My top on 7.0 says "CPU states:" not "CPU:" > > > > Are you sure you're running on 2 cores? > > > > dmesg will tell you and top will have a "C" column with 0 or 1 in it. > > > > If you're running on one core, it will explain the temperature > > discrepancy. > > I'm almost certain that I'm running on 2 cores. > > My /usr/bin/top says that it's version: > >top: version 3.5beta12 Same as mine!?! I'm running: $ uname -rms FreeBSD 7.0-RELEASE amd64 > > It does the a C column with 0 and 1. > > I created a big file full of random data and bzip'd it. > > One copy of the file took 20 seconds. Two copies, two processes ran > in 20 seconds each. Three copies, three processes too 32 seconds. > > Tops tells me that some things are running on CPU0 and others are on > CPU1. > > My config file is a copy of GENERIC and includes 'options SMP'. As > the machine boots it talks about finding both CPUS. > > Here's the config file: > > http://shrimp.alerce.com/bluetoo-info/BLUETOO.txt > > Here's the verbose dmesg: > > http://shrimp.alerce.com/bluetoo-info/dmesg.verbose.txt > > and my rc.conf: > > http://shrimp.alerce.com/bluetoo-info/rc.conf.txt > > and here's top: > > last pid: 1650; load averages: 0.00, 0.04, 0.11up > 0+02:43:22 21:47:06 > 51 processes: 1 running, 50 sleeping > CPU: 0.0% user, 0.0% nice, 0.0% system, 0.0% interrupt, 100% idle > Mem: 22M Active, 518M Inact, 200M Wired, 214M Buf, 3189M Free > Swap: 4063M Total, 4063M Free > > PID USERNAMETHR PRI NICE SIZERES STATE C TIME WCPU COMMAND > 861 root 1 440 5688K 1148K select 1 0:01 0.00% powerd >1336 hartzell 1 440 33756K 4608K select 0 0:00 0.00% sshd > 980 root 1 440 73860K 7192K select 1 0:00 0.00% httpd > 854 root 1 440 9432K 2284K select 1 0:00 0.00% ntpd >1338 hartzell 1 200 10100K 3060K pause 1 0:00 0.00% tcsh > 921 root 1 80 4600K 972K nanslp 1 0:00 0.00% svscan >1019 root 1 440 10696K 3868K select 1 0:00 0.00% sendmail > 900 root 1 440 13416K 2772K select 1 0:00 0.00% nmbd >1104 hartzell 1 50 10100K 2752K ttyin 0 0:00 0.00% tcsh > 943 dnscache 1 440 5624K 2368K select 1 0:00 0.00% dnscache >1333 root 1 40 33756K 4544K sbwait 1 0:00 0.00% sshd > 733 root 1 440 5688K 1368K select 1 0:00 0.00% syslogd > 942 root 1 440 6624K 1560K select 1 0:00 0.00% atalkd > 971 avahi 1 440 15652K 2580K select 1 0:00 0.00% > avahi-daemon > 804 root 1 960 4604K 1424K select 0 0:00 0.00% nfsd >1092 root 1 80 20440K 1896K wait 1 0:00 0.00% login > > g. Well, it certainly seems that you're running on 2 cores so that blows that theory out of the water :) My next theory is that cpu0 is reporting too high a figure because it's got a busted or miscalibrated thermistor (or whatever they use). My machine reports cpu core temps of 22 & 24 respectively. That's hovering about room temperature with powerd enabled and a virtually idle machine. For the record, I've got a Core 2 Duo E6550 2.33GHZ. Another possibility, is that coretemp has a bug in it triggered by your particular CPU. I think the broken temp sensor is more likely though. I don't know if your BIOS records the core temps. If not, it will probably record the CPU temp in which case compare with your coretemp temperatures. That may or may not cast some light on things and whether you have to worry about the machine shutting down due to too high a CPU temperature being erroneously recorded. Regards, -- Frank Contact info: http://www.shute.org.uk/misc/contact.html ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: CPU temp's on core 2 duo, should they be significantly different?
