High core temperatures - is there something wrong?

Questions about other topics - please check if your question fits better in another category before posting here
Forum rules
Before you post read how to get help. Topics in this forum are automatically closed 6 months after creation.
Locked
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

High core temperatures - is there something wrong?

Post by powerhouse »

I don't know what's going on but recently my core temperatures are much higher than they used to be (using sensors applet / lm-sensors).

Rig 1 - i7 3930k CPU with 3.2.0-30-generic kernel running on a Xen hypervisor:

Average idle core temps are ~36-43C, averaging around 40C
Running Prime95 torture with all cores at 100%: 68-78C, averaging ~75C, with a Vcore of 1.26V
NO overclocking !!!

Rig 2 - Core 2 duo E6420 CPU with 3.2.0-31-generic kernel (no Xen, just plain vanilla LM13 Mate installation)

Average idle core temps are ~45C
Minor load: 50-55C
Haven't been able to run Prime95 on that machine, yet.

I swear the CPU temperatures were significantly lower when running LM9 on the second (Core 2 duo) rig. I also had lower temperatures when I first installed LM13 some 3-4 months ago on the i3930K rig, with the kernel that first shipped with LM13.

Is there any chance that coretemp readings are incorrect? I'm a bit worried about the high temperatures, and they don't make much sense considering the cooling I got (Noctua NH-D14 for the 3930K).

It seems strange that CPU temperatures should rise with installing a newer kernel/apps, as has been the case with the second rig (Core 2 duo), and possibly the 3930K as well.

I did remove the heatsinks from both CPUs, cleaned and reseated them with new thermal paste that came with the Noctua cooler. It's unlikely an issue with the hardware.

What seems weird is that my CPU fans on the first rig (3930K with Noctua NH-D14 cooler) read around 580 and 640 for idle load, and only 880-930 when under full load. The Noctua fans can go up to 1200/1300 and I sure don't use the Low Noise Adapter.

I wonder who's controlling the fan speed? The BIOS? Or Linux? Seems like the BIOS is in charge here.

Anyone else with strange core temperature readings as of lately?
Last edited by LockBot on Wed Dec 28, 2022 7:16 am, edited 1 time in total.
Reason: Topic automatically closed 6 months after creation. New replies are no longer allowed.
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

OK, did some experimentin'.

I set the CPU fans to 'manual' in BIOS, and selected the following settings:

CPU max temp: 70C
CPU min temp: 20C
Fan max cycle (or so): 100%
Fan min cycle (?): 20%

Fans are running higher now, though not hitting their max:
idle: ~730-780rpm @ CPU temp ~42C
100% load (Prime95): ~950-1030rpm @ CPU temp 72C

I also removed the top cover to "feel" the heat. Can't say the Noctua cooling fins get hot, they are just a little warm. The memory heat spreaders are another story - they literally burn through your skin. Voltage regulators are somewhere in between, not too hot.

When doing the BIOS fan adjustments I also noted the CPU and MB temperatures:

CPU: 41C
MB: 43C

Both temperatures match the idle temperature readings under Linux.

Well, it's been a hot day today (34-36C) and ambient temperature in the room should be around 27-28C.

I still don't understand why the temperature (under lm-sensors/coretemp) went up to 72C, because if the motherboard was reading the same temperature it should have given the PWM controlled CPU fans 100% duty cycle and they should have reached at least 1080-1170rpm (rated rpm -10%). Since CPU core temps were always above 70C (highest value) the fans should have been blowing at maximum speed.
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

Next experiment: I set the CPU fans as follows:
CPU max temp: 50C
CPU min temp: 20C
Fan max duty cycle: 100%
Fan min duty cycle: 20%

This means the BIOS should run the CPU fans at maximum speed when CPU temperature reaches 50C. Here the results:
idle: ~790-835rpm @ CPU temp ~42C
100% load (Prime95): ~1290-1320rpm @ CPU temp 72C

Right now I'm running Prime95 with the following BIOS settings:
CPU max temp: 60C
CPU min temp: 20C
Fan max duty cycle: 100%
Fan min duty cycle: 20%

and get these results:
100% load: 1230-1260rpm @ CPU temp 73C.

