M2 MacBook Pro Thermal Concerns | Hacker News Reader

M2 MacBook Pro Thermal Concerns

mjtsai.com

12 pointsrukshn11 comments

Wow this really saddens me. The reason I'm donating to Asahi Linux is to get a non thermally constrained small laptop. Currently I have a liquid metaled 9310 XPS 13 with thermal pads to connect the heatsink with the case. I still have to disable turbo boost to not have it throttle under sustained load. And the huge cavaet that you can burn yourself on the underside if it's under heavy you load.

I'd hoped I could stop hacking around soon and get a sane thermal experience. But this shows Apple doesn't really care either. A non combined workload brings it to it's knees.

It generally makes sense for a laptop to allow the CPU to do more work than can be sustainably cooled, since short duration high CPU workloads are common. You could view such a laptop as ‘thermally constrained’, or you could view it as maximizing the performance potential of the chasis by allowing a temporary boost. (Really, the performance of any practical laptop is heavily constrained by considerations of cooling and battery life.)

I think these complaints would make more sense if you specified some specific performance figures that you’d like your laptop to be able to sustain indefinitely. Otherwise you could ‘fix’ the issue just by downclocking the CPU.

There is no short burst high performance needs I have. the only thing where I need high performance is playing older games, compiling and watching 4k video. None of these are burst oriented. I honestly don't know any daily workload that requires short time bursts. Personally I feel Intel/AMD just do it for marketing BS.

The thing is, Apple caters to the average user, not too much to specific needs. One example of burst-loads would be compiling code. Also, does watching 4K video lead to thermal throttling these days with hardware accelerated decoding?

Compiling code is not short burst activity. I would say 90% of the time I spend compiling code is 10+ mins of time.

4K video decoding works on dedicated players. But neither Firefox or Chrome have consistent working support for it on Linux.

What does the average user do that requires such a high short burst performance? Besides I'm not so anti bursting if the laptops would still have Oke non-burst performance. But take the laptop I have now. The 9310 XPS 13. Even disabling turbo boost it throttles itself into to the ground to about 1.1 Ghz under load. How is that acceptable? I'm at least happy I was able to mod mine. But it's a total crapshoot when I buy a new laptop. Reviewers don't test it. There is almost no detail available of sustained performance of laptops.

Apple M1 macbook pros from what I can say CAN sustain it's performance indefinitely. Which is why I was very interested.

> I honestly don't know any daily workload that requires short time bursts.

JavaScript.

Why does JavaScript require short time bursts? In fact I don't see at all how a LANGUAGE requires short time bursts. that's way to generic. Some applications written in JS might require short time burst performance.

The standard use of JS when loading www pages in a browser scales nearly perfectly with frequency during short single-threaded bursts.

I doubt this. Web pages are bottle necked by network latency and throughput. My phone for example is just as snappy loading websites then my desktop. In fact to be sure I just tested loading reddit(A bloated JS heavy site) twice. One with my CPU with turbo boost and one time with my CPU without turbo boost. Chrome reports a load time of about 3.98 seconds with turbo boost and 4.01 with turbo boost. The dominating factor according to the performance tab is just http requests. I don't see how 30ms of less load time is significant. And that is with one of the bloatest JS sites in use today.

Is reddit actually JS-heavy?

On my work desktop with Chromium, loading facebook.com takes me 7.2s with 4.7s of that being scripting. If I throttle the CPU to a 4x slowdown in Chromium devtools it takes 46.2s with 42.4s of that being scripting. (Only a single run of each.)

Unfortunately in the middle of some work stuff so I don't want to reboot to toggle off turbo boost to test.

My personal hypothesis is that TSMC's timelines for Apple slipped and the M2 was originally supposed to be a cooler 3nm chip. Instead we have hotter GPU cores built on the same 5nm process.

Reply on news.ycombinator.com