Building an ARM64 laptop [LWN.net]

Building an ARM64 laptop

Posted Sep 17, 2017 5:57 UTC (Sun) by marcH (subscriber, #57642)
Parent article: Building an ARM64 laptop

> ARM processors need no fans, meaning that the resulting laptop is lighter and will not burn the user's legs.

This is backwards. Fanless laptops (and smartphones) are the ones burning legs and hands since they have no other way to dissipate heat. This has nothing to do with the processor architecture; there are many fanless Intel SoCs too.

to post comments

Building an ARM64 laptop

Posted Sep 17, 2017 9:34 UTC (Sun) by excors (subscriber, #95769) [Link] (2 responses)

See also Nintendo Switch, which has an ARM64 SoC (Tegra X1) and a fan. And that's running at half the X1's maximum CPU frequency, and a third of its maximum GPU frequency.

It seems modern SoCs are far too powerful for sustained use - it's vital to have a good thermal design and then you'll probably still have to throttle it severely. That power is great for burst performance, which helps with typical usage patterns on smartphones, but perhaps not so much for laptops. But none of them should burn the user unless the device is incompetently designed.

Building an ARM64 laptop

Posted Sep 17, 2017 17:22 UTC (Sun) by anton (subscriber, #25547) [Link]

The more powerful smartphone SOCs cannot sustain the performance in smartphones because they cannot dissipate the generated heat. That should be a smaller problem for laptops, which have a much bigger case for dissipating the heat. I have also seen reports of Intel-based laptops not being able to sustain the performance because of lack of heat dissipation, but my own experiments with a Lenovo E31-80 showed that it can sustain the two-core turbo of 2700MHz for at least 10 minutes.

Building an ARM64 laptop

Posted Sep 17, 2017 19:59 UTC (Sun) by marcH (subscriber, #57642) [Link]

Fanless or not, most modern SoCs support peak performance that cannot be sustained over long periods of time and has to be throttled down eventually for thermal reasons. Simply because not all but many many jobs are short, so why not complete them faster and sooner.

> But none of them should burn the user unless the device is incompetently designed.

I think "burn" was meant metaphorically :-) The only thing I meant is: when there's no fan (no convection) then the only way the heat can be dissipated is through the case (conduction). The main article seems to have it backwards; maybe because it was trying to make a... "different" point about performance per watt.

You're being a bit harsh with "incompetently": it takes a lot of complex engineering to extract the most power from a SoC, especially a fanless one, while staying right at the limit before users feel uncomfortable and complain. Example:
https://software.intel.com/en-us/articles/2-in-1-tablet-m...

Building an ARM64 laptop

Posted Sep 18, 2017 5:19 UTC (Mon) by autious (guest, #114303) [Link] (1 responses)

Considering ARM's lower energy use, i could imagine that dissipating the heat generated on a high-end smart phone over the surface of a laptop might be enough to stay fan less without excessive heat?

I'm curious how big the difference in wattage actually is, but a preliminary googling didn't give me any useful results.

Building an ARM64 laptop

Posted Sep 18, 2017 9:49 UTC (Mon) by marcH (subscriber, #57642) [Link]

> Considering ARM's lower energy use,

This world is a bit more complicated than simplistic statements like this.

The ARM ThunderX has a 100W TdP. The x86 Quark D1000 can go configured as low as 1.6mW.

> dissipating the heat generated on a high-end smart phone over the surface of a laptop might be enough to stay fan less without excessive heat?

Yes, but no one is going to underpower a laptop with the TdP of a smartphone. That would be just a waste of ability to dissipate power and limit the performance for no good reason.

If low skin temperature is more important that performance then you can just reconfigure the performance settings, no need to design crippled hardware to stay cool.