(2019) 16" is HOT & NOISY with an external monitor! :(

[Corncab44]

So funny that this all comes down to bugs that apple refuses to fix. Give me power and fans, but don't just run up the wattage for no reason...

 

[Achever]

That's in line with what I saw. Single monitor in clamshell was OK. Once I added a second monitor (exactly the same brand, settings and connected the same way - displayport via a different thunderbolt bus on other side of the MBPro) the power usage jumped until I rotated one of the monitors.

Rotating both monitors caused power usage to drop also. My educated guess (as a software developer) was that the rotation was triggering a different path through the downclocking decision logic and thus maybe bypassing a bug. Others have reported better behaviour in Bootcamped Windows with exactly the same setup they had for MacOS. I never tried this myself before returning the 5500M. But it is more evidence that behaviour may be partially due to a bug in the video drivers but that said maybe the scaling behaviour in MacOS means the GPU is driven differently.

The 5600M can drive these two displays (in standard rotation) at less than 10W suggests some sort of bug also. As I type this email I'm driving the 2 x 4K displays, the internal display plus a 1920x1200 display and the 5600M power usage is under 9W. This arrangement was 24W+ on the 5500M. 5600M stats below:

View attachment 934896

Thanks for the Info! I am getting the 5600m now

 

[Norbert Mikołajczyk]

That's in line with what I saw. Single monitor in clamshell was OK. Once I added a second monitor (exactly the same brand, settings and connected the same way - displayport via a different thunderbolt bus on other side of the MBPro) the power usage jumped until I rotated one of the monitors.

Rotating both monitors caused power usage to drop also. My educated guess (as a software developer) was that the rotation was triggering a different path through the downclocking decision logic and thus maybe bypassing a bug. Others have reported better behaviour in Bootcamped Windows with exactly the same setup they had for MacOS. I never tried this myself before returning the 5500M. But it is more evidence that behaviour may be partially due to a bug in the video drivers but that said maybe the scaling behaviour in MacOS means the GPU is driven differently.

The 5600M can drive these two displays (in standard rotation) at less than 10W suggests some sort of bug also. As I type this email I'm driving the 2 x 4K displays, the internal display plus a 1920x1200 display and the 5600M power usage is under 9W. This arrangement was 24W+ on the 5500M. 5600M stats below:

View attachment 934896

Nice, clamshell mode, right?

 

[sco]

Nice, clamshell mode, right?

The screenshot is 5600M not in clamshell - one of the displays is the internal. For the 5500M numbers anytime I used the internal display plus an external it was 20W+.

 

[Norbert Mikołajczyk]

The screenshot is 5600M not in clamshell - one of the displays is the internal. For the 5500M numbers anytime I used the internal display plus an external it was 20W+.

Awesome stats.

 

[x3sphere]

I noticed that rotating the display changes the refresh rate in some cases - so for those reporting less power usage you may want to check that. I normally run my monitor at 85Hz which gives 18W usage, with it rotated it went back down to 60Hz at 5W usage. But in my case, it wasn't the rotation that fixed anything, only that changing the rotation simply dropped the refresh rate. It wasn't apparent until I actually checked.

 

[yurc]

This problem also exhibit in previous model, such as 2015 15 inch MBP (my device). My work room has no air conditioner and ambient around 32-33 degree.

I made DIY fan cooler from acrylic sheet, similar like typical laptop cooler, outfitted with twin Noctua NF-A12 because they can spin in high rpm without being noisy. I am using USB converter as power and it limited to 5V, but is adequate to produce airflow. Alternatively using single 180mm Noctua model would also works.

Put MBP on top and I left plugged with Dell 4K display through USB-C to DisplayPort cable, from iStats log average temperatures shows around 65 degree, and built in MBP fans no longer producing jet sound anymore compared without assisted cooling ( I prefer this way)
I am running in scaled Hi-DPI mode a.k.a retina (1920x1080@2x). Previously after 30 minutes it starts to warm and fan going full blast.

Apple won’t give a damn when we kvetching about thermal issues, so i act on my own complaint by making ghetto cooling device…

 

[sco]

I noticed that rotating the display changes the refresh rate in some cases - so for those reporting less power usage you may want to check that. I normally run my monitor at 85Hz which gives 18W usage, with it rotated it went back down to 60Hz at 5W usage. But in my case, it wasn't the rotation that fixed anything, only that changing the rotation simply dropped the refresh rate. It wasn't apparent until I actually checked.

I was at 60Hz here irrespective of rotation. I do suspect varying refresh rate could have an impact. From what I have read downclocking of the GPU happens during vertical blanking and refresh rate changes would affect timing of that.
[automerge]1595067517[/automerge]

I made DIY fan cooler from acrylic sheet, similar like typical laptop cooler, outfitted with twin Noctua NF-A12 because they can spin in high rpm without being noisy. I am using USB converter as power and it limited to 5V, but is adequate to produce airflow. Alternatively using single 180mm Noctua model would also works.

