Extreme heat increase after upgrading to 2.6.35

tavianator · 2010-08-23 17:48:11

Okay fine, you're assuming that coretemp gets it right. From the sounds of it, coretemp reads the DTS directly. Tjunction is only read directly on recent CPUs (core i[357], not Core 2 or previous). I assume that it gets the Tjunction value from Windows or the BIOS on older CPUs, so the fault could lie with Windows, coretemp, or Linux.

If there is a bug in Linux, I seriously doubt that it's actually raised the temperature by 15 degrees. More likely, it's just getting the Tjunction value wrong. Anyway, if the delta between Tjunction (max cpu temp) and the reported temperature has stayed the same with 2.6.35, then there's no serious problem. If the delta shrunk, it's a serious bug.

lgolebio · 2010-08-23 17:51:46

More likely, it's just getting the Tjunction value wrong

I agree, this might be a problem. Well we have to wait to find out what exactly is the cause

Last edited by lgolebio (2010-08-23 17:52:12)

architech · 2010-08-24 01:34:24

Another Solution that works 100%

Step 1 Enable ACPI
Step 2 Enable cpufreq
Step 3 Enable powerknowd

I went from 72 c idle to 40 c idle . Also I can play flash via npviewer running like hell and the temperature will never go above 62 c
---------------------------------------------
I found the solution/problem

I believe the culprit is/was the vbox kernel module. After recompiling the virtualbox kernel modules and rebooted, my systems temperature dropped almost 35 degrees. I also installed powernowd with the config of "OPTIONS="-u 95 -l 65", for additional temperature control.

----------------------------------------
tavianator, TigTex, and to who ever else that thinks this is a reporting error, it's not. I can physically tell the difference. My laptop is brushed alluminium, so the difference is highly recognizable. When your laptop runs so hot that the room temperature changes when running a flash video , come talk some logic.

Also, my sh!t started with kernel 2.6.34

On a good day, 68C - 72C Idle!
Nvidia Graphics card underclocked by half
Athlon(tm) X2 Dual-Core underclocked @ 1000Mhz per core + conservative w/ 1.2Ghz max

sensors
k10temp-pci-00c3
Adapter: PCI adapter
temp1:       +68.6°C  (high = +70.0°C)                  

acpitz-virtual-0
Adapter: Virtual device
temp1:       +69.0°C  (crit = +100.0°C) 

lspci | grep -i GeForce\ 9
02:00.0 VGA compatible controller: nVidia Corporation G96 [GeForce 9600M GT] (rev a1)

uname -a
Linux 2.6.34-ARCH #1 SMP PREEMPT Mon Jul 5 22:12:11 CEST 2010 x86_64 AMD Athlon(tm) X2 Dual-Core QL-62 AuthenticAMD GNU/Linux

cpufreq-info 
cpufrequtils 008: cpufreq-info (C) Dominik Brodowski 2004-2009
Report errors and bugs to cpufreq@vger.kernel.org, please.
analyzing CPU 0:
  driver: powernow-k8
  CPUs which run at the same hardware frequency: 0
  CPUs which need to have their frequency coordinated by software: 0
  maximum transition latency: 1000 ns.
  hardware limits: 1000 MHz - 2.00 GHz
  available frequency steps: 2.00 GHz, 1000 MHz
  available cpufreq governors: conservative, performance
  current policy: frequency should be within 1000 MHz and 1000 MHz.
                  The governor "conservative" may decide which speed to use
                  within this range.
  current CPU frequency is 1000 MHz.
analyzing CPU 1:
  driver: powernow-k8
  CPUs which run at the same hardware frequency: 1
  CPUs which need to have their frequency coordinated by software: 1
  maximum transition latency: 1000 ns.
  hardware limits: 1000 MHz - 2.00 GHz
  available frequency steps: 2.00 GHz, 1000 MHz
  available cpufreq governors: conservative, performance
  current policy: frequency should be within 1000 MHz and 1000 MHz.
                  The governor "conservative" may decide which speed to use
                  within this range.
  current CPU frequency is 1000 MHz

Last edited by architech (2010-08-27 20:51:23)

tavianator · 2010-08-24 15:20:56

architech wrote:

tavianator, TigTex, and to who ever else that thinks this is a reporting error, it's not. I can physically tell the difference. My laptop is brushed alluminium, so the difference is highly recognizable. When your laptop runs so hot that the room temperature changes when running a flash video , come talk some logic.
Also, my sh!t started with kernel 2.6.34

Then you don't have the same issue. The OP and others here have a large difference between .34 and .35. If this problem really is reproducible (giant heat increase between .33 and .34), then you should file a kernel bug.