Frank Shute writes: > [...] > My top on 7.0 says "CPU states:" not "CPU:" > > Are you sure you're running on 2 cores? > > dmesg will tell you and top will have a "C" column with 0 or 1 in it. > > If you're running on one core, it will explain the temperature > discrepancy. I'm almost certain that I'm running on 2 cores. My /usr/bin/top says that it's version: top: version 3.5beta12 It does the a C column with 0 and 1. I created a big file full of random data and bzip'd it. One copy of the file took 20 seconds. Two copies, two processes ran in 20 seconds each. Three copies, three processes too 32 seconds. Tops tells me that some things are running on CPU0 and others are on CPU1. My config file is a copy of GENERIC and includes 'options SMP'. As the machine boots it talks about finding both CPUS. Here's the config file: http://shrimp.alerce.com/bluetoo-info/BLUETOO.txt Here's the verbose dmesg: http://shrimp.alerce.com/bluetoo-info/dmesg.verbose.txt and my rc.conf: http://shrimp.alerce.com/bluetoo-info/rc.conf.txt and here's top: last pid: 1650; load averages: 0.00, 0.04, 0.11up 0+02:43:22 21:47:06 51 processes: 1 running, 50 sleeping CPU: 0.0% user, 0.0% nice, 0.0% system, 0.0% interrupt, 100% idle Mem: 22M Active, 518M Inact, 200M Wired, 214M Buf, 3189M Free Swap: 4063M Total, 4063M Free PID USERNAMETHR PRI NICE SIZERES STATE C TIME WCPU COMMAND 861 root 1 440 5688K 1148K select 1 0:01 0.00% powerd 1336 hartzell 1 440 33756K 4608K select 0 0:00 0.00% sshd 980 root 1 440 73860K 7192K select 1 0:00 0.00% httpd 854 root 1 440 9432K 2284K select 1 0:00 0.00% ntpd 1338 hartzell 1 200 10100K 3060K pause 1 0:00 0.00% tcsh 921 root 1 80 4600K 972K nanslp 1 0:00 0.00% svscan 1019 root 1 440 10696K 3868K select 1 0:00 0.00% sendmail 900 root 1 440 13416K 2772K select 1 0:00 0.00% nmbd 1104 hartzell 1 50 10100K 2752K ttyin 0 0:00 0.00% tcsh 943 dnscache 1 440 5624K 2368K select 1 0:00 0.00% dnscache 1333 root 1 40 33756K 4544K sbwait 1 0:00 0.00% sshd 733 root 1 440 5688K 1368K select 1 0:00 0.00% syslogd 942 root 1 440 6624K 1560K select 1 0:00 0.00% atalkd 971 avahi 1 440 15652K 2580K select 1 0:00 0.00% avahi-daemon 804 root 1 960 4604K 1424K select 0 0:00 0.00% nfsd 1092 root 1 80 20440K 1896K wait 1 0:00 0.00% login g. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: CPU temp's on core 2 duo, should they be significantly different?
