CF-31 thermal management issue

[fleecy sheep]

Hello everyone

Rob suggested to me to ask for opinions around here concerning an issue I have.
After having purchased a new CF-31JCGAX1M (i5-2520M; no ATI) from Rob a couple of months ago (excellent customer sevice btw), I can say that I am overall very pleased with the CF-31.
Things I'm still a bit wary of durability-wise are the living hinges on the small flaps on the right-hand side and the partially very thin floor plates, but time will tell... Oh and the touchpad sucks big time!

The problem I experience with my unit pops up whenever I have to do larger computations (in pure shell or in desktop environment with all 4 CPU threads at 100% load for several minutes), which luckily isn't the case often. While the CPU temperature increases, the processor is throttled down massively and the fan annoyingly spins up and down alternately (pretty stupid regulation if you ask me).

Under light load usage (standard pattern) I usually put the CF-31 into concealed mode, which in my case only consists of the fan being turned off, as it would otherwise be constantly on even while the CF-31 is idle (overall CPU load under linux 2-4% at 800 MHz; each thread between 0-9%), independent of the OS used (win7 and linux).
Temperatures are as follows:
- current ambient temperature
24°C/75°F
- CPU core temperatures under linux during idle in concealed mode
51-55°C/124-131°F
- CPU core temperatures under linux during idle in normal mode (fan on at lowest speed) after another 10 mins
48-52°C/118-126°F

From experience I can say that the difference in temperatures between concealed mode (fan off) and normal mode during standard usage is in the ballpark of 4-5°C with the temperatures in concealed mode obviously being higher.

Last friday evening I wanted to find out how the machine would behave while playing counter-strike and starcraft (both are the more than a decade old versions) under linux with wine (just a test, no "real" game), which my old notebook bought in 2000 could cope with, albeit this was directly in windows then.
The results were quite ashaming...
At first everything was fine while playing CS at 59 fps, but after about 3-4 minutes the frame rate had gradually dropped to 5 fps. With starcraft it was no different (no fps measurement) as the game began to stutter after a short initial period of smooth playing.
Don't get me wrong. I do not intend to use the CF-31 as a gaming machine
But this performance is just plain ridiculous.

Additionally I did a test with truecrypt's benchmark utility (buffer size 1024 MB) under linux in normal mode (fan on). CPU core temperatures were at 78-98°C/172-208°F on both cores and alternating with the fan at an amplitude lenght of about 6-7 sec. While the fan exhibited this siren-like pattern of alternating its speed, the processor frequency was switching between 2501/2500 and 1000 MHz.


The ultimate question is: do I have a dud (i.e. insufficient contact between CPU and heat sink, bad CPU, ...) or did the folks at Panasonic "excel" at designing the heat management?

Cynicism put aside, maybe continuous usage at higher CPU load was not a design goal as the unit otherwise performs flawlessly and smoothly during normal operation.

Another thing I noticed is that after extended uptime under normal use the hard disk temperature reaches up to 50°C/122°F, which is said to be detrimental to life expectancy, although I remember (hopefully correctly) a study done by google which comes to the conclusion that higher hard disk temperatures are not necessarily all evil.

Thanks in advance for your input on this issue!

[ADOR]

I know when Rob had a 31 Mk1 with the ATI card he didn't have any problem playing games that were a lot newer. (Alice 2 for one) I haven't seen anyone with the integrated graphics version of the I5 toughbook really try any games out. With the hardware you have it should be working fine, even the integrated graphics have made some big jumps in performance from the 28/29 days. I have done a little research and have seen other people have had problems with there I5 processors in laptops with speedstep bouncing up and down a lot. Heat seemed to cause it a lot of the threads I read.

Here are a few things to check out to narrow this down, these aren't toughbooks but use the same CPU and Chipset.

Heat, if you have a cooling pad or a fan just try it out to see what it does.

One thing pointed to a charger problem causing the below. Do you have a second charger to test this out?

CPU or GPU heat can cause this and power consumption can also trigger your CPU to start throttling. Many Dell laptops will throttle as soon as you switch to battery power. One problem is that some power adapters do not make good contact so your laptop can intermittently switch from AC to battery power, your CPU will throttle and then it will switch back to AC without you really noticing what happened.

Run ThrottleStop in monitoring mode and run a log file so you can at least see what is going on. Try checking off GPU monitoring in the Options window so you have a record of your GPU too. When you are finished gaming, exit the game, and then exit ThrottleStop and then Copy and Paste the log file somewhere convenient like www.pastebin.com so I can have a look. It might be something obvious.

Whatever the problem is, if running ThrottleStop allows your CPU to run at its rated speed then there is no harm using it. ThrottleStop can not bypass the Intel throttling mechanism. If your CPU gets too hot, it will throttle just as Intel intended it to. All ThrottleStop does is it gives a user a way to make sure that the bios isn't doing something dumb for no reason with their CPU. It keeps the CPU running at the speed that Intel intended it to run at.

ThrottleStop can not be used to lock a Core i5 CPU at the maximum multiplier regardless of what CPU-Z might be telling you. ThrottleStop lets your CPU idle down when not loaded. The C3% and C6% data is proof of that. It lets cores go into the sleep states when they are not needed to save power and to keep the heat down.

A lot of the Dells use clock modulation throttling to slow a CPU down internally. A log file will show this if it is happening. I think chipset clock modulation throttling is the usual problem.

Basically, I booted skyrim, loaded a save and just let it sit there. It start up at 30fps. After about a minute it dropped to 8 fps and stayed there. I let it sit at low fps for awhile then I hit escape to exit the game. The second the menu popped up the framerate shot up to 60fps. And if I exited the menu and went back to the game it would go back to 30fps for about 4-5 seconds then drop to single digits again.