On a good day, 68C - 72C Idle!

Clean out your fans, etc.?

Bl@ster · 2010-08-24 16:31:18

Here is the same problem. Installing kernel26 from [archlinuxfr] helped a bit but not too much; After a while, my desktop keeps freezing for a minute. Now it seems okay, but I'm afraid for another "attack" of machine-hyper-loading

architech · 2010-08-24 20:12:42

-- unwanted --

Last edited by litemotiv (2010-08-28 12:26:17)

bangkok_manouel · 2010-08-24 20:25:44

edit: not relevant anymore

Last edited by bangkok_manouel (2010-09-01 07:17:43)

zippy · 2010-08-24 21:27:14

Tried 35.3 . No difference. I disabled my vbox module, and i did drop 5 degrees (WTF ORACLE).. i knew that oracle deal was bad news. Theres really no reason to autoload vboxdrv if i only use virtualbox once a month or so. Anyways, im still "showing" 10+ degrees hotter and my cpu is still reporting its critical temp is 100C as opposed to 85C since around .29 is when i got this laptop. I find it rediculous that temp reporting has been wrong up until know. "No Iphone users! You don't have a bad signal, weve just been showing you the bars wrong!" I'm 100 times more pissed than the HD power management fiasco happened (and is still happenening apparently.) 15 degrees can be life or death. I hope its just visual.

I tried i8krellm to see if i got different info this dell wont report the temp correctly (but it does a hell of a job with the fans.) My old XPS's bios would didnt like teh linux and i8k was the only way to use it and control the fans properly. If any dell users want to try it, load the i8k module or try "i8k force=1" then get i8kutils or i8krellm with gkrellm.

Well i guess i should dig through intel's website and find the real specs and limits for this chip.

zippy · 2010-08-24 21:35:55

Well apparently my T5800 doesnt exist. Its a renamed T7250, and its tjunction is 100C.

http://ark.intel.com/Product.aspx?id=31 … MHz+FSB%29

Imma boot into 7 and monitor it there for a while.

zippy · 2010-08-24 21:50:43

Well coretemp in windows says my proc is a T5870 and its tjunction is supposed to be 100C as per

http://ark.intel.com/Product.aspx?id=37 … MHz+FSB%29

but is says 85C

thefatprecious · 2010-08-26 21:34:45

I have the same issue of high temps post kernel 2.6.35 on my laptop(intel proc) but not on my desktop (amd proc). Reverting to 2.6.34.3 resolved the issue for me. This along with another issue with 2.6.35 is also reported here: https://bbs.archlinux.org/viewtopic.php?pid=813471

BTW, architech I tried your scenario but it did not improve anything on my system.

Renan Birck · 2010-08-27 03:09:13

I'm seeing the same problem. If I boot on kernel26-lts (which I'm doing at the moment) I get the expected readings (about 48°C on light usage), but on 2.6.35 I get temperatures in excess of 70°C!

It seems to be a kernel problem, as 1) the laptop isn't hot to the touch 2) booting with an older kernel, or with a live-CD, has no problems.

I'm using module "coretemp" (Intel Core 2 Duo temperature sensors) on an Acer laptop with the Intel C2D T5550 processor.

architech · 2010-08-28 10:57:18

-- unwanted --

Last edited by litemotiv (2010-08-28 12:25:30)

lgolebio · 2010-08-28 11:44:36

@Architech, please stop spamming. Your solution doesn't work. You posted 3 times about your solution. It is getting annoying, when you are repeating yourself over and over again. PLEASE STOP because it is not helping.

litemotiv · 2010-08-28 12:25:10

architech, your solution is not relevant for this issue. Stop posting it repeatedly, consider this your last warning.

Igolebio, please use the topic report button if you feel any action is required, instead of confronting other users yourself.

TigTex · 2010-09-01 03:44:40

Usualy newer cpu's have a sensor in each core and one sensor that can be read through ACPI. The ACPI sensor is the "complete" cpu temperature and the others are core reading temperatures. Since the sensors are too close, there should be a little diference in both readings (here the diference is only 1ºC). The acpi is the correct temperature and bios uses that to shutdown down your pc in case of extreme heat, in degrees. The "core" temperature is an internal cpu mechanism used to throttle down your cpu in case of overheat, in DTS (distance to max temperature that you have specified).

for example, hwmonitor (windows yup...) has a config file named hwmonitor.conf and you need to set the tjmax of your cpu. If i set for example tjmax to 50ºC, my core temperature is reported as something between -10 and 10ºC, if i write tjmax = 100, the core temperature is between 40 and 60ºC, the same value as the ACPI sensor; the correct value. This is just to prove that IF the tjmax value was changed, the temperatures are reported in a diferent way (more or less heat). If the tjmax was correct and now is incorrect, it's a kernel bug, report it. if tjmax was the same between older and the 2.6.35 kernel, maybe you have the I/O kernel problem that everyone is complaining about.

