You are not logged in.

#1 2020-07-29 04:59:54

VDmvKcW9JamBSir5fNfehqpG
Member
From: Earth
Registered: 2019-12-30
Posts: 59

[SOLVED] Trying to cool down laptop, undervolt, underclock not working

Today, my laptop overheated and shut down. I need my laptop to actually work, so that is a problem. Information:

* The CPU is an i7-740QM, which looks like it's a Clarksfield/Nehalem, if that's relevant.

* When it shut down, I was in the process of setting up a VM in Virtualbox, and I had about 50 Brave tabs and 10 VSCode windows open, along with a handful of other programs here and there, so I believe it was the CPU and not the GPU that overheated, but they share a cooler in my laptop. The fan was at full speed when it happened.

* The laptop is clean inside and out, was on a desk, and was slightly lifted using erasers because the rubber feet have fallen off years ago.

* I *want* to repaste the laptop, but my mom doesn't let me use my bank account for online purchases and I can't go to the store because of Trump Disease. (And I'm under 18, so getting a bank account outside of the control of a parent is not an option.)

The first thing I thought about trying to do was to try to write a value to the MSR registers to raise the thermal shutdown temperature a bit, like maybe to 130 instead of 100. However, I couldn't find anything in the MSR docs to suggest that this is possible, and the people on the Intel discord seemed kinda horrified, didn't want to help me with that, and were also convinced that it was impossible anyways. I also wanted to try messing with a sensor to falsify the reading, but it turns out that most or all of the sensors are on the die and therefore impossible to tamper with. And someone on the Intel discord thinks my VRMs are being baked and shutting down, which defeating my CPU's thermal protections won't help with. So on to the next plan.

The next thing I wanted to try was undervolting because I do not want to sacrifice performance, but obviously I have to decrease peak power consumption. The wiki pointed me to PHC, as I have a pre-Haswell CPU. (And the tools meant for Haswell and newer just spewed out segfaults and MSR errors, depending on the tool.) I installed the phc-intel package, but right after that is when I started running into problems. The wiki said to run

phc-intel setup

, but that just told me that the only valid options were start, stop, status, and set. I figured I'd try running

phc-intel set

, but that didn't appear to do anything at all. The phc-intel manpage also does not appear to exist, the upstream URL listed on the AUR (http://www.linux-phc.org/) appears to be dead, and the site on the internet archive as of December 2019 does not appear to have any information. So I wasn't able to complete that step. I was told to install mprime and run this script, but that just complained about PHC not being loaded. I tried

phc-intel start

, but it told me to edit /etc/default/phc-intel to set the default VIDs for my CPU. My intuition tells me that these were supposed to come from

phc-intel setup

, which didn't work. And for some reason, search engines refuse to give any information about what these defaults are supposed to be, and I don't want to try random values because CPUs are expensive and can be fried by giving too much voltage, which is a risk that I don't want to take. So on to the next plan.

I figured I'd try underclocking from 1.73 GHz (my CPU doesn't seem to turbo even though it's supposed to) to the next lowest step, 1.6 GHz, because that should still be over 90% of my performance and it seems like my system is only barely going beyond its thermal limits. Unfortunately, when I try to set the max speed using cpupower, it seems to be ignored, regardless of what the governor is set to (userspace, performance, and schedutil are available.)

My last resort is to disable a CPU core, but that's more of a performance hit than I want to take, so I'd like to know if I'm doing something wrong here that's preventing the other methods from working.

Last edited by VDmvKcW9JamBSir5fNfehqpG (2020-07-31 18:08:58)

Offline

#2 2020-07-29 12:06:06

Awebb
Member
Registered: 2010-05-06
Posts: 6,273

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

Explain to your mum the maintenance value of thermal paste. Thermal paste is like 5 bucks. You don't need crushed diamonds in dolphin grease. Thermal pads are also cheap.

