Cooling a Thinkpad

Late last year I wrote about the way that modern laptops suck [1]. One of the problems that inspired that post was the excessive heat generated by my Thinkpad T61.

There is a partial solution to this, Fool Control explains how the kernel option pcie_aspm=force can be used on kernels from 2.6.38 onwards to solve a heat regression problem [2]. I applied this to my Thinkpad T61 and the result was that on a cool evening (ambient temperature about 24C) the temperature changed from 85C to 66C on the NVidia video card, and for the “virtual devices” it changed from 80C and 78C to 60C and 61C. I’m not sure exactly what each of those measurements refers to, but it seems that the change was somewhere between 17C and 20C.

This changes the system from being almost unbearable to use to being merely annoyingly warm.

I’m not going to make my laptop be my primary computing device again though, the combination of a desktop system with a 27″ monitor and an Android phone is working quite well for me [3]. But I haven’t yet got version control systems working for all my software. Also Wouter suggested using NBD which is something I haven’t got working yet and probably won’t until I can swap on it and therefore have a diskless workstation. Finally I still haven’t got the “Chrome to Phone” browser extension working such that a page I’m viewing at home can be loaded on my phone.

11 comments to Cooling a Thinkpad

  • Diggory

    Another way you can reduce noise (very noticibly) and some heat is to get an SSD. Also net you a nice performance boost (why would you want RAID?).

    Just read your previous article, and I must say I agree with the others disagreeing with you. Display hotplugging over displayport works very nicely with the NVIDIA driver (hint: disper and autorandr), such that I’ve been considering switching my 14″ T410 for something smaller like the X220. Don’t know about pricing in Australia, but here it’s possible to get an X220 for around 1150 CHF (no OS, through a university program). Only real advantage of a desktop from my point of view is quad-core processors, though hopefully Ivy Bridge will help sort that out.

  • etbe

    Diggory: While SSD is generally good for reducing noise and heat I don’t think it would make a great difference in this case. My Thinkpad is annoyingly warm (for my fingers on the keyboard and the base resting on my lap) in the area around the CPU while the area around the HDD doesn’t feel warmer than the rest of the system. When the system is idling in a room that’s about 25C (warmer than is comfortable but not THAT warm when compared to some of the things that my Thinkpad is expected to deal with) the HDD and CPU fan produce a similar volume, but the HDD has a higher frequency which is more easily blocked by headphones when listening to music or watching a movie. The fact that in spite of the cooling fan the CPU makes a large section of the case hot while the HDD with no fan doesn’t seem to warm the case suggests that using a SSD won’t extend battery life in any way that I would notice. As for performance, the laptop has 5G of RAM and could be expanded to 8G of RAM for less money than buying a SSD. HDD speed isn’t a real problem.

    My Thinkpad doesn’t support DisplayPort but I can apparently buy a docking bay to give it dual-link DVI to support a monitor such as the 27″ Dell monitor I’m using (see the above URL).

    The dual-core 64bit workstation I’m using to type this came from an e-waste pile. While I could spend money on a SSD and a docking bay to keep using the Thinkpad 24*7 it was a lot cheaper to just use a free PC. The PC only supports 3.25G of RAM which is rather annoying (less than the 5G from my Thinkpad), but apart from that it’s better in every way for desktop use.

  • Diggory

    I know the T61 only has VGA-out; I used to own one. And yes, you’re probably right about an SSD: I didn’t notice so much performance increase (perhaps faster searches); what I did notice was lack of spin-up time and noise. NVIDIA driver power profiles can reduce power usage considerably at the cost of performance (even noticable sometimes on a composited desktop). Doesn’t sound like I can tell you much you don’t know though; nice find in the e-waste pile!

  • I’m the author of that article on FoolControl, out of curiosity which kernel version are you using?

    @Diggory, argument about SSD is not valid because I have SSD in my notebook.

  • Robert

    I use NFS for netbooting a few clients. Highly recommended, although the performance is not stellar (lack of metadata caching I think, can’t be sure) – makes managing the setup a breeze, though, when you need to chroot and install some packages.
    About e-waste piles… seriously? I can never find any, they all end up on auctions and are never given away for free or left around :P

  • Frank

    So does that mean your previously article about why laptops suck was wrong, and it’s really just your choice of operating system that sucks?


  • etbe

    Adnan: 3.1.0 from Debian/Unstable.

    Robert: Some of the stuff I get through luck, some of it is through being friendly with the people who manage the rubbish. I recently got a fully quad-core system with 4G of RAM and a 2*DVI video card – but it was white-box so I’m giving it to a friend.

    Frank: Designing laptops to be as fast as desktop systems instead of being as energy efficient as tablets is still a bad idea. Thin laptops which have noisy fans and which bend and eventually break is just a bad engineering choice, even with the ACPI change my Thinkpad is still unreasonably hot.

  • Diggory

    @etbe: why do you still just measure perceived heat and not battery drain or something a bit more measurable? If what you’re proposing is a switch to ARM-based processors or something similar, I agree with you and I think MS do too — even with all the power saving features of modern laptops they still inevitably use around 10W on standby, which is a heck of a lot more than my ARM-based mobile.

    @Frank: power-optimising laptops is a bit more work on Linux than Windows or Mac OSX, but still entirely possible.

  • etbe

    Diggory: In most cases battery drain isn’t an issue, the laptop in question doesn’t get a lot of use away from mains power. Heat however is a real problem, the bottom of the case can get too hot to be in contact with my legs which thus makes it impossible to use the laptop on my lap.

    ARM won’t solve all the problems, there is also a lot of other hardware that uses power in a laptop. If you want a laptop that has a GPU capable of driving a large display and which has large banks of fast RAM and a spinning media storage device then you will still have power problems regardless of the CPU.

    Why do you think that it’s more work on Linux? Surely the MS and Apple people have put in a lot of work to optimise their OSs for power use. But we don’t get to see public git logs of it.

  • Diggory

    @etbe: because that heat’s got to come from usage of electrical energy, and apart from your display, nearly all of that electrical energy is going to go into heat. It’s just a lot more measurable than saying “my laptop’s hot” (or not).

    No, the CPU’s not the only thing. But why don’t you just buy a netbook then? Many of those are optimised for low power usage, if not being quite as low power as a tablet.

    And I really don’t understand your point about git logs… the point is just that I’ve run into power problems far more on linux than with Windows.

  • etbe

    Diggory: Almost all the energy gets converted to heat and light, probably the majority is heat. But the real issue is where the heat is located. Increasing the temperature of a large volume mass of the device by a few degrees is no big deal while increasing the temperature of a small part that touches the user by 20C is a significant problem.

    If the same Thinkpad had a copper case which distributed heat more evenly then the temperature wouldn’t be a big deal.

    I own an EeePC 701 Netbook, it also has cooling problems. Netbooks dissipate less heat but the vents are smaller so it’s still a problem, see the above URL for more information.

    My point about git logs is that all Linux development is done out in the open and we know how hard it is. We don’t know how much work is done on Windows. Also you should note that Debian/Stable was working a lot better than Debian/Unstable on the Thinkpad in question. I would not bet money that a development version of Windows would work well on the Thinkpad in question.