Trashing Chromebooks | Hacker News Reader

Trashing Chromebooks

systemcall.org

261 pointsmafuyu61 comments

Those swollen batteries are rather worrying - unless these laptops are being subjected to temperatures far outside of their design spec or the charging circuitry is (very slightly) overcharging them, they should not be swelling at all. Although in more gentle conditions they'd probably last long enough for the few years design life these have.

The article implies the issue was caused by indefinite high temperature, and the heat sinks performed correctly. I'm no battery engineer, but I assume the temperature spec is only for some finite interval.

Batteries are bags of scary-sounding chemicals. I do not trust them.

Yeah that's pretty crazy. If the battery produces gases like that, shouldn't it be vented?

> If the battery produces gases like that

Run.

Yes spot on.

When I was a teenager I used to race RC cars. The things had pretty hefty NiCd an NiMh cells. Occasionally one would vent and catch fire in someone's car or more usually attached to a fast charger.

Flames, smoke, nasty chemicals spilling out and everything for about 2 feet in each direction pretty much destroyed or covered in chemical gunk.

Now there are LiIon and LiPoly cells with an even higher energy density and discharge rate. If one of them goes, I'm out of the place. I trust a 240v mains line over those things.

Typical outcome (artificially started with a hammer):

https://www.youtube.com/watch?v=QjkW3KUz5uo#t=116

just last week a house got burnt down overnight due to an electric bicylce's bettery catching fire while left charging. Luckily nobody got hurt or worse. Moreover, a quick glance over news sites indicates this happens a couple of times a year.

Run indeed.

Something tells me you don't want the lithium in your batteries reacting with the water in the air.

Hey, not a chemist :-) But wouldn't they be "reacting with water in the air" rather quickly if they burst?

I've had several (three or four) MacBook batteries swell up like that from normal/slightly heavy use after only a couple of years. This was in the 2007-2008 era or so, though, when you could actually remove and replace them.

I think it might happen more than people think.

This happened to me too. Ended up just disposing of the battery and using the laptop with the charger as a direct power source.

My father's 2010 MBP did this a couple of weeks ago. He opened it to stick an SSD in it and saw that the thing had bulged inside the chassis. Might have been like it for a year.

FWIW, the Macbook Pros from a couple years back can also develop this problem, where the battery would swell ever-so-slightly and cause the trackpad to stop functioning.

The Samsung Chromebook is a fanless machine, relying on cooling from the ambient air around it. Having to remove the lid switch to allow the machine to power on with the lid closed should have been a pretty strong signal of this.

Since they are far, far into voided-warranty land already, I would say just remove the lid completely and thus get better airflow and cooling for the board.

As a curiosity -- why not use QEMU to test ARM? Or is that risking incompatibility with real ARM hardware?

In practice, both (qemu and hw) buildbots are useful. qemu is a bit faster and easier to scale, hw is the ground truth.

Google Native Client buildbots are exactly like that:

qemu: http://build.chromium.org/p/client.nacl/builders/precise_64-...

pandaboard: http://build.chromium.org/p/client.nacl/builders/oneiric_32-...

And yes, Panda is very unstable under heavy load. :(

Needlessly provocative title/link bait.

The Chromebooks are not trashed, they are simply re-installed with Linux and have their batteries removed so they can rest in racks for long-term testing.

tl,dr; There aren't enough reliable ARM servers yet, so for developing Linaro (optimized release of gcc), they use Chromebooks.

I like how any interesting article I've seen about Chromebooks in the last year or two starts with removing Chrome OS from the machines.

Agreed. I use an HP Chromebook 14 with an Ubuntu chroot as my main machine, and I love the Chromebook, but I have no love for Chrome OS.

It's just too damn limited to be appreciable.

If you put the devboards in an enclosure with vents on opposite sides and a small exhaust fan on one vent, you will get much better longevity, as that keeps air moving across the hot parts. I don't know if it's a better solution than the chromebooks though.

A fan seems a much simpler solution.

