Update: My problem is different from many here so I opened a new question called Surface
Pro 2017 with i7 overheating under high CPU loads (both Utilization and Speed) Probably best to ignore this...
I opened my new Surface Pro i7/512GB/16GB machine yesterday. It has shut down several times today.
HEADLINE: Each time it shut down I got a Kernel-Power event "The system was hibernated due to a critical thermal event."
Details:
I can make this happen at will. My Visual Studio solution has 6694 unit tests. If I run them in parallel (which that's what you're supposed to do, have the hardware work as hard as possible) then somewhere between 35 sec and 2 min it will shut down. This
has the i7-7660U CPU @ 2.50 GHz. The CPU will peg at 100% and about 3.77GHz until it shuts down.
I saw the BTHUSB event each time, but that was about 30 seconds after the Kernel-Power Event.
Here is the Kernel event:
Event 88, Kernel-Power
The system was hibernated due to a critical thermal event.
Hibernate Time = 2017-06-27T21:20:07.908155400Z
ACPI Thermal Zone = Intel(R) Dynamic Platform Thermal Framework
_HOT = 375K
Now the CRAZY part... If I keep moving my mouse, rather than pegging at utilization of 100% and a speed of 3.77GHz, it fluctuates a bit and the CPU speed will be between 3.52 and 3.68GHz - and utilization will fluctuate a bit too. And my tests will run to
completion! Taking about 3 min 20 sec. I have 5 Kernel-Power events that all match the times of my machine turning off and I have run all my unit tests 4 times by doing something else on the machine. It seems to be the combination of 100% utilization along
with high (3.77+GHz) speed that does it. I turned off the feature that has the tests run in parallel and I'm only about 50% CPU utilization but Speed goes as high as 3.94GHz. And the tests took 4 min 30 sec to run instead of 3 min 20 sec.
When I realized this I unplugged the power (it was plugged in) and the CPU seldom had a speed over 3.0GHz on battery. And the utilization was not always 100%, even though the unit tests were set to run in parallel. Then the tests took 3 min 45 sec.
So I have 3 workarounds:
- Keep moving the mouse while my tests are running. (fastest but silliest)
- Don't run the tests in parallel (and let them run 35% longer). (Best when docked)
- Disconnect my new Surface Pro from the power and run tests in parallel (and let them run only 14% longer). (Best if not docked)
This has been the only condition for me so far that has shut the new Surface Pro down.
I hope this helps someone at Microsoft figure out what to do.
If others face this, look at the Event Viewer and see if you have Kernel-Power events at the time of the shutdown. If not, then perhaps there are other issues than the one I face and we have several things going on.