Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Is it crazy to want to repaste my CPU again on my 2018 Mac Mini?

E

Engender

macrumors regular
Original poster
Oct 6, 2007
190
36
Hello!

I just repasted at the end of January. I used Thermal Grizzly's Hydronaut. It prevent my CPU from throttling when browsing the web. Now, WindowServer is slowing my system down to the point where I need to restart every few hours.

I'm looking at using ARCTIC MX-6. Is this just stupid after a month?

Maybe I should get a new CPU cooler?
 
  • Wow
Reactions: russell_314
Something is wrong with your system, and it doesn’t involve thermal paste. Perhaps you did not apply it correctly? Thermal paste should last years not a month. Is the fan coming on?

Is the computer actually slowing down where you can notice it without some sort of benchmark or are you just not happy with the temperatures? Those Intel Mac mini’s get extremely hot and it’s normal. Short of doing something crazy like putting an external water cooler on this thing, you’re not going to get great temperatures.



My first thought is there’s some issue with the OS causing it to use the CPU too much. Maybe try completely reinstalling macOS and not immediately adding whatever applications you have back. If everything works after that, then add them one by one until you see a change if any. Troubleshooting is a PITA sometimes.
 
  • Like
Reactions: Queen6 and MacCheetah3
What version of the OS are you using?

Mojave is starting to show its age on the 2018 Mini -- particularly with regards to Safari, which seems to be getting "stuck" now and then and sending the CPU and fans into overdrive.
 
I'm running the latest version of Sequoia. Nothing is wrong with the system beyond the WindowServer slowdown.

That's a good idea about one-at-a-time. In fact, I run MacOS from an external NVME drive . . . so I've got the 256 GB internal drive to mess around with. For some reason, I just don't want to do that. But, I think that using the internal drive for work-critical apps (Word & Safari, really) is what I should do.
 
Engender said:
Hello!

I just repasted at the end of January. I used Thermal Grizzly's Hydronaut. It prevent my CPU from throttling when browsing the web. Now, WindowServer is slowing my system down to the point where I need to restart every few hours.

I'm looking at using ARCTIC MX-6. Is this just stupid after a month?

Maybe I should get a new CPU cooler?
Click to expand...

I would break this into two parts:
1) Is your CPU usage (e.g. as visible in Activity Monitor) as expected for what you are doing?
2) Is your CPU temperature as expected for your CPU usage?

Also which CPU do you have?

As an example, my i3-8100B Mac Mini is currently running 36 deg C. This is with just the Flurry screen saver active while iTunes plays random music in the background. A few idle Brave windows and a few other apps. I have never repasted my CPU but I do have my Mac Mini turned on its side as someone else promoted in another thread on this site and that has made a big difference. When I start a session of CPU intense programs, I also preemptively crank the fan to full blast via Macs Fan Control.

But if you're not seeing < 50 deg C (and ideally < 40 deg C) at idle, I'd say something is off. Also, Safari with no tabs open shouldn't drive much CPU. If you have a lot of tabs open or some of them are runaway, that will drive CPU which will drive temperature which will thermal throttle the whole thing. These days if you are browsing without adblockers and the like, you will likely end up with some runaway tabs. Which will quickly add up if you keep more than a few tabs open.
 
My CPU is the i5 version. As I'm typing this, it's 77 degrees celsius. But, this is an improvement from the 90 degrees plus that it was at idle before the initial repaste.

I run off of my internal NVME drive last night on a fresh install of Sequoia, and I didn't have the WindowServer issue I've been having on my current build. I think it has something to do with my mouse, which is a Logitech MX Master (I have both the original and the 3, and I use them interchangeably). So, Logi Options+ may be the culprit for the WindowServer issue.

Now I wonder if I could push the heat down by repasting the CPU again . . . Is there any downside to doing it again so soon? Or, rather, is another repaste so soon going to hurt anything—assuming I don't destroy the logic during the repaste?

EDIT: Hey bzgnyc2, could you do me a favor and run pmset -g thermlog and tell me if your Mini limits the CPU when you wake your computer from sleep?
 
Last edited:
Engender said:
My CPU is the i5 version. As I'm typing this, it's 77 degrees celsius. But, this is an improvement from the 90 degrees plus that it was at idle before the initial repaste.
Click to expand...

Sounds like that definitely helped but still sounds high for just typing in a browser. I would be curious what does Activity Monitor -> CPU shows when View -> All Processes is selected?