Do you want a virtual overmelting cpu? set tjmax to 1000ºC. your temperatures would be something like 940ºC while idling instead of 40ºC (in a 100ºC tjmax cpu), but its just a virtual calculated value: tjmax - dts.

In my opinion, if you dont know your cpu thermal specification just read the acpi temperature (usualy acpitz sensor). If it shows 80ºC, thats your cpu temperature, the core temperature is always something near that.

someone said that i'm wrong but they have a different situation: if your cpu usage (or the load average) is higher with 2.6.35 and that makes your pc hotter, yes, i'm wrong, it's a known issue. read here https://bbs.archlinux.org/viewtopic.php?id=103346

edit: after thinking... maybe there is a problem with the C-states and 2.6.35. Your cpu may be idling at active C-states instead of deep ones and that may cause extra cpu heat. powertop reads the cpu C states. C0 is active cpu (working/with load), and C3 or C6 are the best powersaving modes (less heat). Usualy your cpu idles at C2 and C3 power states. Can someone confirm that your cpu is idling at higher power states (C1 instead of C2 or 3)?

Last edited by TigTex (2010-09-01 04:04:44)

lgolebio · 2010-09-01 06:42:34

@TigTex I think you are right !!

[lukasz@xpsm1530 ~]$ sensors
acpitz-virtual-0
Adapter: Virtual device
temp1:       +48.0°C  (crit = +88.0°C)                  

coretemp-isa-0000
Adapter: ISA adapter
Core 0:      +63.0°C  (high = +100.0°C, crit = +100.0°C)  

coretemp-isa-0001
Adapter: ISA adapter
Core 1:      +63.0°C  (high = +100.0°C, crit = +100.0°C)

My CPU is...

[lukasz@xpsm1530 ~]$ uname -p
Intel(R) Core(TM)2 Duo CPU T5750 @ 2.00GHz

The intel spec shows that max Tjunction is 85*C http://ark.intel.com/Product.aspx?id=33915, and sensors shows 100*C. This might be the problem !!

And also 100*C - 85*C = 15*C !!!! This is exactly why my CPU temp rised about 15*C after upgrade !!! Now tell us how to fix it

Last edited by lgolebio (2010-09-01 06:48:31)

TigTex · 2010-09-01 16:13:30

I really don't know how to fix it But your real cpu temperature is 48ºC and not 63ºC.

According to http://www.mjmwired.net/kernel/Document … n/coretemp

Maybe you can set the "temp1_crit" manualy writing something like this inside /etc/sensors3.conf

chip "coretemp-isa-*"
      set temp1_crit 85

And then reload sensors with

/etc/rc.d/sensors restart

It's only an idea... At the moment I don't have any pc here with a DTS sensor... only pentium 4 and older.

Last edited by TigTex (2010-09-01 16:19:35)

lgolebio · 2010-09-01 17:18:02

Nope, it doesn't work. Maybe we should wait for kernel 2.6.36

Dave158 · 2010-09-01 23:32:29

So I take it this is just a kernel issue?

TigTex · 2010-09-02 20:33:53

careful with double posting dave

If you dont see an increased cpu load and the acpitz sensor shows a normal temperature, yes, it's just a reading bug. Probably there are workarounds, but I don't have any cpu with the coretemp sensor to test...

If the temperature is really higher it might be this bug https://bbs.archlinux.org/viewtopic.php?id=103346

Dave158 · 2010-09-02 23:33:25

I only posted once. Most likely a glitch with my browser sending the data. Sorry. Also, okay. Thanks.

jarryson · 2010-09-14 16:14:38

i have little bit CPU heat increase here.

but i am using BFS instead of CFS, and i did not seen any increase of cpu usage.

before when laptop is idle, the cpu fan wont run, but now it running and running, a little annoying.

Renan Birck · 2010-09-23 00:50:05

Sorry for bumping the thread, but I think I discovered where the problem lies:

This commit: http://git.kernel.org/?p=linux/kernel/g … a24deb09c9 - specially those lines:

+       switch (c->x86_model) {
+       case 0xe:
+       case 0xf:
+       case 0x16:
+       case 0x1a:
+               dev_warn(dev, "TjMax is assumed as 100 C!\n");
+               return 100000;
+               break;

And in fact, "grep model /proc/cpuinfo" give me "model: 15", that is, 0xF, thus returning 100000 (100°C), which is not the right Tjmax.