Then again, I am playing with Odroids now and don't like to hear that they will fail if I max them out for days on end.

Even a small refrigerator seems simpler than hacking up a chromebook.

In high school I tried building a PC into a refrigerator, and the results weren't pretty. The fridge could not keep up with the heat being generated by the CPU and over time it got too hot and shut down. The electric bill was also quite expensive, keeping a fridge running full blast 24/7.

Maybe hacking up the fridge to have the Freon running right over top of the CPU, but I'd rather take my chances tearing apart a Chromebook than tearing apart a fridge.

http://www.overclockers.com/extreme-water-cooling-using-refr...

Yeah it's super old.

There's also the guy who using his pool water to water cool a few machines.

http://www.thebuehls.com/pool_cooling/

You cannot put heat generating devices into refrigerators. The thermal load they are designed to handle is quite small. They cannot handle a continuous internal heat source.

This seems like a job for one of those wacky mineral-oil-submersion aquarium PC things. http://www.pugetsystems.com/submerged.php

There are some definite advantages; the thermal resistence from components to mineral oil is something like 4x lower than to air. However, you need to be careful as I've seen mineral oil travel up cables and over the top of the aquarium (was the worst with IDE cables, I expect it would be bad with these fancy nylon or fabric wrapped cables you see today as well).

My current ARM build machine of choice is the 2G quad core Utilite http://utilite-computer.com/web/utilite-models with internal SATA SSD.

Only annoyance is stupid power connector, which really needs fixing as it comes out too easily.

Its fast, and I run it fairly heavily although builds are not continuous, but it is on 24x7.

I feel like the NVIDIA Jetson TK1 would be a better alternative to the odroid products.

https://developer.nvidia.com/jetson-tk1

jetsons are not inexpensive.

Neither are any of the alternatives, really. The Jetson TK1 is a great little board, and it's significantly faster than the original ARM Chromebook (4x2.3GHz Cortex-A15 vs 2x1.7GHz). It also has a fan, so it should stay cool under prolonged load.

The TK1 is probably worth testing out and looks at least marginally faster in most tests, but it's worth pointing out that it seems nowhere near as much faster as its specs might suggest, and stability running under constant load is still an open question. Here's a couple benchmark comparisons courtesy of Phoronix: http://openbenchmarking.org/result/1405010-KH-NVIDIATEG94,13...

(Still, if it were stable it seems like it's much preferable to having to individually mangle Chromebooks and the TK1 comes w/ Ubuntu preinstalled on it's eMMC. The bootloader can also be flashed to boot from SATA)

So what about cross compiling on x86? My company make SW for switching hardware which is mostly PPC (P2020) and we compile everything on large x86 systems.

I don't understand your question... the article is about automated software build testing on ARM, not the build step itself.

isn't it? their complaint about the beagles is that they don't have enough ram to be compiling anything without refactoring the build system.

Being able to cross compile tells you exactly nothing about the state of your native build environment on the target platform. I mean, just because I can successfully cross compile C code for an attiny doesn't prove that gcc has been successfully ported to run on an attiny.

Having worked on the Samsung chromebook firmware I second this, you can configure the chromium-os build system to cross compile any piece of the software/firmware stack (or to use a prebuilt package) and most deving was done this way.

If it is at all an option, check out the chroot and build system of chromium-os and see if you can import your toolchain packages, the environment should be strait forward to extend wih your own custom dependancies.

Huh? Why not buy these? http://www.amazon.com/RK3188-Quad-core-Bluetooth-headphone-s...

In my experience (and the experience of other users; Google for some blogs), the RockChip devices are pretty low quality. Often it's tricky to get Linux on them and if you do, it won't be with a mainline kernel and hardware support may be patchy.

well, maybe Linaro could do something about this, if only author had known somebody working for Linaro ....

Why not use Android sticks instead? For $60 you can get a quad-core with 2GB RAM, USB + HDMI + WiFi + BT and passively cooled. You can place dozens of them on a single rack. What would be Chromebook advantage here?