Also, to make sure we're comparing Apples to Apples, which temperature are you referencing? From Mac Fans Control, are you referring to the CPU Core (Average), CPU PECI, or CPU Proximity?

Engender said:
Now I wonder if I could push the heat down by repasting the CPU again . . . Is there any downside to doing it again so soon? Or, rather, is another repaste so soon going to hurt anything—assuming I don't destroy the logic during the repaste?
Click to expand...

That I can't answer as I've never had to do that in >5 years of ownership...

Engender said:
EDIT: Hey bzgnyc2, could you do me a favor and run pmset -g thermlog
Click to expand...

Current uptime of 6 days, that command reports:

Note: No thermal warning level has been recorded
Note: No performance warning level has been recorded
2025-02-27 14:51:13 -0500 CPU Power notify
CPU_Scheduler_Limit = 100
CPU_Available_CPUs = 4
CPU_Speed_Limit = 100

Engender said:
and tell me if your Mini limits the CPU when you wake your computer from sleep?
Click to expand...

I am not sure what you mean by that. I put my machine to sleep at the end of each day and wake it each morning and don't notice any slow down/limits.

I should note my Mini is currently installed with Mojave and Monterey. If Sequoia is at the root of your thermal issues, my configuration is not currently comparable (other than having a very recent Boot ROM / iBridge version).
 
Keep thermlog running. When you wake your Mac from sleep, see if CPU_Speed_Limit drops. Mine drops to like 19.

CPU PECI as show by TG Pro, which I'm also using to manage the fans.

1740688897933.png
 
Engender said:
Keep thermlog running. When you wake your Mac from sleep, see if CPU_Speed_Limit drops. Mine drops to like 19.
Click to expand...

No change to mine coming in and out of sleep. This is under Mojave (have not rebooted back into Monterey yet).

Engender said:
CPU PECI as show by TG Pro, which I'm also using to manage the fans.
Click to expand...

With various iLife apps open plus a few windows/tabs in Orion and a few more in Brave plus Netflix streaming the 4K test pattern plus iTunes running the projectM visualizer (an OpenGL screensaver/etc), my CPU PECI varies runs in the 50-60 deg C range. If I replace the Netflix streaming with a Speedometer run from Browserbench.org, it goes up to 70 deg C.

Engender said:
View attachment 2486556
Click to expand...

Well there's your problem. You have too many background processes chewing up CPU. Spotlight / mds should be ~ 0% if you're more than a few hours past boot / mounting a new drive. Not sure why all those other processes (e.g. airportd, opendirectoryd, runningboardd, etc) are using more than 1% CPU.

Then whether related or in addition, whatever is happening in Safari is also likely driving WindowServer (I've seen select website peg not only themselves but WindowsServer).

My little 2 core / 4 thread i3 MacBook Air is > 75% idle right now with two browsers with several windows each, two mail programs, and Word open. I don't see anything you are running that justifies a 6 core system running < 25% idle.

After your system boots and settles, but before you start any other app/etc, run this command in a Terminal window:
Code: 
top -u -s 60 -S

Mine is > 99% once it settles (which can take 20 minutes to two hours with recent macOS like Monterey).
 
bzgnyc2 said:
After your system boots and settles, but before you start any other app/etc, run this command in a Terminal window:
Code: 
top -u -s 60 -S
Click to expand...

Do you suggest I remove the startup apps as well before I try this?

EDIT: Also, I would love to see your activity monitor!
 
Engender said:
Do you suggest I remove the startup apps as well before I try this?
Click to expand...

I would do as-is to capture your current state and then try without them -- especially if your %idle is <98% with the startup apps loaded. If your system isn't very idle as configured with regular applications running but then actually runs idle when you remove your startup apps, then you know to look further at those to see if they are justified and/or need fixing/updating. If not idle in either case then you know you have more investigating to do.

Engender said:
EDIT: Also, I would love to see your activity monitor!
Click to expand...

Here is the top rows + top 10 processes about 2 hours after boot from some testing I was doing with Monterey on my Mac Mini back in December:

Processes: 325 total, 2 running, 323 sleeping, 995 threads
Load Avg: 0.83, 0.89, 0.91
CPU usage: 0.7% user, 0.17% sys, 99.75% idle
SharedLibs: 453M resident, 80M data, 35M linkedit.
MemRegions: 21626 total, 700M resident, 166M private, 339M shared.
PhysMem: 4362M used (1428M wired), 3829M unused.
VM: 10T vsize, 3153M framework vsize, 0(0) swapins, 0(0) swapouts.
Swap: 0B + 0B free.
Purgeable: 51M 342(3) pages purged.
Networks: packets: 34241/3664K in, 18769/3193K out.
Disks: 65883/1240M read, 24180/441M written.