Here is how I can explain this:

Linux lm-sensors (coretemp module) reads a different (higher) temperature from what the motherboard reads. At 100% CPU load the motherboard reads a CPU temperature somewhere between 50-60C, but Linux' lm-sensors displays core temperatures well above 70C. There is a difference of at least 15C, probably closer to 20C.

I have a nice system (see signature) and though it might sound like overkill for many, I actually am able to use its power. 100% CPU load is no rarity and since I often stress the CPU to maximum load (not with a benchmark/stress test tool but real applications) I really don't like to see CPU temps above 70C.

Should I trust the motherboard manufacturer (Asus) that they can read the CPU temps properly and that their BIOS fan control software works as advertised?

Or should I believe the lm-sensors / coretemp readings and worry? Or install liquid nitrogene cooling to be on the save side?
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
homerscousin

Re: High core temperatures - is there something wrong?

Post by homerscousin »

I haven't played with anything related to fan speed or anything cooling related because I don't have a need. One of the first things I noticed when I built my latest rig is how much cooler it was compared to the Pentium D. My power supply used to feel warm putting my hand on the enclosure. Now, it feels like any metal surface, very cool as it pulls the warmth from my hand. I just now touched my mem sticks. They are both cool. Conky reports my 4 cores at 21-22 C, N.B. at 35.5 and hdd at 26.5. Ambient temp is 22 C. CPU use while typing this is 2% at idle speed- 1.6 Ghz.

My CPU is not oc'd and I use the on chip graphics and the stock Intel heatsink and fan. I have no side covers on my case, not done with that yet, and have a 120 mm fan mounted to blow onto the mb. And it is running at 5v. That's about it. I see your vcore is 1.26v? That seems very high. Mine is close to 1.0v. Typed 'sensors' at terminal just now. Vcore is .87v. Even during the 90+ days a few weeks ago core temps were in the low 30s.

Check that Vcore. I didn't think any Ivy Bridge chip ran that high unless someone changed it for overclocking purposes.
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

@homerscousin: Thanks for the reply and reference data! This is helpful.

I'm not using Ivybridge - my Intel i7 3930K is a Sandybridge-E design. They are power-hungry, with a TDP of 130W :shock: .

However, I'm also using one of the top air coolers money can buy. The chassis - a Corsair 500R - comes with plenty of fans installed, so I really don't see a reason why I should get high temps, at least not under "normal" operation like web browsing etc. I'm currently in the high 30C/low 40C with Firefox and a pdf file open. I also checked my CPU load and there are no processes that would give reason to increased temp readings.

My memory is hot even under idle load, and it gets real hot when running Prime95. But that is somewhat expected - I have 32GB of 1600MHz DIMMs using all 8 slots.

My Vcore now is 0.84V. Only when running all cores at or near 100% do I reach 1.26V, which isn't so bad for this CPU.

The CPU temps I mentioned before are always based on the highest individual core temp. Even under full load of all 6 cores, temperature between cores varies up to 6C.

Even the high ambient temperature we have now doesn't explain the high CPU temps I get with two different PCs. My old Core2 PC gives now CPU temp readings of 50C+. I hardly ever reached these temps before upgrading the OS, even during stress test.

Which kernel version do you run? Can you post:

Code: Select all

uname -r
And please also let me know which lm-sensors version you got:

Code: Select all

sensors -v
For reference, mine are:
  • 3.2.0-31-generic
    sensors version 3.3.1 with libsensors version 3.3.1
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
homerscousin

Re: High core temperatures - is there something wrong?

Post by homerscousin »

I just checked back so here is kernel: 3.2.0-29, the default and sensors: sensors version 3.3.1 with libsensors version 3.3.1
I thought any Intel i5 or i7 3xxx was ivy, but I don't know em all. Is your CPU 32 or 22 nm?. Ok, so Vcore looks normal at idle.
I don't know. I could believe that 6 core naturally running a little warmer than a 4, but my 2 memory stix aren't even close to warm. But, I have yet to do anything remotely stressful.
How do you monitor temps and cpu use? Did you install Conky? Maybe google high temps with your motherboard?
I'll probably check back tomorrow.
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

@homerscousin: Thanks for checking the kernel release and lm-sensors.