I see two things going on in your log

1) BAT_Mw I'm guessing this is a measurement how much power is being drawn from the battery. If I'm right, you either have a problematic connection to your charger, or the charger is undersized for your laptop (or the laptop itself limits the amount of power it can draw from the charger)


2) CKMOD (clock modulation) starts stepping down from 100.0 to 12.5 within two seconds of your GPU temp reaching 77c. It stays at 12.5 until it's back down to 72c where it steps back up to 100.0. This is where your throttling happens. Jumping to the menu screen doesn require much computation power from the GPU, so you get max frame rate, but it's still throttled so going back to the game you'll see the framerate drop again.


Most people conclude that if it is throttling that it must be throttling for a reason but no, sometimes Dell laptops throttle excessively for no reason at all.

If you set ThrottleStop clock modulation to 100.0% and put a check mark in that box, that will help prevent your laptop from throttling like this. Remember to turn ThrottleStop on and take it out of monitoring mode. I also like to put a check mark in the Set Multiplier box and I increase that to the maximum. Read the ThrottleStop docs and make sure your power plan is set up correctly.

This should help make your laptop usable again. Why some laptops throttle and some don't is anyone's guess. Slight differences in power consumption or heat can make the difference. If you are not sure if you have ThrottleStop setup correctly then post a screen shot of it and post another log file while gaming with ThrottleStop working to correct this problem. Most users find that ThrottleStop makes a significant difference and your M6500 is not the first M6500 that I have heard about with throttling problems.

Make sure BD PROCHOT is not checked as well. That stands for bi-directional processor hot. This signal path allows your GPU to send a signal to your CPU telling it that it is overheating. Blocking this signal can help block the throttling. I am not sure if this Dell laptop model uses this method. Many of the Asus gaming laptops do. Your GPU is not that hot and neither is your CPU so no worries.

Thanks for all the help, I've been using throttlestop for awhile now to get it running normally so I guess I'll just continue doing that. I've tried tweaking the bios settings in every way it allows and updated to the latest bios without any luck . There's an option to turn off speedstep in the bios but turning it off runs the pc at the lowest possible stepping (it even says it will do that in the bios description).

Oh well, thanks so much for your time I'm just glad it wasn't something dumb I was doing haha.

The user found using the throttle stop program would help out in his situation.

http://www.overclock.net/t/1385956/inte ... -underload

Another user on another board started having problems and pulled his mainbattery and cmos for about 10 minutes, cleared all power by cycling the power button and his machine went back to acting like normal.


Here is a start.

[Rob]

Thanks for your help ador!

About the touchpad: keep one thing in kind: the touchpad is NOT a capacitive one like EVERY other laptop out there! Its resistive! This is so when it's -1,000 degrees out and you have gloves on you can still use the touchpad (without having to have lame, thin, stupid, "smartphone compatible" gloves)

Be sure to turn the sensitivity all the way up. If that's not enough then go to toughbooktalk.com/public_downloads and look for the mouse sensitivity fix that one of our guys made.

Thanks!

[fleecy sheep]

Cause for problem narrowed down...
Luckily it's not hardware related.

ADOR's reply made me again tweak the governor settings for the CPU (lower freqency) in linux with only little improvement. But this time this did raise my suspicions.
So I took the time to try out CS in win7 and everything was running smoothly at high fps during the about half an hour long test, after which I immediately ran the truecrypt benchmark tool, which also gave me much higher results than the test with the linux variant.
I had previously thought that the problem (throttling and siren-like fan pattern) had also occured some months ago in win7 (I rarely use it) but my mind seems to have played tricks on me.

This brings me to the conclusion that the culprit is the 3.2 linux kernel, which according to the facts offers inefficient CPU power management. But I can live with that as hopefully some future kernel update will cure this issue.

@Rob: I knew the touchpad is a resistive one and the valid reasons for its use over a capacitative one in the fully rugged toughbooks. It's a PITA to use nonetheless (I had already modified the parameters for the corresponding linux kernel module to make it suck less), although I have to admit that I generally abhor touchpads (the touchscreen is pretty useful though) and absolutely prefer to use a mouse with as many function keys as possible whenever possible as this highly increases workflow. The mouse I use with my CF-31 is a Logitech M555b (bluetooth) and I can fully recommend it, although it unfortunately doesn't have as many buttons as the Logitech MX518 I use with my Desktop (great corded mouse; no longer in production, but successor should be available).

Thanks ADOR, thanks Rob for your help.
I hope I may give something back in the future, i.e. concerning linux issues, as I have an otherwise fully operational dual boot linux/win7 setup on the CF-31.

[ADOR]

We could always use another Linux guy. Jeff is or main one right now. I can tweak a lot in windows but I am still behind the times in Linux.

[ADOR]

What version of linux are you using? I have been doing more reading up on this problem on the linux side trying to narrow it down further.

[fleecy sheep]

Kubuntu 12.04.2 LTS. Kernel version according to Ubuntu nomenclature is 3.2.0-xx, but with backported upstream patches. If I recall correctly, this is the first LTS release with rolling kernel upgrades, not just security patches.

In fact I've seen some indication the problem might have already been fixed in the current upstream kernel 3.10, but I haven't tested it yet, since I've become too lazy to compile my own kernel again and again with every single update and the easierer way of trying it out would have meant to install a pre-compiled mainline testing kernel from the Ubuntu kernel team, who unfortuately don't sign their testing stuff, which to me is a show stopper.

[fleecy sheep]

To celebrate the creation of the Linux subforum (where this thread might be better-placed), I gleefully declare the above problem as no longer existent since running a 3.8-kernel!

And only the best wishes for the new year to all of you