On Monday 23 June 2008 16:06:20 George Hartzell wrote: > DA Forsyth recently mentioned the coretemp driver, which fetches the > core temperatures for Core 2 Duo chips. > > I'm in the middle of building up a Shuttle SG31G2 (7-STABLE) and > loaded the driver to see what it told me. > > I've noticed that cpu.0 is consistently hotter than cpu.1, even on an > unloaded machine. Is that because that core's doing housekeeping work > whilst the other is truly idle? > > dev.cpu.0.temperature: 44 > dev.cpu.1.temperature: 29 > > If I background a pair of "dd if=/dev/urandom of=/dev/null" so that > the cpu's are busy, both go up but cpu.0 stays hotter. > > I'm asking because I'm worried that this could be a sign that I didn't > get the heatsink goop spread out sufficiently well > > Thanks, > > g. For what is worth .. my readings: [EMAIL PROTECTED]:~ # kldload coretemp [EMAIL PROTECTED]:~ # sysctl kern.version kern.version: FreeBSD 7.0-RELEASE #0: Sun Feb 24 19:59:52 UTC 2008 [EMAIL PROTECTED]:/usr/obj/usr/src/sys/GENERIC [EMAIL PROTECTED]:~ # sysctl hw.model hw.model: Intel(R) Core(TM)2 Duo CPU E4500 @ 2.20GHz [EMAIL PROTECTED]:~ # sysctl dev.cpu.0.temperature; sysctl dev.cpu.1.temperature; date "+%H:%M:%S" dev.cpu.0.temperature: 25 dev.cpu.1.temperature: 24 00:08:29 [EMAIL PROTECTED]:~ # dd if=/dev/urandom of=/dev/null & [1] 5482 [EMAIL PROTECTED]:~ # date "+%H:%M:%S" 00:08:48 [EMAIL PROTECTED]:~ # sysctl dev.cpu.0.temperature; sysctl dev.cpu.1.temperature; date "+%H:%M:%S" dev.cpu.0.temperature: 36 dev.cpu.1.temperature: 36 00:10:39 [EMAIL PROTECTED]:~ # j [1] + 5482 Running dd if=/dev/urandom of=/dev/null [EMAIL PROTECTED]:~ # date "+%H:%M:%S" 00:11:13 [EMAIL PROTECTED]:~ # sysctl dev.cpu.0.temperature; sysctl dev.cpu.1.temperature; date "+%H:%M:%S" dev.cpu.0.temperature: 40 dev.cpu.1.temperature: 37 00:11:38 [EMAIL PROTECTED]:~ # sysctl dev.cpu.1.temperature ; sysctl dev.cpu.1.temperature ; date "+%H:%M:%S" dev.cpu.1.temperature: 35 dev.cpu.1.temperature: 35 00:13:58 [EMAIL PROTECTED]:~ # j; date "+%H:%M:%S" [1] + 5482 Running dd if=/dev/urandom of=/dev/null 00:14:20 [EMAIL PROTECTED]:~ # sysctl dev.cpu.1.temperature ; sysctl dev.cpu.1.temperature ; date "+%H:%M:%S" dev.cpu.1.temperature: 39 dev.cpu.1.temperature: 39 00:14:30 [EMAIL PROTECTED]:~ # sysctl dev.cpu.1.temperature ; sysctl dev.cpu.1.temperature ; date "+%H:%M:%S" dev.cpu.1.temperature: 37 dev.cpu.1.temperature: 37 00:14:57 [EMAIL PROTECTED]:~ # fg dd if=/dev/urandom of=/dev/null ^C29752816+0 records in 29752816+0 records out 15233441792 bytes transferred in 378.928688 secs (40201342 bytes/sec) [EMAIL PROTECTED]:~ # Hope it helped :) -- Blessings Gonzalo Nemmi ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: CPU temp's on core 2 duo, should they be significantly different?