I will try with the 3.2.0-29 or an older 3.2 kernel release to see if that makes a difference.

My memory sticks are so close together that they literally touch each other. I have 2 banks with 4 DIMMs in each bank, altogether 8 DIMMs installed. I don't think that this compares to two DIMMs. I wanted to install 8GB DIMMs to populate only 4 slots, but they weren't on stock here. I had enough issues getting the memory to work - it cost me nearly 2 months - so I was really happy when the computer lab had these Kingston DIMMs that did work flawlessly even under extended stress tests.

I use the hardware monitoring applet in the task bar, which uses lm-sensors and specifically coretemp to read the core temperatures and fan speeds. I installed conky but this gives exactly the same results (conky also uses lm-sensors and the coretemp module).

I suspect there is a kernel problem, or coretemp issue.
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
wayne128

Re: High core temperatures - is there something wrong?

Post by wayne128 »

hi powerhouse,

you have such a high end rig.. how nice..

mine is simple, Dell desktop, just intel i3, two core plus two virtual core.
I opened up side casing, add one fan and that is all, I think i3 is quite cool even running full throttle.
this is already 2+ year old.

it used to run cool, on kernel 2.6.32 -2.6.35
from kernel 2.6.38 onwards they cooked it .
and win7 run real cool, something like 36degc when Linux OS 38-60 depending on distros and kernel.

many post on kernel issue from 2.6.38.. I think they finally resolved on 3.3 level, but then my own testing said some 3.3 level are still running hot.

you might want to test very old kernel 2.6.32 and very new kernel 3.4 level for your own comparison.

for quite a while I have been using the Jupiter to run my Linux OS with either power on demand or power safe mode.. all because I do not think I want to see a 68degC core temp when I can run it with 38degC

so, for normal running with just browsers ( many tabs), the cpu is usually at low end, like 1.2G ( range is 1.20 to 2.93G), that keep cpu core on cool side.


anyway, just paste some info here, kernel is 3.4.10, with lm_sensors installed
# sensors -v
sensors version 3.3.0 with libsensors version 3.3.0

Code: Select all

wayne # uname -r
3.4.10-1.fu2012.i686
wayne # sensors
coretemp-isa-0000
Adapter: ISA adapter
Core 0:       +37.0°C  (high = +89.0°C, crit = +105.0°C)
Core 2:       +36.0°C  (high = +89.0°C, crit = +105.0°C)

it8721-isa-0a10
Adapter: ISA adapter
in0:          +0.92 V  (min =  +1.32 V, max =  +0.70 V)  ALARM
in1:          +3.06 V  (min =  +1.48 V, max =  +0.91 V)  ALARM
in2:          +2.04 V  (min =  +1.46 V, max =  +2.59 V)
+3.3V:        +3.36 V  (min =  +2.78 V, max =  +0.36 V)  ALARM
in4:          +3.00 V  (min =  +1.27 V, max =  +2.68 V)  ALARM
in5:          +2.22 V  (min =  +2.65 V, max =  +2.26 V)  ALARM
in6:          +2.22 V  (min =  +2.46 V, max =  +1.12 V)  ALARM
3VSB:         +3.34 V  (min =  +2.93 V, max =  +6.07 V)
Vbat:         +3.05 V  
fan1:        1010 RPM  (min =   21 RPM)
fan2:         814 RPM  (min =   10 RPM)
temp1:       -128.0°C  (low  = -70.0°C, high = +47.0°C)  sensor = thermal diode
temp2:         -8.0°C  (low  = -74.0°C, high = -77.0°C)  sensor = thermistor
temp3:        -68.0°C  (low  = +126.0°C, high = +47.0°C)  sensor = disabled
intrusion0:  ALARM
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

wayne128 wrote:hi powerhouse,

you have such a high end rig.. how nice..
Yeah, :) .
wayne128 wrote:mine is simple, Dell desktop, just intel i3, two core plus two virtual core.
I opened up side casing, add one fan and that is all, I think i3 is quite cool even running full throttle.
this is already 2+ year old.

