0

EDIT 3: It's nothing to do with which GPU the system is using. The thing that seems to be triggering the failures is wifi. If I switch wifi off from the Apple menu then the CPU ramping stops, the graphics performance returns to normal and the system fans spool down.

With wifi on the ramping behaviour as described in previous edits occurs.

Additionally it logs multiple failure messages in the logs every few seconds, even though the wifi connection appears to be active and stable. A sample of the logging messages is provided below:

default 13:53:55.742497+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: BusState.txt 1603544035.157804 default 13:53:55.743014+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: MMIO_Log.txt 1603544035.157804 default 13:53:55.743643+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: LowLatencyRxCompRing 1603544035.157804 default 13:53:55.744249+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: LowLatencyTxCompRing 1603544035.157804 default 13:53:55.744836+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: TxFlowRing5.bin 1603544035.157804 default 13:53:55.745560+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: TxFlowRing4.bin 1603544035.157804 default 13:53:55.746138+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: TxFlowRing3.bin 1603544035.157804 default 13:53:55.746767+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: DebugCompletion.bin 1603544035.157804 default 13:53:55.747272+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: DebugSubmission.bin 1603544035.157804 default 13:53:55.747884+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: TxCompletion.bin 1603544035.157804 default 13:53:55.748532+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: RxCompletion.bin 1603544035.157804 default 13:53:55.749110+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: RxSubmission.bin 1603544035.157804 default 13:53:55.749735+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: ControlCompletion.bin 1603544035.157804 default 13:53:55.750196+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: ControlSubmission.bin 1603544035.157804 default 13:53:55.750878+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: DeviceInfo.xml 1603544035.157804 default 13:53:55.751271+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: InterfaceState_apsta.txt 1603544035.157804 default 13:53:55.751813+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: InterfaceState_prox.txt 1603544035.157804 default 13:53:55.752351+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: InterfaceState_infra.txt 1603544035.157804 default 13:53:55.752973+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: CoreState.txt 1603544035.157804 default 13:53:55.753464+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: InterfaceState_llw0.txt 1603544035.157804 default 13:53:55.753926+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: fChannelSwitchDictionary1.xml 1603544035.157804 default 13:53:55.784527+0100 corecaptured Session Name: BCMWLAN Net Roam Failure~status=3,reason=4, Blob Name: fChannelSwitchDictionary[0].xml 1603544035.157804

(this is only a sampling as the logs get very noisy very quickly)

It appears that each of these log events also generates a corresponding set of XML files in /Library/Logs/CrashReporter/CoreCapture/WiFi/ though I've not been able to decipher them yet.


EDIT 2: The CPU ramping and sluggish behaviour in certain apps has returned. Again, no applications appear to be exhibiting particularly high processor usage. However, WindowServer and kernel_task seem to be hitting unusually high utilisation figures. The graph at the bottom of Activity Monitor is showing mostly red (indicating the majority of CPU utilisation is system rather than application related)

One thought that occurs is that maybe there's a bug in the AMD GPU drivers? This behaviour goes away whenever the AMD GPU is disabled


EDIT: Today the system seems to be behaving perfectly normally, even though I've not changed a thing!

enter image description here

Unlike with the graph posted below, there actually were a couple of applications running (though they were all minimised and idling). This is more what I expect a Power Gadget graph to look like under light load. Blender is also behaving normally, as is the screen saver.

Obviously I'm glad that the problem isn't manifesting itself just now, but I still have no idea what may have caused it or why.


In the last couple of days my MBP (16 inch, Catalina 10.15.7, AMD Radeon Pro 5500M 8 GB GPU) has started suffering from unusually high fan noise and lagging graphics (most notable in GPU-intensive applications like Blender and the Drift screensaver, though oddly enough Minecraft seems to be unaffected).