Anybody that has a "Merom"-based CPU and dual-boots with Windows (I can't dual-boot here, as it would break my entire partition scheme, sorry ) can post a Core Temp screenshot showing the right Tjmax?

Thanks.

uhuu · 2010-11-22 18:49:12

Renan Birck wrote:

Sorry for bumping the thread, but I think I discovered where the problem lies:
This commit: http://git.kernel.org/?p=linux/kernel/g … a24deb09c9 - specially those lines:
+       switch (c->x86_model) {
+       case 0xe:
+       case 0xf:
+       case 0x16:
+       case 0x1a:
+               dev_warn(dev, "TjMax is assumed as 100 C!\n");
+               return 100000;
+               break;
And in fact, "grep model /proc/cpuinfo" give me "model: 15", that is, 0xF, thus returning 100000 (100°C), which is not the right Tjmax.
Anybody that has a "Merom"-based CPU and dual-boots with Windows (I can't dual-boot here, as it would break my entire partition scheme, sorry ) can post a Core Temp screenshot showing the right Tjmax?
Thanks.

Thank you Renan Birck

Replacing the code

+       switch (c->x86_model) {
+       case 0xe:
+       case 0xf:
+       case 0x16:
+       case 0x1a:
+               dev_warn(dev, "TjMax is assumed as 100 C!\n");
+               return 100000;
+               break;

with the code

+       switch (c->x86_model) {
+       case 0xe:
+       case 0xf:
+               dev_warn(dev, "TjMax is assumed as 85 C!\n");
+               return 85000;
+               break;
+       case 0x16:
+       case 0x1a:
+               dev_warn(dev, "TjMax is assumed as 100 C!\n");
+               return 100000;
+               break;

in coretemp driver fixed it for me

Last edited by uhuu (2010-12-04 09:45:36)

Arch Linux

#26 2010-08-23 17:48:11

Re: Extreme heat increase after upgrading to 2.6.35

#27 2010-08-23 17:51:46

Re: Extreme heat increase after upgrading to 2.6.35

#28 2010-08-24 01:34:24

Re: Extreme heat increase after upgrading to 2.6.35

#29 2010-08-24 15:20:56

Re: Extreme heat increase after upgrading to 2.6.35

#30 2010-08-24 16:31:18

Re: Extreme heat increase after upgrading to 2.6.35

#31 2010-08-24 20:12:42

Re: Extreme heat increase after upgrading to 2.6.35

#32 2010-08-24 20:25:44

Re: Extreme heat increase after upgrading to 2.6.35

#33 2010-08-24 21:27:14

Re: Extreme heat increase after upgrading to 2.6.35

#34 2010-08-24 21:35:55

Re: Extreme heat increase after upgrading to 2.6.35

#35 2010-08-24 21:50:43

Re: Extreme heat increase after upgrading to 2.6.35

#36 2010-08-26 21:34:45

Re: Extreme heat increase after upgrading to 2.6.35

#37 2010-08-27 03:09:13

Re: Extreme heat increase after upgrading to 2.6.35

#38 2010-08-28 10:57:18

Re: Extreme heat increase after upgrading to 2.6.35

#39 2010-08-28 11:44:36

Re: Extreme heat increase after upgrading to 2.6.35

#40 2010-08-28 12:25:10

Re: Extreme heat increase after upgrading to 2.6.35

#41 2010-09-01 03:44:40

Re: Extreme heat increase after upgrading to 2.6.35

#42 2010-09-01 06:42:34

Re: Extreme heat increase after upgrading to 2.6.35

#43 2010-09-01 16:13:30

Re: Extreme heat increase after upgrading to 2.6.35

#44 2010-09-01 17:18:02

Re: Extreme heat increase after upgrading to 2.6.35

#45 2010-09-01 23:32:29

Re: Extreme heat increase after upgrading to 2.6.35

#46 2010-09-02 20:33:53

Re: Extreme heat increase after upgrading to 2.6.35

#47 2010-09-02 23:33:25

Re: Extreme heat increase after upgrading to 2.6.35

#48 2010-09-14 16:14:38

Re: Extreme heat increase after upgrading to 2.6.35

#49 2010-09-23 00:50:05

Re: Extreme heat increase after upgrading to 2.6.35

#50 2010-11-22 18:49:12

Re: Extreme heat increase after upgrading to 2.6.35

Board footer