it used to run cool, on kernel 2.6.32 -2.6.35
from kernel 2.6.38 onwards they cooked it .
and win7 run real cool, something like 36degc when Linux OS 38-60 depending on distros and kernel.
So you are saying that the temperature readings under Linux are correct and - depending on the distro/kernel - Linux will run the CPU hot. Now that I read this, I remember having read similar comments a while ago.
wayne128 wrote:many post on kernel issue from 2.6.38.. I think they finally resolved on 3.3 level, but then my own testing said some 3.3 level are still running hot.

you might want to test very old kernel 2.6.32 and very new kernel 3.4 level for your own comparison.
Easier said than done. I don't want to break my current system as it's taken me some time to get it working. I could run a live distro and check. Do you know if kernel 3.4 is already somewhere in the Linux Mint or Ubuntu repos?
wayne128 wrote:for quite a while I have been using the Jupiter to run my Linux OS with either power on demand or power safe mode.. all because I do not think I want to see a 68degC core temp when I can run it with 38degC
Never heard of Juniper. Need to find out more, also if it runs with my 3930K on a Xen hypervisor. If I didn't mention it before, my system is running Xen hypervisor 4.1.2 with Linux Mint 13 Mate 64bit as dom0 and Windows 7 Pro as domU, using VGA passthrough for full graphics acceleration under Windows. I also passed through a bunch of PCI devices and use the GPLPV drivers under Windows to boost I/O performance. Results are quite impressive, and I don't want performance or stability issues.
wayne128 wrote:so, for normal running with just browsers ( many tabs), the cpu is usually at low end, like 1.2G ( range is 1.20 to 2.93G), that keep cpu core on cool side.
Ok, Juniper adjusts the CPU clock. I need to see if Xen has similar features.
wayne128 wrote:anyway, just paste some info here, kernel is 3.4.10, with lm_sensors installed
# sensors -v
sensors version 3.3.0 with libsensors version 3.3.0

Code: Select all

wayne # uname -r
3.4.10-1.fu2012.i686
wayne # sensors
coretemp-isa-0000
Adapter: ISA adapter
Core 0:       +37.0°C  (high = +89.0°C, crit = +105.0°C)
Core 2:       +36.0°C  (high = +89.0°C, crit = +105.0°C)

it8721-isa-0a10
Adapter: ISA adapter
in0:          +0.92 V  (min =  +1.32 V, max =  +0.70 V)  ALARM
in1:          +3.06 V  (min =  +1.48 V, max =  +0.91 V)  ALARM
in2:          +2.04 V  (min =  +1.46 V, max =  +2.59 V)
+3.3V:        +3.36 V  (min =  +2.78 V, max =  +0.36 V)  ALARM
in4:          +3.00 V  (min =  +1.27 V, max =  +2.68 V)  ALARM
in5:          +2.22 V  (min =  +2.65 V, max =  +2.26 V)  ALARM
in6:          +2.22 V  (min =  +2.46 V, max =  +1.12 V)  ALARM
3VSB:         +3.34 V  (min =  +2.93 V, max =  +6.07 V)
Vbat:         +3.05 V  
fan1:        1010 RPM  (min =   21 RPM)
fan2:         814 RPM  (min =   10 RPM)
temp1:       -128.0°C  (low  = -70.0°C, high = +47.0°C)  sensor = thermal diode
temp2:         -8.0°C  (low  = -74.0°C, high = -77.0°C)  sensor = thermistor
temp3:        -68.0°C  (low  = +126.0°C, high = +47.0°C)  sensor = disabled
intrusion0:  ALARM
Thanks for the info!

In summary, you say some kernels and distros between 2.6.38 and 3.3.x (perhaps even 3.4.x) aren't good at throttling the CPU and produce higher CPU core temperatures. Well, I've been trying to show that my system seems to report the wrong core temperatures. I don't think my system's core temps are as high as reported.