PIDCOMMAND%CPUTIME#TH#WQ#PORTSMEM
0​
kernel_task
0.3​
00:18.8​
207/4
0​
0​
114M-
141​
bluetoothd
0​
00:04.2​
8
4​
215​
3216K
1128​
top
0​
00:00.2​
1/1
0​
27​
2236K+
1​
launchd
0​
00:04.1​
2
1​
1706+13M
235​
airportd
0​
00:01.4​
9
7​
188​
2940K+
472​
sharingd
0​
00:01.4​
3
1​
221​
6052K
133​
distnoted
0​
00:01.0​
2
1​
135​
440K
402​
distnoted
0​
00:01.1​
2
1​
253​
652K
148​
cfprefsd
0​
00:01.4​
2
1​
482​
768K
82​
logd
0​
00:01.3​
2
1​
1008​
4112K

Note the above is with the -s 60 option to top and after at least two refreshes (the first two refreshes of top are cannon fodder since it is measuring itself for a lot of those first two iterations).


P.S.Switching from Activity Monitor to top(1) for debugging this since you seem pretty technical and top has less overhead while giving more control over the refresh speed.
 
Here is what I have, but note that I am working on writing a motion, so not at idle:




Processes: 656 total, 6 running, 650 sleeping, 2754 threads


18:02:24 Load Avg: 5.51, 8.01, 9.31
CPU usage: 28.74% user, 17.78% sys, 53.47% idle
SharedLibs: 1047M resident, 157M data, 88M linkedit.
MemRegions: 1965 total, 60M resident, 4988K private, 3148M sh
PhysMem: 29G used (5211M wired, 0B compressor), 35G unused.
VM: 27T vsize, 5051M framework vsize, 0(0) swapins, 0(0) swap
Swap: 0B + 0B free
Purgeable: 4622M 42030(877) pages purged.
Networks: packets: 2521256/2726M in, 2006835/1423M out.
Disks: 531975/10G read, 209308/4948M written.


PID COMMAND %CPU TIME #TH #WQ #PORT MEM
173 WindowServer 40.5 10:11.86 13/1 6 4219 619M+
1043 Safari 21.9 07:58.63 18 10 2287- 176M+
146 launchservic 17.7 06:50.77 16 15 9767+ 4204K-
181 runningboard 8.8 03:27.29 9 8 790- 4448K
1 launchd 8.6 03:31.78 5 4 3812- 18M
1051 com.apple.We 8.4 03:35.27 10 5 154- 70M+
573 fontd 8.3 03:30.13 4 3 143- 3864K-
179 loginwindow 7.0 02:54.57 4 3 381 19M
483 lsd 6.4 02:31.41 3/1 2/1 606 25M+
2008 com.apple.We 5.5 00:23.22 9 4 82- 332M-
4655 com.apple.We 4.7 01:44.79 11 3 84+ 361M+
1052 com.apple.We 4.1 00:49.68 28 4 462+ 414M+
0 kernel_task 4.0 02:44.86 357/6 0 0 491M-
559 Activity Mon 3.7 02:04.53 7 5 1434 69M+
1013 Terminal 3.4 00:21.28 9 4 328+ 86M+
535 pkd 2.8 01:07.59 2 1 100 5268K
104 logd 2.8 01:16.22 6 5 2122- 6452K-
376 com.apple.Ap 2.4 00:44.55 8 7 307 1828K
5612 com.apple.We 2.0 00:39.01 9 1 73+ 450M-
537 trustd 1.7 00:42.20 2 1 218- 2468K-
526 ContinuityCa 1.4 00:49.61 9 4 349 4052K
340 daemon 1.4 00:37.31 15 3 284 74M

How did you get yours to paste as a table?
 
Engender said:
Okay, it looks like activity monitor but slower. Now what?
Click to expand...

The slower was deliberate. By using the "-s 60" option to top, it averages the usage over 60 seconds rather than Activity Monitor's 5 seconds maximum. This avoids the overhead of measurement from distorting the results significantly. Yes, there's kind of a Heisenberg uncertainty principle at work when doing some of these measurements...