On Mon, Jun 23, 2008 at 01:52:07PM -0700, George Hartzell wrote: > > Josh Carroll writes: > > [...] > > I'd recommend taking the heat sink off and seeing how the thermal > > grease is spread on the CPU's head spreader and on the heatsink > > itself. If it looks lopsided or extremely thick on one side of the CPU > > package or extremely thin (to the point where you can still see the > > sheen of the heatsink or heat spreader), then re-mount the heatsink > > and try to make sure it's evenly distributing the pressure down on the > > CPU package. > > This is a Shuttle XPC box. I pulled the heatsink/cooler assembly and > there didn't seem to be any obvious asymmetries in how the the grease > was distributed. I swirled it around a bit, reassembled, and am > seeing the same kind of spreads. > > Here's the machine pretty much idle > > dev.cpu.0.temperature: 44 > dev.cpu.1.temperature: 28 > > Where top says: > > last pid: 1217; load averages: 0.02, 0.51, 0.43up > 0+00:14:57 13:49:47 > 52 processes: 1 running, 51 sleeping > CPU: 0.0% user, 0.0% nice, 0.0% system, 0.2% interrupt, 99.8% idle > Mem: 22M Active, 13M Inact, 95M Wired, 1788K Cache, 15M Buf, 3797M Free > Swap: 4063M Total, 4063M Free My top on 7.0 says "CPU states:" not "CPU:" Are you sure you're running on 2 cores? dmesg will tell you and top will have a "C" column with 0 or 1 in it. If you're running on one core, it will explain the temperature discrepancy. > > A could of dd if=/dev/urandom etc... quickly pushes it up, but the > delta remains: > > dev.cpu.0.temperature: 51 > dev.cpu.1.temperature: 39 > > Top says: > > last pid: 1243; load averages: 0.98, 0.65, 0.48up > 0+00:16:07 13:50:57 > 54 processes: 3 running, 51 sleeping > CPU: 0.4% user, 0.0% nice, 92.5% system, 0.0% interrupt, 7.1% idle > Mem: 22M Active, 13M Inact, 95M Wired, 1788K Cache, 15M Buf, 3797M Free > Swap: 4063M Total, 4063M Free > -- Frank Contact info: http://www.shute.org.uk/misc/contact.html ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: CPU temp's on core 2 duo, should they be significantly different?
Josh Carroll writes: > [...] > I'd recommend taking the heat sink off and seeing how the thermal > grease is spread on the CPU's head spreader and on the heatsink > itself. If it looks lopsided or extremely thick on one side of the CPU > package or extremely thin (to the point where you can still see the > sheen of the heatsink or heat spreader), then re-mount the heatsink > and try to make sure it's evenly distributing the pressure down on the > CPU package. This is a Shuttle XPC box. I pulled the heatsink/cooler assembly and there didn't seem to be any obvious asymmetries in how the the grease was distributed. I swirled it around a bit, reassembled, and am seeing the same kind of spreads. Here's the machine pretty much idle dev.cpu.0.temperature: 44 dev.cpu.1.temperature: 28 Where top says: last pid: 1217; load averages: 0.02, 0.51, 0.43up 0+00:14:57 13:49:47 52 processes: 1 running, 51 sleeping CPU: 0.0% user, 0.0% nice, 0.0% system, 0.2% interrupt, 99.8% idle Mem: 22M Active, 13M Inact, 95M Wired, 1788K Cache, 15M Buf, 3797M Free Swap: 4063M Total, 4063M Free A could of dd if=/dev/urandom etc... quickly pushes it up, but the delta remains: dev.cpu.0.temperature: 51 dev.cpu.1.temperature: 39 Top says: last pid: 1243; load averages: 0.98, 0.65, 0.48up 0+00:16:07 13:50:57 54 processes: 3 running, 51 sleeping CPU: 0.4% user, 0.0% nice, 92.5% system, 0.0% interrupt, 7.1% idle Mem: 22M Active, 13M Inact, 95M Wired, 1788K Cache, 15M Buf, 3797M Free Swap: 4063M Total, 4063M Free g. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: {Spam?} Re: CPU temp's on core 2 duo, should they be significantly different?
At 02:24 PM 6/23/2008, Josh Carroll wrote: > Not sure if the core duos work the same as older 2 CPU and 4 CPU > motherboards, but there are some BIOS functions that always use the first > CPU. So you never get true SMP because the hardware uses the first CPU more > to service interrupts. True, but interrupt handling and minimal background processing should not cause a core to be 15 C hotter. I guess the original poster can mention the load on both cores (or post a top snapshot) so we can see if there is some load on the system. It can cause it to be hotter because the first CPU is servicing all motherboard hardware like the ethernet, video, etc. It is usually the ethernet that causes the most cpu activity. -Derek -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: CPU temp's on core 2 duo, should they be significantly different?