Perhaps my assumptions are wrong? Here is my thinking:

1. The BIOS controls the fan speed depending on its CPU temperature reading. This happens with no OS involvement, i.e. outside Linux.
2. I did some fan speed adjustments and tests. When I set my fan speed in the BIOS to "manual", with 70C as maximum CPU heat and fan speed at 100%, I never reach the maximum fan speed (as reported by lm-sensors) even when the CPU core temps are above 70C (~75C, as read by lm-sensors).
3. Only when I adjust the fan settings in the BIOS to have 50C as maximum CPU temperature, again with a fan speed setting at 100%, does the fan speed reach the maximum rated fan speed of the fans (~1200-1300rpm which matches the spec sheets of the fans).
4. I currently have the fan settings at 60C maximum CPU temperature, again 100% fan speed. With that setting, and running Prime95 (100% load on all cores) I get close to the maximum fan speed, but still some 100-200rpm lower than maximum.
This is why I believe that the actual CPU temperature (or better the CPU temperature as read by the BIOS) is some 15-20C lower than the temperature reported by lm-sensors.
5. When I run the fan at 100% speed when reaching 50C CPU temperature, that means when I have maximum cooling at relatively low CPU temperatures, I still get lm-sensors readings of above 70C. If the lm-sensors readings were correct, it would mean that my CPU cooler has a lousy performance and isn't able to cool down the CPU even at maximum cooling performance. There are plenty of tests and CPU cooler comparisons that indicate that my Noctua NH-D14 cooler is among the top coolers and provides cooling similar or even better than water coolers. The temperatures I'm getting from lm-sensors would match a CPU that's been overclocked to around 4.5GHz or above (from a base clock of 3.2GHz). Neither top nor xm top indicate any overclocking, and I doubt that Linux could just overclock the CPU on its own.

So to me it looks like this:
a. Either lm-sensors gets it wrong (temperatures are actually lower than what's displayed),
b. Or the BIOS doesn't know how to read CPU temps correctly (CPU temperatures are actually higher), or doesn't apply the fan settings correctly.

I'll try to get an IR temperature probe to put an end to this.
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
wayne128

Re: High core temperatures - is there something wrong?

Post by wayne128 »

So you are saying that the temperature readings under Linux are correct and - depending on the distro/kernel - Linux will run the CPU hot. Now that I read this, I remember having read similar comments a while ago.
I cannot be sure about that. But I choose to trust those readings.
For me, I use simple hand feeling, when I worked last time I can access to IR thermometer, scopes, lab equipment etc but now I no longer could.
Do you know if kernel 3.4 is already somewhere in the Linux Mint or Ubuntu repos?
you can use distrowatch to check your choice distros and see what default kernel they used.
for example, on LinuxMint, extracted from table, maya had 3.2, lisa 3.0, julia 2.6.35, isodora 2.6.32
Feature 13 maya 12 lisa 11 katya 10 julia 9 isadora
linux (3.6) 3.2 3.0 2.6.38 2.6.35 2.6.32

I do not think kernel 3.4 is in Mint13 repos, but I cannot be too sure because I do not run Mint13.

I knew ubuntu 12.10 beta2 is with kernel 3.5 cos I tested a ubuntu/xubutu12.10beta2.
you can see 3.5 listed in distrowatch

Easier said than done. I don't want to break my current system as it's taken me some time to get it working. I could run a live distro and check.
Never heard of Juniper. Need to find out more, also if it runs with my 3930K on a Xen hypervisor. If I didn't mention it before, my system is running Xen hypervisor 4.1.2 with Linux Mint 13 Mate 64bit as dom0 and Windows 7 Pro as domU, using VGA passthrough for full graphics acceleration under Windows. I also passed through a bunch of PCI devices and use the GPLPV drivers under Windows to boost I/O performance. Results are quite impressive, and I don't want performance or stability issues.

Ok, Juniper adjusts the CPU clock. I need to see if Xen has similar features.
I think you do not need to change your installed system.

