Mac Studio M2 Ultra and kernel_task constantly between 25-30% CPU usage after the computer has gone to sleep

[bubeli]

Good morning, I am disturbing you out of curiosity and evetually asking those who own a Mac Studio M2 Ultra if they experience the same thing.

Basically, I've noticed that the kernel_task process, which when the system is idle takes up about 5-10% of the CPU, after waking up from sleep takes up 25-30% consistently. Nothing dramatic, but I am interested to know if a value of 25-30% in idle can still be considered normal.

The computer is a Mac Studio M2 Ultra 60 core/64GB (the basic Ultra model, in short).
macOS 13.14.1

I also have a MacBook Pro M2 Max with exactly the same software installed, but this problem does not happen. I have tried with two monitors (standard configuration), with only one. Attaching it to both the Mac Studio and the MacBook but precisely the thing happens only on the Mac Stdio.
With kernel_task at 25-30% I seem to detect an average CPU temperature 2 degrees celsius higher.
If instead of letting the computer go to sleep, I only activate the screensaver even for an hour kernel_task stays at 5-10%.

Thanks for any answers and best regards.

 

[haralds]

Might be backing up.

 

[EDS66]

I think sleep is broken in Ventura on M2 Ultra Macs.. I have the 60 core GPU model. If it goes to sleep and then wakes up, the CPU clocks go crazy. Everything slows down. Both GeekBench and Cinebench start reporting weird and low CPU speeds, e.g, 2.4 or 2.3 GHZ, sometimes 3.0 GHz. Reported speeds are all lower than they should be and random. Running benchmarks after M2 Ultra goes to sleep results in proportional low scores. For example GB 6 reports 2710 on fresh start on my (actually it ranges from 2680 or so to 2710, which was the highest I got.) After it is put to sleep and awakened, GeekBench 6 scores plummet 1000 points and I get results in the 1800 range. Same with Cinebench. Here scores drop from 2890 (highness I got without sleep) to low 2000s after sleep. Same with Speedometer 2.1. The latter reports 430 (the highest I got on it). After sleep it drops about 100 points. Once the computer is rebooted everything goes back to normal. I have disabled sleep and have not seen any aberrations since then. When I am done with working on it, I just chose Lock Screen.

Most likely this is a firmware issue with M2 Ultra and sleep. It's almost like after it sleeps efficiency cores are used primarily and performance cores are either idle or are not clocking up to where they should be. Benchmark results seem to point to this, but of course there could be other causes.

I returned my first one thinking this was a one off. The second one is doing the same thing. I reported this to Apple. Most likely this will be fixed in one of the upcoming OS updates. I think your issue and the issue with the SATA drives in the new Mac Pro stem from the same root. Something is really broken in Ventura sleep on desktops with M2 Ultra.

Edit: I also have another Mac -- Mac Mini M2 pro. This abnormal behavior is not present on it. Sleep or no sleep benchmarking utilities report correct CPU speed, and benchmark results do not very significantly before and after the computer sleeps.

 

[bubeli]

Good morning,
small update on the topic.
Apple, after keeping my Max Studio in service for over 12 days, wanted to return the computer to me, saying they found no hardware defects, but not even reassuring me that the problem would be fixed with a subsequent firmware/software update. In short, they saw the problem, but not knowing how to fix it they wanted to leave me with a computer that essentially could not go to sleep.
At my insistence, they then replaced my Mac Studio just in case. The new one also has the same problem.
The Mac enriched last Friday--and the knocking thing is that yesterday's Ventura update seems to have fixed the problem. So the technicians knew about the problem and also that it would be fixed.
Now, I am happy that Apple was very resilient, replacing the Mac Studio for me, even though the legal time for the right of return had expired a few weeks ago.
But what I wonder: couldn't the technicians/engineers have passed the communication on to the service center or to whoever answers Customer Service on the phone? It would have saved us all a lot of time, because knowing that they were working on the problem I would obviously have waited. Whereas so, faced with an initial unwillingness to resolve the quetion, I persisted and asked for a replacement, even though I knew that 99% of the time I would still have the problem. But it was worth a try, given the cost of the device.
I will conduct further tests this evening, but it appears precisely that the problem has been solved.
Greetings.