I have a feeling those sticks wouldn't stand up as well to the sort of load this use case would require of them.

I suspect the latest smartphone SoCs would be as powerful as a Chromebook.

Bay Trail is about the same speed as Core 2 Duo or 1st gen Phenom. I assume Snapdragon 800, Mediatek MT6595 and new A57 chips are there as well, or even better. Soon we should start seeing chips approaching the speed of i3 on smartphone SoCs.

These sticks definitely do not have the cooling capacity to run at full rated heat output 24/7. As you note, they're passively cooled. They don't produce any less heat than a Chromebook on the same SoC, they just throttle and downclock faster.

I meant "passively cooled" as their advantage. You can always place some Noctua on top of them if needed, you just don't have to. I have one of RK3188-based sticks which is enclosed in a passive case. I also run the latest Bay Trail NUC in passive mode 24/7 and its temperature is higher on heavy load than of that Android stick. Performance wise BT is about 2x as fast, though it depends on a task. 9W vs 14W at full load including OpenGL, 1W vs 4W at idle.

I got the impression that most of the sticks and devboards are only "mostly" debugged, and seem to fail under 24/7 intensive use. As a mass production consumer product, the Samsung Chromebook has gotten a lot more debugging work.

Also those sticks are mostly intended for Android use, whereas this application uses regular Linux.

Are there not ARM Chromeboxes around?

There's the Samsung Chromebox Series 3,

http://www.samsung.com/uk/consumer/pc-peripherals/chrome-dev...

Edit: Nevermind, seems unavailable now. And not ARM. Derp.

Unfortunately the Samsung Chromebox Series 3 uses an Intel processor, not an ARM processor.

A slightly less "friendly" alternative to an ARM-based Chromebox would be the NVIDIA Jetson TK1. It's a developer board with an actively-cooled Tegra K1 SoC. We've been experimenting with these boards over the last two weeks and haven't encountered any stability problems. We're not pushing them as hard as Linaro does but we're certainly putting them through their paces.

It seems 'cheap ARM linux platform more powerful than a Raspberry Pi' is the chromebook's primary use these days.

I'm surprised they didn't go for the various rk3188-based tv-player android devices. There's got to be one hackable enough for Ubuntu.

The name of the game with ARM devices is often support and stability. There are plenty of ARM boards out there with just barely enough driver support to get Ubuntu running, let alone hacking it on yourself. Like the post said, Chromebooks already come with good hardware and software support.

What about a chromebox?

http://www.asus.com/au/ASUS_Chromebox/

That isn't an ARM box.

http://www.asus.com/au/ASUS_Chromebox/Chromebox/specificatio...

Intel® Core™ i7-4600U Processor or Intel® Core™ i3-4010U Processor or Intel® Celeron 2955U Processor

doh. Sorry, got lost in the excitement ;)

Was this written in the 2000s? "The only problem is, there aren’t any ARM rack-servers or workstations." There have been commercial rack mountable ARM servers for years now. Even the simplest Google search shows a wide range of ARM rack mountable servers. You can even special order them from Dell. What are don't seem to be is great value. I am sure a few netbooks would be better value. (ps re-installing the operating system is not what anyone except maybe ma and pa, would call trashing one).

Googling "Dell ARM Server" shows me a bunch of articles from 2010 about a proof-of-concept.

What do you recommend that people order from Dell, for less than $2000? (Note that the Chromebook in the article retails for ~$200, and IIRC the other ARM boards mentioned are also like ~$100-$300.)

Maybe there's a way to hack Ubuntu onto a cloud-connect stick?

I've tried to order them from Dell and gotten no where, both through a business rep and one of their partners that does servers and storage.

The response back was that Dell was only interested in selling in very large quantities to customers that had very specific workloads.

If you only need a handful for a lab or build farm, they don't care to sell to you.

http://exxactcorp.com/testdrive/ARM/

Can't see where you buy it or the price; just the remote test drive option. Is it actually a product yet?

Reply on news.ycombinator.com