We need Purism to make a statement about this issue. The other design compromises are acceptable but if the phone is physically incapable of running for extended periods at near max rated cpu frequencies this project will fail.
Looking at the figures here, I’ld say it’s fine. It is on the hotter side of the spectrum (compared to android phones). But it is acceptable.
And we can always slap a piece of metal onto the CPU to the case to conduct heat.
There was talk about this in the other CCC-thread. They need not make statements at this time but concentrate on the work. They got a few weeks to sort them out and another few to test (if Q3 deadline is to hold). It would be nice to hear something, but if they only got some pcbs last week, a week or three is more realistic at this stage. No point in saying anything before that, since nothing is certain or known before.
Besides: if “work * time = 1”, then we should expect major work done the closer the deadline nears
In this issue they test the device unser stress.
I don’t think they stresses device during the presentation. It was just laying around without Interactions.
And at the end of the presentation it wasn’t usable anymore.
So it looks like its still a major issue. Probably more than it is described in the issue entry.
The uncontained dev-kit should have exposed thermal issues in early testing. Why proceeded to ‘final’ pcb’s whithout addressing this problem?
Because it’s not related to the PCB but to the CPU drivers.
On one hand, the i.MX 8M Quad silicon issues meant not only that the temperature of the battery was way too high, but also that it could discharge too quickly. On the other hand, the i.MX 8M Mini is made in an updated silicon process (14nm instead of 28nm) and has reduced features, resulting in a lower power consumption–but it’s new, and there could be unknown issues and risks. So a lot of research was made, and our development team started evaluating the i.MXM8 Mini to see if it could be used within our requirements–free software only, no binary blobs, mainline based software stacks.
And then the month of February began, and something else unexpectedly happened: NXP released a new software stack for our first CPU choice, the i.MX 8M–and all of the power consumption and heating issues suddenly disappeared!
I’m sure they’ve worked / are working on the drivers and dev-kit, but it’s not untill the final pcb and housing (which is probably part of the heat transfer and dissipation) are worked on, that it’s possible to know how it’s going to perform - different airflow (compared to dev-kit), accumulating heat within the case, heat transfer to and from other parts (especially under different stress/use profiles)…
Well, we can do what android have been doing for ages - disable CPU cores when we don’t need then and have them back online when we do. It’s a simple enough hack to build.
wouldnt that be funny if i died before ever getting to see my L5 ? lol Im a mile from the coast where this now projected cat 4 hurricane is supposed to make landfall .
That would be a bummer if I never even got to see it . Well purism donate my L5 to a nice person on here if I die hahaha
Or you know, evacuate. Kind of the silver lining with hurricanes.
whos going to protect from looters ? plus my job has a generator and i might have to work . i have no where to go anyways
At that point, I say screw the job. My L5 is already paid for anyway
Could be because of my family, but I wouldn’t stay with the hurricane barreling down on me.
I just dont have anywhere to go and my bronco has an untraceable coolant leak my mechanic said he couldnt find . I was here for jeanne and frances in 04’ 145mph sustained with gusts upto 195mph . So this aint my first rodeo . Last cat 4 I was in I saw a 60ft singlewide trailer doing cartwheels down my road end over end . It was wild. plus I got enough food and gear to make it if I survive the hurricane. Itll be fun
The CPU makes up for everything else pulling way too much power (see the open issues), and gets up to 99% slower in the process. It only shows that the thermal driver works. This test (stress -c 4) currently makes the phone (lockscreen) unusably slow. A few minutes after stopping “stress -c 4”, the cpu still won’t be able to cool down and stays at 80°C + 99% idle-injection. Indeed this is important, but right now, results in really bad user experience.
This sounds already frightening and it was only 3 weeks ago.
80c? that cant be good
The presentation on the Chaos Communication Camp of the Chaos Computer Club was on 2019-08-22 so just a week ago. At about 44:20 of the presentation Nicole Faerber wonders about the development device: “Now it is off. I don’t know why.” and assures: “It works perfectly well.” and says: “It’s pretty hot so power management is still an area we have to work on.”
- The power management problem has officially been mentioned.
- I can’t judge how well she is up to date on that issue.
- I can’t judge how the development device compares to the end product.
- The development device was off in that situation.
- It is unclear why it was off. Problems of heat and power management are only one possibility. Maybe it was simply a mistake of usage.
- The device was hot.
My experience of performances on ramps is limited but as far as I know it can get very hot up there because of the spot lights. That may promote problems of overheat.
We could speculate all day long. As this is potentially a major issue and it is stated officially just a week ago by the CEO (she is?, isn’t she?) of Purism I think Purism should go for transparency and clarify the situation. No matter if overheating was the reason that the device was off.
I really don’t want to make panic. I just think a clarification is better than speculation.
Web page with video:
Direct link to video:
They are running stress tests, which is an artificial workload, in order to test the throttle feature. Normal workloads should not do this. The throttle is there in case applications misbehave, websites do too much, or there is a real workload that causes this (such as encoding video without hardware acceleration).
It looks like reducing the frequency during idle is still in progress (devfreq). It might be the case that they had this working from the vendor’s code from months ago, but they are still pushing the code upstream to make maintenance easier. If they switched from a custom kernel to a mainline kernel, then that would explain why power management is all of a sudden an issue. If that is the case, then the work required is just moving code upstream, or going back to a custom kernel.
My gut feeling says, that they are currently waiting for the full prototype, so they have something to show. Then they will mention some unexpected problems with it, and postpone until Q1 or even Q2 next year.
Guess we will learn this in a few days.
But god I hope my gut is a big fat liar
If heat causes the cpu to be throttled down to 1%, it should not have shut down during the presentation, but only run terribly slow.
A likely cause could be that the battery just ran empty.
Or cpu throttling might not have been active yet at all, e.g. when the device is connected to mains power.
In the video it looked like the device was plugged into power cable.