Linux 5.11 Gets New Framework To Help Avoid Burning Your Skin On Hot Devices

[phoronix]

Phoronix: Linux 5.11 Gets New Framework To Help Avoid Burning Your Skin On Hot Devices

While the Linux 5.11 merge window has been over for one week where new features are normally added, a power management pull request sent in today for mainline is adding some tardy features including the Dynamic Thermal Power Management (DTPM) framework that in part is designed to help ensure users don't burn themselves with hot devices...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-5.11-DTPM

[AlDunsmuir]

Cool!

[muncrief]

Oh man, this brought back some amusing, but painful, memories.

For most of my career we only had full custom or gate array ICs, and in the era of Apollo DN4000 and other similar workstations simulation was not that accurate. So after simulating a device we often had to prototype it with TTL components. After all, the NRE costs were somewhere around $50,000, and companies did not like it when that first 10 device run failed.

And wow, those little TTL puppies could get extremely hot, sometimes even melting or exploding. In fact we often "debugged" errant circuits by passing thermometers over them.

But occasionally we discovered those errors with our skin! The worst time I remember was when I was prototyping a 3000 gate device, got a little too lazy with my wire wrap gun, and two dangling wires were crossed. I reached my arm over to toggle a switch and blam!, I seared my forearm in an instant.

Ahhh, the good ol' days

[rabcor]

Nobody needs this, and nobody wants this, this is the kind of feature that just shouldn't be bothered with, wtf is it doing in the kernel, where's linus torvalds to slam down his fist on this shit?

The internal component temperature of a device cannot be a proper indicator for external thermals (such as on a laptop case) what are they gonna do? cover the exterior surface of the laptop with thermal sensors? Or just prevent the CPU from going past 45°C?...

This 'feature' is inventing a solution, to a problem that has already been solved, my CPU frequently runs in the 90s (no pun intended) for extended periods of time (gaming!) in my laptop, the surface however does not reach temperatures where I could burn myself. This feature is entirely pointless, and implementing it correctly requires undesirable hardware modificatoins, it's just plain stupid. Stop being stupid.

[ultimA]

Nobody needs this, and nobody wants this, this is the kind of feature that just shouldn't be bothered with, wtf is it doing in the kernel, where's linus torvalds to slam down his fist on this shit?

The internal component temperature of a device cannot be a proper indicator for external thermals (such as on a laptop case) what are they gonna do? cover the exterior surface of the laptop with thermal sensors? Or just prevent the CPU from going past 45°C?...

This 'feature' is inventing a solution, to a problem that has already been solved, my CPU frequently runs in the 90s (no pun intended) for extended periods of time (gaming!) in my laptop, the surface however does not reach temperatures where I could burn myself. This feature is entirely pointless, and implementing it correctly requires undesirable hardware modificatoins, it's just plain stupid. Stop being stupid.

Have to kind of agree here. If you can burn yourself touching the outer case then it is not a problem of a missing software feature, but a bad thermal design of the laptop.

On the other hand, this wouldn't be the first (and certainly not going to be last) case where manufacturers try to fix a hardware issue in software.

[ssokolow]

Nobody needs this, and nobody wants this, this is the kind of feature that just shouldn't be bothered with, wtf is it doing in the kernel, where's linus torvalds to slam down his fist on this shit?

Didn't you hear? In some countries, it's illegal to sell hardware unless it has this. I'm pretty sure those people want the option of buying said hardware, even if they then rip out the DPTM stuff, rather than having to go for grey-market imports.

...especially now that we're starting to see geo-fenced "We can't legally sell this to people in your area" restrictions on sites like AliExpress where you'd expect to go for your grey-market imports.

[elvenbone]

Nobody needs this, and nobody wants this, this is the kind of feature that just shouldn't be bothered with, wtf is it doing in the kernel, where's linus torvalds to slam down his fist on this shit?