Seriously, your CPU is ancient. Thermal paste in Laptops has a lifespan of 2-5 years and is often applied with the least care by manufacturers. It's like the one weak spot of laptops. Not renewing it can cause failures by overheating. Source: I've had my own repair shop with a decade of experience.

Last edited by Awebb (2020-07-29 12:10:07)

Offline

#3 2020-07-29 13:57:12

VDmvKcW9JamBSir5fNfehqpG
Member
From: Earth
Registered: 2019-12-30
Posts: 59

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

My mom doesn't listen to any reasoning on any topic and automatically assumes that her initial idea is correct and no amount of convincing can change that, especially if a teenager or someone else that she considers herself to be above is holding the opposite opinion. (I'm honestly surprised she isn't a flat earther.) In this case, she assumes that because I'm a teenager, I *will* just blow all the money on drugs and vapes and then get arrested. So explaining the importance of thermal paste doesn't work. I'm also not entirely sure how much life this laptop has left. (I'll replace it with a new one when I go to college next year. After that, the current one will replace my raspberry pi as my server, but I don't know how long it will last when doing that), so I'm not entirely convinced that spending money on it is a good idea anyways. Is it really not worth it to even try the software workarounds?

Offline

#4 2020-07-29 21:36:20

ondoho
Member
Registered: 2013-04-30
Posts: 692
Website

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

VDmvKcW9JamBSir5fNfehqpG wrote:

* The CPU is an i7-740QM, which looks like it's a Clarksfield/Nehalem, if that's relevant.

* When it shut down, I was in the process of setting up a VM in Virtualbox, and I had about 50 Brave tabs and 10 VSCode windows open, along with a handful of other programs here and there

a 10 year old quad core... not too bad, but you're asking too much of it.
If your RAM is accordingly limited this probably sent both your CPU and your hard drive (swap) in a frenzy.
- VS code i sheavy
- Brave is heavy (more so with 50 tabs) - use a browser that allows you to restrict javascript, e.g. FF with uMatrix or NoScript.
- VB is heavy - and then it also depends which OS you set up in it.

Do less at a time.

It's hard to believe that you cannot go out and buy thermal paste (it's a long time I was that young, and I wasn't interested in computers back then), but if you're brave you can search the web for ... alternatives.

PS: I see no proof that your laptop does actually overheat, or that that is the reason for the shutdown. Please set up lm_sensors and give us some numbers.

Last edited by ondoho (2020-07-29 21:39:40)

Offline

#5 2020-07-30 04:37:51

VDmvKcW9JamBSir5fNfehqpG
Member
From: Earth
Registered: 2019-12-30
Posts: 59

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

I don't think I'm asking too much of it, as the system was still quite responsive when it overheated. I have 8 GB of RAM. and while swap does get used, it doesn't hammer the SSD that often. I think this is a case of Dell cheaping out and not putting enough cooling for the CPU+GPU combo that they equip the machine with to save 3¢/unit. I was setting up another Arch installation in the VM at the moment of the crash, and I was just mounting the filesystems, so I don't think it was hammering my CPU too badly. I also like really like VSCode and Brave. I cannot completely get rid of Javascript because a lot of sites require it, but I'll consider restricting it.

I saw some alternatives on the web. Mustard and toothpaste seem to be popular. However, I don't want ants to be attracted to my computer, and I'm not convinced they'll outperform even the shit Dell used. Plus, they seem to wear out quickly and I'm too lazy to repaste every week, and they can be conductive and I don't want to risk ruining my motherboard.