All you need for comparison test are
1. various distros that has kernel of your choice.
2. a USB stick ( slow running Live or installed) or USB drive ( much faster when run installed Linux OS)
3. or if you just partition out 10G hard disk , installed test distros and perform your comparison tests

For 1, and with kernel 3.4, just use Fuduntu. those I posted earlier were from the latest Fuduntu 2012.04.
Jupiter was installed and works out of the box. Fuduntu developer, Andrew Whatt, is also the developer for Jupiter.
So you should not need to search high or Low. Just get Fuduntu :) and off you can start testing, even running on Live.
http://www.fuduntu.org/blog/2012/10/01/ ... 4-release/
In summary, you say some kernels and distros between 2.6.38 and 3.3.
That was my experience on various distros and reading lots of people who ran into hot temp ( especially on laptop) when they started with 2.6.38 kernel.
If you have not read about them, one good source is from phoronix
http://www.phoronix.com/scan.php?page=n ... &px=OTM3NQ
And there are many possible 'work around' in their reports as well as other people experienced posted in many forum.

Also, I had old test data on Asus laptop using Mint for comparing temp, you can see a jump in temp between LM10 to LM11
http://forums.linuxmint.com/viewtopic.p ... ilit=+asus

(perhaps even 3.4.x) aren't good at throttling the CPU and produce higher CPU core temperatures. Well, I've been trying to show that my system seems to report the wrong core temperatures. I don't think my system's core temps are as high as reported.
Not too sure.

So to me it looks like this:
a. Either lm-sensors gets it wrong (temperatures are actually lower than what's displayed),
b. Or the BIOS doesn't know how to read CPU temps correctly (CPU temperatures are actually higher), or doesn't apply the fan settings correctly.

I'll try to get an IR temperature probe to put an end to this.
[/quote]

Sound like you are running into 'engineering, R&D mode of investigation' :)
You might not be able to conclude technically, because often some of the electrical control might not be linear, they might simply saturate at some point due to hardware limitation, or firmware/software bug..or simply hardware imposed limits be designed ( to prevent thermal runaway).
I do not have proper equipment and did not spend time to investigate with R&D or techician mind set.
So I cannot answer your questions and cannot comment much on your observations with fan control/temp.
If I have a small lab and with proper equipments plus your rig (:)), I can do more research on this topic.
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

wayne128: Thanks for all your feedback and help.

I'm not sure about anything with regard to the temperature readings I get. The only thing I'm sure is that I don't want to fry a $500 CPU :shock: .

I agree with you that anything is possible and that my approach is not scientific and doesn't proof anything.

I will try your suggestions to see.

Thanks again!
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
homerscousin

Re: High core temperatures - is there something wrong?

Post by homerscousin »

Not quite solved yet.
So to me it looks like this:
a. Either lm-sensors gets it wrong (temperatures are actually lower than what's displayed),
b. Or the BIOS doesn't know how to read CPU temps correctly (CPU temperatures are actually higher), or doesn't apply the fan settings correctly.
I think I would guess there first. As far as various kernels resulting in excess heat would'nt that have to be reflected in a corresponding increase in cpu use?
You also run Win 7. Are temp readings similar there?

On a side note, that is a nice rig. Are you by chance doing CAD work on it? I just wondered what you needed the horsepower for other than: I like it, I can afford it so I got it.
Intel will be releasing a 12 core Xeon extreme at the end of next year I read.
Onsen

Re: High core temperatures - is there something wrong?

Post by Onsen »

powerhouse wrote:So to me it looks like this:
a. Either lm-sensors gets it wrong (temperatures are actually lower than what's displayed),
b. Or the BIOS doesn't know how to read CPU temps correctly (CPU temperatures are actually higher), or doesn't apply the fan settings correctly.
I am guessing a.

my sensors: sensors version 3.3.1 with libsensors version 3.3.1
my kernel: 3.2.0-23-generic

Since running Linux Mint 13 / Mate these sensors seemed to be off. Starting up one morning, the BIOS said my CPU had about room temperature (20C-23C) but just about a minute later, when I was in the system I got 60C idle. That just doesn't seem right. I didn't pursue this any further, though.
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

homerscousin wrote:Not quite solved yet.
So to me it looks like this:
a. Either lm-sensors gets it wrong (temperatures are actually lower than what's displayed),
b. Or the BIOS doesn't know how to read CPU temps correctly (CPU temperatures are actually higher), or doesn't apply the fan settings correctly.
I think I would guess there first. As far as various kernels resulting in excess heat would'nt that have to be reflected in a corresponding increase in cpu use?
You also run Win 7. Are temp readings similar there?