Good one I tried a Thermaltake Massive V20 here with the 5500M and it did’t help.

My previous laptop was a 2015 and had to keep ambient under 24 deg. C to stop it throttling.

 

[pr33tz]

Am I right in thinking the current state is:

5300/5500 + int display = running normal
5300/5500 + ext display = running normal
5300/5500 + int display + ext display = running hot
5600 + any config = running normal

Is this thought to be a hardware issue or something that could eventually be fixed via an update?

 

[mrmachine79]

Am I right in thinking the current state is:

5300/5500 + int display = running normal
5300/5500 + ext display = running normal
5300/5500 + int display + ext display = running hot
5600 + any config = running normal

Is this thought to be a hardware issue or something that could eventually be fixed via an update?

Correct. Hardware, by design.

 

[09872738]

Correct. Hardware, by design.

Some report the issue is not present under Bootcamp/Windows, what strongly suggests it might be a driver issue, therefore at least partly software related

 

[Achever]

Some report the issue is not present under Bootcamp/Windows, what strongly suggests it might be a driver issue, therefore at least partly software related

I think we need to pressure Apple to demand no worse performance in comparison with windows/bootcamp
Affected people should open a support ticket and reopen it again and again until Apple starts to move and we see similar driver performance

 

[Corncab44]

Correct. Hardware, by design.

If it's purely a hardware issue, then how do we explain these anecdotal reports of benefits seen by switching refresh rates, rotating displays etc.?

 

[mrmachine79]

Some report the issue is not present under Bootcamp/Windows, what strongly suggests it might be a driver issue, therefore at least partly software related

Not really. It's present in windows because (elevated baseline power consumption). The baseline heat is just lower because windows does not use as much GPU to render every element of the OS interface like macos does.
[automerge]1595080815[/automerge]

If it's purely a hardware issue, then how do we explain these anecdotal reports of benefits seen by switching refresh rates, rotating displays etc.?

Those reports are only in clamshell. 100% internal plus external reports confirm high idle power consumption and every single reported response from apple engineering, no matter how high it is escalated, is "working as designed".

This behaviour has been present since forever with dGPU macbooks and in amd desktops. It's just worse in the 16 inch because it's a faster GPU running faster memory clock producing more heat. Hbm2 memory used less power and runs at a lower frequency and therefore produces less heat.

Even if there's a bug where some single displays in clamshell also exhibit high power consumption, fixing that won't fix high power consumption with external and lid open. If you're happy to only use clamshell with external displays, it works half the time already and you have options.

For lid open, eGPU or 5600m are the only options if you want to uncripple the CPU or fan noise bothers you.

It's been 8 months since 16 inch was released and years and years since the first macbooks with and dGPU. A software fix is not coming. No matter how much people wish or hope or complain.

 

[DaveXX]

For lid open, eGPU or 5600m are the only options if you want to uncripple the CPU or fan noise bothers you.

It's been 8 months since 16 inch was released and years and years since the first macbooks with and dGPU. A software fix is not coming. No matter how much people wish or hope or complain.

As i said this is not a MBP 16 problem. Its the same with MBP 15 2018. The only difference may that the MBP 16 will speed up the fans sooner. So with TG Pro its no problem to get a different behaviour if you want but the problem is the same.

 

[applesed]

Not really. It's present in windows because (elevated baseline power consumption). The baseline heat is just lower because windows does not use as much GPU to render every element of the OS interface like macos does.
[automerge]1595080815[/automerge]

Those reports are only in clamshell. 100% internal plus external reports confirm high idle power consumption and every single reported response from apple engineering, no matter how high it is escalated, is "working as designed".

This behaviour has been present since forever with dGPU macbooks and in amd desktops. It's just worse in the 16 inch because it's a faster GPU running faster memory clock producing more heat. Hbm2 memory used less power and runs at a lower frequency and therefore produces less heat.

Even if there's a bug where some single displays in clamshell also exhibit high power consumption, fixing that won't fix high power consumption with external and lid open. If you're happy to only use clamshell with external displays, it works half the time already and you have options.

For lid open, eGPU or 5600m are the only options if you want to uncripple the CPU or fan noise bothers you.

It's been 8 months since 16 inch was released and years and years since the first macbooks with and dGPU. A software fix is not coming. No matter how much people wish or hope or complain.

sorry, what are the options for clamshell + external = no high power draw? Do you mean 5300 or the 13 incher?

 

[Corncab44]

sorry, what are the options for clamshell + external = no high power draw? Do you mean 5300 or the 13 incher?

There are no options to achieve this on the 16, unless you pay 600-700 for a 5600.
[automerge]1595088499[/automerge]
New strategy -- Ignore the fans. It's liberating. Turn up music really loud and feel them purrrrrrrr.

 

[bomby0]

There are no options to achieve this on the 16, unless you pay 600-700 for a 5600.
[automerge]1595088499[/automerge]
New strategy -- Ignore the fans. It's liberating. Turn up music really loud and feel them purrrrrrrr.

