mini Mac 2018 Throttling

[magicMac]

Thought i'd run some tests to see just how bad the throttling is, particuarly under different CPU and fan settings. Have shared the full results below.

What I have found is disabling TurboBoost doesn't have much of impact to multicore processes because the CPU is power limited when all cores are computing, and therefore remains closer to base clock. Disabling TurboBoost is a quick win in those cases but otherwise does introduce a big hit in performance for single threaded processes, since this is where the CPU turbo's the most.

So disabling intel turbo may not be an option for some; interestingly this single core drop equated to a significant drop in browser performance according to JetStream2 (although unnoticeable in most cases). I have personally settled leave Turbo enabled but forcing max fan speed, this has significantly boosted my multicore performance with the expensive of that contiuous noise.

But this got me thinking, silent operation is a nice thing and the miniMac is desperate for a cooling mod. This should be relatively simple considering the internal components are made up of just a power supply, Mainboard/RAM, Fan/Heatsink and an WiFi/Bluetooth Antenna. Has anyone else on the forums attempted a custom chassis with this machine yet ? As much as I like the small form factor, as the lowest cost mac I can't help but wonder how much of a better performing machine it would be when the components are removed and placed into a larger chassis with either a more effective fan/heatsink combo or a water-cooling setup.


Results
===

CPU: Intel i7-8700B paired with user installed crucial 32GB RAM
14nm 6-core chip with 3.2GHz Base freq. and 4.6Ghz Turbo freq.

Geekbench5 64bit + JetStream2 in Safari used for testing


Test 1
Turbo Enabled, Default Fan Speed:
singlecore 1083
multicore 5388
JetStream 175
Max Temp on Geekbench: 80~
Idle Temp: 50+

Test 2
Turbo Disabled, Default Fan Speed:
singlecore 810
multicore 5112
JetStream 115
Max Temp on Geekbench: 60~
Idle Temp: 43

Test 3
Turbo Disabled, Max Fan Speed:
singlecore 827
multicore 5186
JetStream 120
Max Temp on Geekbench: 47~
Idle Temp: 36

Test 4
Turbo Enabled, Max Fan Speed:
singlecore 1171
multicore 6227
JetStream 179
Max Temp on Geekbench: 52
Idle Temp: 36.5

Test 5
Repeat of test 4 after full OS reboot, closing all background apps and 30 mins of no activity
singlecore 1218
multicore 6256

 

[IngerMan]

I don't think the 2018 Mac Mini has a throttling or cooling issue. I just ran the test at default settings and it handles the temps very well, It got up to 83 C but only for a second and the fans ended up just over 3K, you can hear them very well over 3K but it was short and when the test ended it was 40's C in seconds. I don't see much of a benefit for performance blasting the fans full. My default score is very close to your Max fan score.

GB5 is not the program to test heat because of the sporadic bursts. Try Cinebench for your testing.

 

[magicMac]

I don't think the 2018 Mac Mini has a throttling or cooling issue. I just ran the test at default settings and it handles the temps very well, It got up to 83 C but only for a second and the fans ended up just over 3K, you can hear them very well over 3K but it was short and when the test ended it was 40's C in seconds. I don't see much of a benefit for performance blasting the fans full. My default score is very close to your Max fan score.

GB5 is not the program to test heat because of the sporadic bursts. Try Cinebench for your testing.

View attachment 952952

Cinebench is even worse for thermal throttling on the mac mini, but your'e right it's a better way to demonstrate the issue.

With turbo on - fan on max:
The chip can only sustain 3.49GHz to keep within the 100C limit, it only peaks at 4.6Ghz for 2-3 seconds at the start of the test and the nose dives as the temperature hits 100C almost immediately.
-CPU remains steady around 3.49Ghz
-Chip draws ~70W and maxed out temp at 100C.
Score: 2797

With turbo off - fan on max:
The chip runs at 3.2GHz - which is the base clock of i7-8700B
-CPU remains fixed at 3.2Ghz
-Chip draws 55W max and temps gradually increasing until reaching 81C max towards the end
-Repeating this test with fan set to normal, I get score 2478 and temps gradually increasing until ~95C towards the end
Score: 2487

So with Turbo on, I only sustain 300MHz higher clock rate yielding 12.5% more performance, yet it's reaching boiling point of water and using 28% more power.

Ultimately this is not just about performance but the lifespan of the computer as well.