On a side note, that is a nice rig. Are you by chance doing CAD work on it? I just wondered what you needed the horsepower for other than: I like it, I can afford it so I got it.
Intel will be releasing a 12 core Xeon extreme at the end of next year I read.
I tried Fuduntu on my old PC, as Wayne128 suggested, and it doesn't make a difference with the core temp readings, at least not something real noticeable (could be a few deg lower, though). Nice OS, by the way.

I can't read the CPU temps under Windows 7, since it is running as a virtual machine (Xen domU). Although I get practically bare-metal performance under the Win7 guest OS, hardware sensors are still out of reach since the CPUs (cores) are virtualized.

I planned and bought the rig for doing photo editing (RAW files) under Windows 7 using Lightroom, Capture NX2, Photoshop, etc. Windows 7 runs as a VM on a Xen hypervisor, with VGA passthrough to have native access to a dedicated graphics card. Photo editing can use a lot of resources, and I hate to upgrade PCs every 2-3 years at most.
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

Onsen wrote:
powerhouse wrote:So to me it looks like this:
a. Either lm-sensors gets it wrong (temperatures are actually lower than what's displayed),
b. Or the BIOS doesn't know how to read CPU temps correctly (CPU temperatures are actually higher), or doesn't apply the fan settings correctly.
I am guessing a.

my sensors: sensors version 3.3.1 with libsensors version 3.3.1
my kernel: 3.2.0-23-generic

Since running Linux Mint 13 / Mate these sensors seemed to be off. Starting up one morning, the BIOS said my CPU had about room temperature (20C-23C) but just about a minute later, when I was in the system I got 60C idle. That just doesn't seem right. I didn't pursue this any further, though.
Thanks! It really looks like lm-sensors gets it wrong. Which CPU do you have?
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
Onsen

Re: High core temperatures - is there something wrong?

Post by Onsen »

AMD / E 350 /1.6GHz / Dualcore.
It's a zbox and the only thing I did different from a basic install was that I used full disk encryption.
pickles99

Re: High core temperatures - is there something wrong?

Post by pickles99 »

Since you are using the Sandy Bridge chipset could it be the "Intel Insider" system heating you up?

From what's publicly available there's a chip with it's own operating system inside that chipset for "Digital Rights" where Intel inspects everything going over the chipset.

Your high load application you are using that's heating everything up could be prompting "Intel Insider" to really work the processor hard going over what you are doing?

I'd put the processor in a different motherboard/chipset. If you want computing power don't use Sandy Bridge. It was made to audit what you're doing. You lose computing power.
powerhouse
Level 6
Level 6
Posts: 1144
Joined: Thu May 03, 2012 3:54 am
Location: Israel
Contact:

Re: High core temperatures - is there something wrong?

Post by powerhouse »

pickles99 wrote:Since you are using the Sandy Bridge chipset could it be the "Intel Insider" system heating you up?

From what's publicly available there's a chip with it's own operating system inside that chipset for "Digital Rights" where Intel inspects everything going over the chipset.

Your high load application you are using that's heating everything up could be prompting "Intel Insider" to really work the processor hard going over what you are doing?

I'd put the processor in a different motherboard/chipset. If you want computing power don't use Sandy Bridge. It was made to audit what you're doing. You lose computing power.
Never heard about that. By the way, I'm not using Intel Sandybridge, but Sandybridge E, which is different in many ways. The CPU has 2011 contacts, for one thing. It's a 6 core CPU (not 4 core as for Sandybridge), etc. etc. etc.

I also doubt that anything is slowing it down - it surely doesn't feel like it.

As far as I know, there are no different motherboard/chipsets available but the X79 chipset.
Subjects of interest: Linux, vfio passthrough virtualization, photography
See my blog on virtualization, including tutorials: https://www.heiko-sieger.info/category/ ... alization/
Locked

Return to “Other topics”