Activity monitor doesn't show any application making unusual demands of the CPU, but Intel Power Gadget is showing a weird behaviour in the CPU clock speed and power usage graphs. Instead of being relatively stable as they've normally been when I've ran the power gadget in the past, the CPU clock and power draw are oscillating up and down in a way that seems very abnormal. This is true even with no application software other than Finder running.

Screenshot

Above is a screen capture of Intel Power Gadget, showing the weird CPU frequency and power graphs. This happens both when the machine is connected to a monitor, or when it's using its own display and is running something which triggers the AMD GPU. It doesn't appear to happen when the Intel GPU is active (so when there's no external display and nothing making demands of the AMD GPU).

enter image description here

The above is what Activity Monitor reports as happening when Intel Power Gadget is showing the weird CPU ramping. As you can see nothing is making any particular demands of the system

I've not installed any new applications recently. I have upgraded the Wacom tablet drivers and have thought that maybe that was the issue (I have had problems with the drivers in the past), but the problem persisted after uninstalling the drivers. A scan with a virus scanner turned up nothing untoward, nor did a scan with Clean My Mac. I have tried both an NVRAM reset (hold power button for 10 seconds) and an SMC reset (hold Command Option Shift and Power for 7 seconds) without success

This issue is degrading performance in Blender almost to the point of unusability, as every 30 seconds or so the display freezes or drops below 1 FPS for a couple of seconds. Note that this has nothing to do with the complexity of the scene in question, this behaviour happens regardless of what's being displayed, and even happens when the only thing there is the default cube! It's as if something is trying to trigger something in the GPU in an endless cycle but I don't know what. I've tried looking at the Console logs but I don't see anything that could obviously be an issue (though admittedly I don't know what I'm looking for)

This issue has also caused more fan noise in the last couple of days than I've been used to prior to now. That isn't a huge issue, but it's very annoying. Also the Drift screensaver doesn't look very good freezing every few seconds!

5
  • There's nothing here that's indicative of a problem - all that's being shown is what the CPU is doing and there are applications - depending on what's being asked of them, will spike the CPU like this. Secondly, fan noise is a symptom of nothing. Your fans come on because your SMC (in it's algorithm) noticed a temp condition that told it to spin the fan up - it's not abnormal. Lagging graphics can be due to any number of things - the question is - what are you running at the time you experience it? – Allan Oct 12 '20 at 3:53
  • @Allan As stated in the question, the only thing running when that graph was captured was Finder and Intel Power Gadget. This is definitely abnormal behaviour because normally the Power Gadget graphs are pretty flat under those conditions. Activity Monitor offered no insight either because it didn't show anything making particular demands of the CPU. I'm also aware that the fan spins up because the system is heating up, the question is why is it heating up when it's not doing anything? – GordonM Oct 12 '20 at 7:03
  • Apps that you're aware of...what about the background processes? We need to see Activity Monitor or from the command line, the command ps -Ax. You could also run top or download and run htop (it's more robust than the built in top). Remember, a graph is a top line "tool." To see what's going on, we have to dig deeper. – Allan Oct 12 '20 at 7:09
  • Added an Activity Monitor screen grab. As you can see the CPU usage graph there is pretty flat and the only things using significant CPU are the kernel, corecaptured (which I believe is part of Apple's screenshot utility), Activity Monitor and Intel Power Gadget. The Power Gadget was showing the same sawtooth CPU clock and power as in the originally uploaded screen shot – GordonM Oct 12 '20 at 7:45
  • Additionally, the CPU performance graph in Power Gadget is a lot more normal when running on the Intel GPU (On the Macbook's monitor) and when About This Mac isn't displaying the AMD GPU in its summary. When you connect an external monitor or do some other action that activates the AMD GPU the power gadget graph starts spiking again. Additionally the screen saver doesn't stutter or lag when on Intel GPU. This was verified with a day of running on batteries and no external monitor – GordonM Oct 12 '20 at 16:05

You must log in to answer this question.

Browse other questions tagged .