If you look around the web you'll see I'm not the only one this appears to be a common issue with the recent series of Intel chips stuck on 14nm tech. It definitly has a throttling issue.

 

[IngerMan]

I ran the App after I suggested it, its been a while but I scored over 2900 and the temps never exceeded 85C and ran in the low 80’s. My fan speed did top 4K midway through the test, I do have a $5 Nupad I am sitting on which may help for air flow. I also know mine is dust free for I just upgraded the ram. If your maxing out at 100C then that’s a good 15C higher than mine? 2915 was not my highest score, I bested over 3000.

There is also another thread that talks about using this extra fan for heat management.

SPEED Mini Cooling Base | SPEED Designs

The ultimate cooling solution for your Mac mini. Keep your Mac mini cool with your heaviest video editing and business demands.

www.speed-designs.com

 

[magicMac]

I ran the App after I suggested it, its been a while but I scored over 2900 and the temps never exceeded 85C and ran in the low 80’s. My fan speed did top 4K midway through the test, I do have a $5 Nupad I am sitting on which may help for air flow. I also know mine is dust free for I just upgraded the ram. If your maxing out at 100C then that’s a good 15C higher than mine? 2915 was not my highest score, I bested over 3000.

An i7-8700 should reach 3547 on cinebench when it's able to properly use the 4.3GHz "6-core turbo speed". Also keep in mind you have an eGPU there, but yeh the base is a good idea probably helps with airflow a little bit. Can you run the test with Intel Power Gadget in the background as it will be interesting to see the actual core temps and sustained turbo speed during the benchmark ?

 

[IngerMan]

An i7-8700 should reach 3547 on cinebench when it's able to properly use the 4.3GHz "6-core turbo speed". Also keep in mind you have an eGPU there. Can you run the test with Intel Power Gadget in the background as it will be interesting to see the actual core temps and sustained turbo speed during the benchmark ?

Will do, but I am done tonight on the MM. I will post it tomorrow. If you would like something else let me know before the AM, I already have it back in the freezer preparring for the next test 😂. That was a joke....

I am not saying that heat is not a factor in the MM, heat is a factor in all apple computers. I have had many, and this MM has performed better for me in regards to heat than my previous 2 iMacs, and all my many Apple laptops. The only Apple device that has performed better in regards to heat is the 2020 ipad Pro I am typing on. But that is another story.

Speed mini cooling thread starts post #238

Cooling solutions for the Mac minis?

Here is another option. https://www.speed-designs.com/speed-mac-mini-cooling-base

forums.macrumors.com

 

[frou]

Turbo clocks beyond a single loaded core are not even an advertised spec by Intel themselves: https://www.techpowerup.com/237740/on-intels-decision-to-no-longer-disclose-all-core-turbo

So while it's kinda interesting to think how the chip performs under ideal cooling conditions (the 4.3 GHz all-core figure), it's pretty categorical that there isn't anything Going Wrong.

 

[magicMac]

Turbo clocks beyond a single loaded core are not even an advertised spec by Intel themselves: https://www.techpowerup.com/237740/on-intels-decision-to-no-longer-disclose-all-core-turbo

So while it's kinda interesting to think how the chip performs under ideal cooling conditions (the 4.3 GHz all-core figure), it's pretty categorical that there isn't anything Going Wrong.

I do know the 4.3Ghz figure is unlikely to be achieved for a long time, as is the 4.6Ghz single core turbo figure. But if you look at other desktop PC's with the same CPU it is able to reach around a steady 3.8-4GHz figure. I'm not saying the mac mini has an especially bad thermal design, after all aluminium dissipates heat quite well. This is largely an Intel thing but it can be mitigated by bringing temps down as much as possible to either prolong the life of the chip or to maximise the turbo benefits.

Interesting project here:

Project | Mac mini 2018 Hexa Core Cooling Analysis | Hackaday.io

hackaday.io

 

[IngerMan]

I ran the test with Intel power gadget. Frequency peaked at 4.+ in the beginning and sustained around 3.6 +/- .1 for the remainder. My temps was in the high 90's on intel oddly my iStat showed (istat top menu bar temp)in the mid 80's, obviously a different sensor I have picked on iStat. Scored 2922 running both apps.

update: Checked my istat sensor and I was using CPU Proximity, changed it to CPU PECI and it matches the Intel temps now which I confirmed after rerunning the test.