Exactly, this should be considered in the hardware cooling solution, case design and component-internal throttling logics directly.
The continuous maximum power operation of the computer's internals (which may be subject to multiple components', like SSDs' and CPU's, throttling logic) gives a defined heat radiation and heat transfer distribution. That heat distribution, together with heat dissipation via convection by passive and active cooling, and consideration of covered case surfaces (the laptop's bottom may be covered by the cotton of the user's clothes) lead to some point of equilibrium. If that equilibrium shows outer case temperatures of over 45°, well then the overall design is bad.
But yeah, just fix it in Software by crippling performance, when you have the device on your lap instead of a glass or metal table. Reviewers won't notice because they don't review it on the couch, in bed or on the train. And the user will buy it, believing in getting a performant device.

EDIT: Sorry for the repetition of arguments, I was too slow.

[V1tol]

I heard some expensive laptops do have some thermal sensors on their cases. And they deal with the problem by hardware-BIOS-UEFI monitoring stuff. I don't think having this in kernel is a good idea, especially if it will have "universal" approach. Yes, my notebook is hot, but I use it mostly like desktop (I don't do heavy stuff on battery = no heat). And I don't want that thing to cripple performance of my machine. Looks like another kernel argument for me (I hope they will add a knob to disable it).

[extremesquared]

I guess this spells the end of the IOT toaster.

[oiaohm]

The internal component temperature of a device cannot be a proper indicator for external thermals (such as on a laptop case) what are they gonna do? cover the exterior surface of the laptop with thermal sensors? Or just prevent the CPU from going past 45°C?...

That is interestingly right and wrong. Yes it right that you cannot get the exact temperature of the case external thermals. But you can get a max possible tempeture of the case by thermal dynamics delta calculations. There are calculable worst case delta values to a cooling solution. If cpu dia temp is 100C the cooling solution has a worst case delta of 60C so the surface temp of cooler/case has to be at max of 40C or lower. The reality is thermal dynamics have very strict rules that higher heat to cool has to be transferred to lower heat area. Some of this is knowing how much surface area the heat will be spread over in the cooling solution some is understanding what the transfer saturation points are.

Worst case delta is workable out. Best case delta not so much. Its really simple to forgot items like heat pipes have dry out and material have max thermal transfer rates this is why you can make a worst case delta value because it calculated on the max of the materials so giving you a max possible temp at the surface.

rabcor the key thing here to avoid human burns is amount of thermal transfer. Remember seeing humans walking on ultra hot coals the trick is the ash is a thermal isolation. So in theory you could have a laptop with a cpu running at 3000+C on your lap and not burn you just as long as the thermal conductivity in the human direction is low enough. This is all calculable. Yes you could be monitoring the 3000+C and know that it safe for human on top of the low thermal conductivity material due to how thermally resistant that material is This maths will not give you the exact temperature on the surface it will give you the max possible for human safety that good enough.

There are a lot of things you can calculate about a thermal solution to give delta values to take X temperature sensor inside a device/computer and give a max value for another temperature sensor in the device be absolutely sure that that sensor is lower than the calculated value. Fun part is its max possible values not what the current true value at that point is.

So you are right that you cannot give exact external thermals of a laptop case by calculated delta methods but the calculated delta methods can tell you if there is a possibility to burn a human or not because if the calculated delta max value is lower than burn human its not going to burn human. Yes this will leave a little performance on the table generally the max by delta and the real value is inside 5 C of each other.

Do note delta methods do have advantages due to being able to calculate deltas between all your thermal sensors inside your device if you are seeing one exceed calculated max something is damaged either the sensor or the cooling solution.

You think about it you put a thermal sensor on the case its out in a very exposed area it has a higher risk of getting broken. Calculate delta solution that only gives a max possible value for temperature has the advantage of being able to keep the thermal sensors deeper in the device so more protected. This is a trade off between being exact with higher risk of breakage to being good enough leaving a little performance on the table using delta method with less risk of being broken. Yes some high end high performance laptops put sensors in the case itself so they can extract every bit of performance they but there is a higher risk if those sensors go bad that human will get burnt.

Delta method allows using more than 1 sensor so a single defective sensor does not cause a safety problem due to being able to use delta values to compare between them so you can detect a sensor returning impossible hot or impossible cool value. We don't want a Boeing 737 MAX on our laps or hands if possible yes single sensor failure issue.