So the questions now are:
1) what was your % idle (or the whole CPU usage line) once the system settled after boot?
2) If the system never seemed to settle and/or %idle never gets >98%, what are the top processes?
3) Did removing your startup apps change the situation?
 
Last edited:
Engender said:
Here is what I have, but note that I am working on writing a motion, so not at idle:
Click to expand...

Okay that won't quite give us an Apple to Apple comparison but understand of course work comes first. When you do get a chance, a reboot and test would provide a clean baseline.

In the meantime, I take it from the below that you are using a web-based tool to prepare that motion? Is it Office 365 Online or Google Docs? I use local Word so I can't speak to how either of those work in different browsers these days. In any case, assuming that is the only site open in the below, it does appear to draw a lot on your system. You might try accessing that site from Brave rather than Safari to see if it makes a significant difference. Not that you have to switch but just to see if that makes a difference...

Engender said:
Processes: 656 total, 6 running, 650 sleeping, 2754 threads

18:02:24 Load Avg: 5.51, 8.01, 9.31
CPU usage: 28.74% user, 17.78% sys, 53.47% idle
SharedLibs: 1047M resident, 157M data, 88M linkedit.
MemRegions: 1965 total, 60M resident, 4988K private, 3148M sh
PhysMem: 29G used (5211M wired, 0B compressor), 35G unused.
VM: 27T vsize, 5051M framework vsize, 0(0) swapins, 0(0) swap
Swap: 0B + 0B free
Purgeable: 4622M 42030(877) pages purged.
Networks: packets: 2521256/2726M in, 2006835/1423M out.
Disks: 531975/10G read, 209308/4948M written.


PID COMMAND %CPU TIME #TH #WQ #PORT MEM
173 WindowServer 40.5 10:11.86 13/1 6 4219 619M+
1043 Safari 21.9 07:58.63 18 10 2287- 176M+
146 launchservic 17.7 06:50.77 16 15 9767+ 4204K-
181 runningboard 8.8 03:27.29 9 8 790- 4448K
1 launchd 8.6 03:31.78 5 4 3812- 18M
1051 com.apple.We 8.4 03:35.27 10 5 154- 70M+
573 fontd 8.3 03:30.13 4 3 143- 3864K-
179 loginwindow 7.0 02:54.57 4 3 381 19M
483 lsd 6.4 02:31.41 3/1 2/1 606 25M+
2008 com.apple.We 5.5 00:23.22 9 4 82- 332M-
4655 com.apple.We 4.7 01:44.79 11 3 84+ 361M+
1052 com.apple.We 4.1 00:49.68 28 4 462+ 414M+
0 kernel_task 4.0 02:44.86 357/6 0 0 491M-
559 Activity Mon 3.7 02:04.53 7 5 1434 69M+
1013 Terminal 3.4 00:21.28 9 4 328+ 86M+
535 pkd 2.8 01:07.59 2 1 100 5268K
104 logd 2.8 01:16.22 6 5 2122- 6452K-
376 com.apple.Ap 2.4 00:44.55 8 7 307 1828K
5612 com.apple.We 2.0 00:39.01 9 1 73+ 450M-
537 trustd 1.7 00:42.20 2 1 218- 2468K-
526 ContinuityCa 1.4 00:49.61 9 4 349 4052K
340 daemon 1.4 00:37.31 15 3 284 74M
Click to expand...

Engender said:
How did you get yours to paste as a table?
Click to expand...

Oh yeah...actually had previously saved the output of top to a file, then cut the statistics lines off manually, processed it with a command called "awk", pasted it into Excel, formatted it there, and then pasted it and the statistics lines here...that part's not going to be worth your time reproducing...
 
No, I'm using Word to write. Although, I do use a web-based timer from Clio to track how long I'm working on the motion.

I've got Firefox, but not Brave. Do you think Firefox would work for testing purposes? Or does Firefox also use webkit?

I did get a kernel panic the other day that pointed to webkit . . .

Edit: Also, man awk?
 
Engender said:
No, I'm using Word to write. Although, I do use a web-based timer from Clio to track how long I'm working on the motion.
Click to expand...

You should monitor the system with and without that timer running. If a web-based timer is responsible for most of the load on your system that's crazy.

Engender said:
I've got Firefox, but not Brave. Do you think Firefox would work for testing purposes? Or does Firefox also use webkit?
Click to expand...