> Not sure if the core duos work the same as older 2 CPU and 4 CPU > motherboards, but there are some BIOS functions that always use the first > CPU. So you never get true SMP because the hardware uses the first CPU more > to service interrupts. True, but interrupt handling and minimal background processing should not cause a core to be 15 C hotter. I guess the original poster can mention the load on both cores (or post a top snapshot) so we can see if there is some load on the system. Josh ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: CPU temp's on core 2 duo, should they be significantly different?
At 02:06 PM 6/23/2008, George Hartzell wrote: DA Forsyth recently mentioned the coretemp driver, which fetches the core temperatures for Core 2 Duo chips. I'm in the middle of building up a Shuttle SG31G2 (7-STABLE) and loaded the driver to see what it told me. I've noticed that cpu.0 is consistently hotter than cpu.1, even on an unloaded machine. Is that because that core's doing housekeeping work whilst the other is truly idle? dev.cpu.0.temperature: 44 dev.cpu.1.temperature: 29 If I background a pair of "dd if=/dev/urandom of=/dev/null" so that the cpu's are busy, both go up but cpu.0 stays hotter. I'm asking because I'm worried that this could be a sign that I didn't get the heatsink goop spread out sufficiently well Thanks, g. Not sure if the core duos work the same as older 2 CPU and 4 CPU motherboards, but there are some BIOS functions that always use the first CPU. So you never get true SMP because the hardware uses the first CPU more to service interrupts. -Derek -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
Re: CPU temp's on core 2 duo, should they be significantly different?
> I've noticed that cpu.0 is consistently hotter than cpu.1, even on an > unloaded machine. Is that because that core's doing housekeeping work > whilst the other is truly idle? > > dev.cpu.0.temperature: 44 > dev.cpu.1.temperature: 29 I notice some differences on my quad-core (Q6600) CPU, too: dev.cpu.0.temperature: 35 dev.cpu.1.temperature: 34 dev.cpu.2.temperature: 27 dev.cpu.3.temperature: 30 The differences also stay around the same when I yes > /dev/null 4 times to load up each core: dev.cpu.0.temperature: 47 dev.cpu.1.temperature: 45 dev.cpu.2.temperature: 38 dev.cpu.3.temperature: 42 The discrepancy isn't as much as in your case, though. 15 C is pretty significant. It could be that either your heat spreader or heat sink are concave or convex causing one of the cores to get hotter. On my dual-core box, here are the idle temps: dev.cpu.0.temperature: 33 dev.cpu.1.temperature: 31 and under load: dev.cpu.0.temperature: 48 dev.cpu.1.temperature: 46 I'd recommend taking the heat sink off and seeing how the thermal grease is spread on the CPU's head spreader and on the heatsink itself. If it looks lopsided or extremely thick on one side of the CPU package or extremely thin (to the point where you can still see the sheen of the heatsink or heat spreader), then re-mount the heatsink and try to make sure it's evenly distributing the pressure down on the CPU package. Josh ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"
CPU temp's on core 2 duo, should they be significantly different?
DA Forsyth recently mentioned the coretemp driver, which fetches the core temperatures for Core 2 Duo chips. I'm in the middle of building up a Shuttle SG31G2 (7-STABLE) and loaded the driver to see what it told me. I've noticed that cpu.0 is consistently hotter than cpu.1, even on an unloaded machine. Is that because that core's doing housekeeping work whilst the other is truly idle? dev.cpu.0.temperature: 44 dev.cpu.1.temperature: 29 If I background a pair of "dd if=/dev/urandom of=/dev/null" so that the cpu's are busy, both go up but cpu.0 stays hotter. I'm asking because I'm worried that this could be a sign that I didn't get the heatsink goop spread out sufficiently well Thanks, g. ___ freebsd-questions@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-questions To unsubscribe, send any mail to "[EMAIL PROTECTED]"