In this article, I’ll investigate on MacBook Pro 16” cpu throttling issues when using one or more external display.

Let’s start with the software and hardware specifications of this test:

Hardware:

• MacBook Pro (16-inch, 2019)
  • 2,3 GHz Intel Core i9 8 core
  • 32 GB 2667 MHz DDR4
  • AMD Radeon Pro 5500M 4 GB
  • Intel UHD Graphics 630 1536 MB
• Monitor Asus VP28U (3840 x 2160) plugged via Apple USBC to HDMI Connector model A2119
• Monitor LEN L27q-10 (2560 x 1440) plugged via Apple USBC to HDMI Connector model A2119

Software:

• macOS Catalina 10.15.6 (19G2021)
• iStat Menus 6.40
• Intel Power Gadget 3.7
• “yes” for the stress test
• htop to monitor cpu usage

Test environment:

For each test, I simply span 16 (8 cores + hyperthreading) “yes > /dev/null” in the terminal:

yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null & \
yes > /dev/null &

then wait about 1-2 minutes for the stats to “stabilyze”.

The temperature in the room is 27 C°

The setup is in this picture (I didn’t use the rightmost display in the tests) :

Test 1: notebook mode: no external display, lid open

Results:
3.2 GHZ
note that “radeon high side” is at 0 watt, since only the Intel gpu is used

Test 2:  1 * 4K external display + internal display (lid open)

Results:
2.6 GHZ
note that “radeon high side” is 20 watt, which is one of the highest value I saw during the test

Test 3:  1 * 4K external display, lid closed

Results:
2.9 GHZ
“radeon high side” is 10 watt

Test 4:  1 * QHD (2560 x 1440) external display, lid closed

Results:
2.5 GHZ
we see that the “radeon high side” is about 20 watt, so it’s normal that the performance are similar to Test 2 (a little bit worst, but may be related to the better thermal management with the lid open in the test 2?)

Test 5:  1 * QHD (2560 x 1440) external display *forced to 1080p resolution*, lid closed

Results:
2.8 GHZ
note that “radeon high side” is 10 watt, results similar to test 3

Conclusion:

in this test, you can see three different behaviors:

• Behavior 1: Radeon GPU not used: best thermal management and less cpu throttling
• Behavior 2: Radeon GPU “partially loaded” (10 watt): still decent thermal management , medium cpu throttling
• Behavior 3: Radeon GPU “fully loaded” (20 watt): high cpu throttling

It seems that the Radeon GPU usage (and then, the heat and the throttling) is impacted by:

• the usage of the internal display only (in this case, Radeon GPU is not used) causes Behavior 1, so best performances and less cpu throttling
• the usage of *only one* display, with the lid closed, and a resolution of 4K or FullHD, causes Behavior 2 (medium throttling)
• the usage of more than one display *or* a resolution different from 4K or FullHD (in my tests I used 2560 x 1440) causes Behavior 3, with the high cpu throttling

Other thoughts:

• it seems that opening the lid can improve a little bit the thermal management only when more than one display is used. Opening it when only one display is used, adds a second display, the Radeon goes in “Behavior 3” and the overall situation get worst

So What?

For the best performance, use your MacBook Pro *without ANY External display, use the integrated display only*.

If you need an external display, the best trade-off is with:

• only one external display
• lid closed
• resolution of 4K or 1080p

The worst performance are when:

• you use more than one display (two external or one external + integrated)
• you use a different resolution than 4K or 1080p (need to do more tests with different resolutions, I only tested 2560 x 1440 by now)

Let me know in the comment if you have any question or you have made similar tests with different results.

Bye

Fabio