That would be a different test but still useful. Firefox uses it's own engine known as Gecko. Brave uses the Chromium engine. There's also Orion which uses its own snapshot of WebKit. Orion is like Safari+ and where I feel most at home. However, I've found Brave a bit faster especially on Intel systems. Plus I find some websites are a bit more optimized for Chromium-based browsers given their popularity on Windows systems. It shouldn't be that way but...if I come across a website that is driving my system crazy, I test it with Brave. But I also have Firefox with uBlock Origin and NoScript extensions if I want to really sandbox a site or get granular with its permissions. Anyway, the short of it is that if you work with a site that drives a lot of %CPU, it's worth seeing if that holds in others browsers (e.g. Firefox or Brave/Chromium/Chrome if you have them).

As an example, whenever someone makes me join a Teams meeting, I now join with Brave. I have the Teams app but I hate it so I refuse to use it unless absolutely required. Then Teams doesn't support Safari/WebKit so I used to use Firefox. But Teams in Firefox spun up the fans on my little laptop like a turbine. Accessing a Team meeting from Brave is so much smoother it's night and day. Though for most websites it's not that dramatic.

Engender said:
I did get a kernel panic the other day that pointed to webkit . . .
Click to expand...

Interesting -- WebKit shouldn't really be able to induce a kernel panic. That suggests a hardware issue or an Apple bug in that version of Sequoia.

Engender said:
Edit: Also, man awk?
Click to expand...

I can get a little old school...
 
bzgnyc2 said:
Interesting -- WebKit shouldn't really be able to induce a kernel panic. That suggests a hardware issue or an Apple bug in that version of Sequoia.
Click to expand...
I did the Apple Hardware test after the kernel panic, and it did not find any issues.

I did upgrade the RAM in my Mini . . .
 
OP:

Looking at the image you posted in #9 above...

My recommendation is:
TURN OFF SPOTLIGHT.
Kill it forever.
That will take care of item numbers 2, 3 and 4 in your Activity Monitor list above.

If you need to search for something, use either EasyFind or Find Any File.
Both are small and FREE.
And they work BETTER THAN Spotlight.

sudo mdutil -a -i off
(turns off spotlight)
You need to disable SIP (system integrity protection) first, then run the command.
 
Engender said:
Roger that.

For bzgnyc2, here is my activity a few minutes after boot:
View attachment 2486791

EDIT: Why is Fontd using 6.1% of CPU?​

Click to expand...

Interesting, though I would wait longer for it to settle -- which I've seen take up to two hours in the past starting with Catalina and later macoS -- and check the same.

However, I am curious if you had any tabs/windows open in Safari when you took that screenshot? If you did, the open website may have been running a JavaScript applet that was constantly scanning your font library, which seems a bit rude.

Otherwise likely "fontd" was just taking another pass through all your fonts for its own reasons and/or your font cache is corrupted. Then you might try booting in Safe Mode and then rebooting again normal:

Start up your Mac in safe mode - Apple Support

To help determine whether an issue is caused by software that loads as your Mac starts up, you can start up your Mac in safe mode.
support.apple.com support.apple.com

Clears some system caches, including font caches and the kernel cache. These are automatically created again as needed.​
​

If that still doesn't resolve everything you might start up Console (under Applications/Utilities) and let it stream. Perhaps start with filter on Errors and Warnings and see what kind of recurring errors/warnings you see. That your logd is running 2% and runningboardd 5.8% of a CPU is a little suspicious. Mine average 0.1% and 0% respectively.

Also your #1 activity in your snapshot was "launchservicesd". On my slower laptop that settles at 0.7%. Yours may be constantly trying to restart a service and failing. Which likely would also be generating lots of log entries which may explain the previous.


P.S.kernel panic, etc, I agree it's probably not driving your CPU temperature, etc so let's put that aside for now. It is an odd data point for now but I would assume it's a one-off unless it happens again.
 
Hey bzgnyc2, here is the activity from three hours after booting, without loading anything except iStat, TG Pro, Carbon Copy Cloner, Some Logitech app for my webcam, BetterDisplay, and Adobe Creative Cloud. Oh! And ClipMenu—an app that is really old but really, really useful.
1740862539390.png

EDIT: And the only fault I'm seeing on console is
sharedcache fault 16:04:10.218235-0500 com.apple.WebKit.WebContent checkinWithServer Failed bootstrap_lookup2 for name of coreservicesd, kern_return_t=#1100/0x44c Permission denied name=com.apple.CoreServices.coreservicesd

Okay, I'm also seeing an error with imagent and siriinferenced. Maybe I should nuke Siri?
 
Last edited:
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back