After the laptop shut down, it refused to reboot for several minutes until it cooled down significantly, measured by putting my finger on the heat pipe and seeing how long it took to get painful. Upon rebooting my laptop, my BIOS told me that it powered off due to a "System board thermal trip" and gave me a lecture telling me to clean the laptop, give it enough airflow, and contact tech support if that doesn't work. I think that's evidence that *something* overheated. I have lm-sensors installed, but was not monitoring it at the time. Unfortunately, my laptop won't run the fan at full speed and thermal throttles at about 80°C as a result unless I use a utility from the AUR (dell-bios-fan-control), but using that utility to make my fan work correctly causes the CPU temperature seen by lm-sensors to freeze until the AUR utility releases control of the fan. I currently have it in a wrapper script that runs in the background and pins the fan to full speed once the CPU goes above 75°C and releases control once the GPU (because the CPU sensor gets stuck) goes below 65°C, which is acceptable because they share heat pipes. So either I get horrible throttling OR an accurate temperature reading. However, in addition to my sensor labeled as CPU, I have four sensors labeled as Core 0-3. Is it acceptable to monitor these sensors instead? I don't think they get stuck.

Offline

#6 2020-07-30 04:42:00

Awebb
Member
Registered: 2010-05-06
Posts: 6,273

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

ondoho wrote:

but if you're brave you can search the web for ... alternatives.

Brave = stupid in this case.

Offline

#7 2020-07-30 05:52:30

VDmvKcW9JamBSir5fNfehqpG
Member
From: Earth
Registered: 2019-12-30
Posts: 59

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

On idle, my CPU and GPU both run at about 60°C, plus or minus a few degrees, with the fan running on low speed intermittently. After 10 minutes of mprime (I would run more, but I have to go to bed), my CPU reached 87°C and my GPU reached 86°C. Are these temps acceptable or too high?

Offline

#8 2020-07-30 12:46:32

ondoho
Member
Registered: 2013-04-30
Posts: 692
Website

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

VDmvKcW9JamBSir5fNfehqpG wrote:

Are these temps acceptable or too high?

Check your CPU's specs on the 'net, that will tell you.

You should really use your eloquence on your mother, not on us.

But it is conceivable that your old intel CPU requires some special attention.
Maybe you should go pack to post #1 and see if all those things are what you really need (maybe it would work better without interfering, the kernel would just do the right thing?) and if you set them up correctly.

I am sorry, I am a little pampered, I have been using only all-Intel devices for the past 5 years. I would never dream of under/overclocking or faffing around with CPU fans. This sort of stuff Just Works.

Offline

#9 2020-07-30 17:36:10

VDmvKcW9JamBSir5fNfehqpG
Member
From: Earth
Registered: 2019-12-30
Posts: 59

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

The CPU is rated to handle up to 100, and lm-sensors claims that 100 is the critical temperature and 84 is the "high" temperature, so my temperatures after 10 minutes of mprime appear to be in the high range, but don't look dangerous yet. What exactly is meant by "special attention"? Searching for "clarksfield" and "nehalem" on the Arch Wiki shows nothing of interest.

I also discovered that I do have a couple bucks of gift card money on my Amazon account. Can I get away with the cheapest thermal paste I can find, or should I try to get good paste? And does it have a shelf life? I probably won't need it for a few more years, so will it still be good then or will it only be good for this job?

Offline

#10 2020-07-31 09:07:18

ondoho
Member
Registered: 2013-04-30
Posts: 692
Website

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

VDmvKcW9JamBSir5fNfehqpG wrote:

What exactly is meant by "special attention"?

What you wrote in post #1. Pre-Haswell? More recent intel CPU/GPU's "Just Work" on Linux.

And does it have a shelf life?

I just used mine again after years - so yes, it has a shelf life.

Offline

#11 2020-07-31 18:08:18

VDmvKcW9JamBSir5fNfehqpG
Member
From: Earth
Registered: 2019-12-30
Posts: 59

Re: [SOLVED] Trying to cool down laptop, undervolt, underclock not working

Nice. I'm now willing to buy thermal paste, but I just want to make sure I truly need it first. I'll set up a temperature alarm script so I know when I need to kill something to prevent this from happening again. If my temperature alarm goes off too often, I'll buy some thermal paste.

Offline

Board footer

Powered by FluxBB