I can hear the 5500 rpm vacuum cleaner through my headphones

I think the real lesson here is Apple using discrete GPUs is a disaster. If I stick to Mac, there's no way I'll ever get a Mac with a dGPU. I miss my 2013 13" MBP

 

[09872738]

Not really. It's present in windows because (elevated baseline power consumption). The baseline heat is just lower because windows does not use as much GPU to render every element of the OS interface like macos does.

Sounds plausible. Are you in a position to explain what changed in M5600 equipped models (where users report very different power consumption numbers)?

 

[bill-p]

5600M uses a completely different memory architecture (HBM2). Likely, that's the reason why it doesn't see this issue.

Apple has historically never really addressed any software issue with dGPU. Given that they are now pushing for ARM, I'd guess they want to do away with dGPU and eGPU altogether. So that gives them even less incentive to work on this issue even if it's just a software fix.

We honestly just have what we have. Take it or leave it.

 

[mrmachine79]

sorry, what are the options for clamshell + external = no high power draw? Do you mean 5300 or the 13 incher?

There are many reports of people having low power consumption with lid closed (clamshell) and an external display, either using the right display or cable or using switchresx to change refresh rate to get low power draw.

What there is no option to fix is high power draw when using the internal display and an external display.

In this configuration the dgpu is forced to active to drive the external display, and memory is set to max speed by design to avoid graphical corruption because down clocking the memory can only occur during v-blank and there are issues timing that with multiple displays at different refresh rates.

Something about rendering an app on dgpu and displaying it on the igpu or something about the internal display refresh rate is triggering this behaviour with lid open that can't be fixed with switchresx.

Sounds plausible. Are you in a position to explain what changed in M5600 equipped models (where users report very different power consumption numbers)?

The issue is caused by memory speed being locked at max which uses more power and generates more heat. The 5600m uses hbm2 memory which is faster at a lower clock speed and uses 3-4 times less power, therefore even when locked at max speed it uses less power and generates less heat.

An egpu will cost less and use even less power and generate even less heat inside the MBP to give you maximum CPU performance and near constant min fan speed under medium load, but you will have to quit and relaunch some apps (happens automatically after clicking a disconnect button) before you unplug.

 

[mrmachine79]

As i said this is not a MBP 16 problem. Its the same with MBP 15 2018. The only difference may that the MBP 16 will speed up the fans sooner. So with TG Pro its no problem to get a different behaviour if you want but the problem is the same.

Agree. This has been an issue with all macbooks with and dGPU. Earlier dGPU models were just slower and used less power and generated less heat at max memory speed. I found that the fans were also frequently at max and Radeon power draw was high (but not quite as high, about 15W) on my old 2019 15 inch. And performance is much better on the 16 inch even with its hotter dGPU thanks to its 12W better cooling capacity.

 

[arche3]

2 external monitors macbook pro in clamshell. seems ok I suppose. this is at idle. 5600m graphics

 

[sco]

2 external monitors macbook pro in clamshell. seems ok I suppose. this is at idle. 5600m graphics

So similar figures to mine with 4 displays - so 5600M idle at full clock speed looks to be around 11W. Suggests that the HBM does make a difference as others have suggested. I agree with suggestions that hardware limitations are having an impact but I still think in some cases software issues are at play here - especially in the rotation scenario that reduces power usage.

 

[chaiboy09]

2 external monitors macbook pro in clamshell. seems ok I suppose. this is at idle. 5600m graphics

Nice. Can you share the laptop lid open readings of Radeon wattage. Also what are the fan speeds?

For me with 5500 I am getting 9 W\1900 RPM with one 4K monitor in clamshell mode. Need to see if that shoots up if I add second monitor to 18 W\3000 RPM (currently this the reading I see with Laptop lid open and connected to 4K).
[automerge]1595131513[/automerge]

That's in line with what I saw. Single monitor in clamshell was OK. Once I added a second monitor (exactly the same brand, settings and connected the same way - displayport via a different thunderbolt bus on other side of the MBPro) the power usage jumped until I rotated one of the monitors.

Rotating both monitors caused power usage to drop also. My educated guess (as a software developer) was that the rotation was triggering a different path through the downclocking decision logic and thus maybe bypassing a bug. Others have reported better behaviour in Bootcamped Windows with exactly the same setup they had for MacOS. I never tried this myself before returning the 5500M. But it is more evidence that behaviour may be partially due to a bug in the video drivers but that said maybe the scaling behaviour in MacOS means the GPU is driven differently.

The 5600M can drive these two displays (in standard rotation) at less than 10W suggests some sort of bug also. As I type this email I'm driving the 2 x 4K displays, the internal display plus a 1920x1200 display and the 5600M power usage is under 9W. This arrangement was 24W+ on the 5500M. 5600M stats below:

View attachment 934896

Can you please post the radeon\Fan reading for single monitor in clamshell mode with 5600? I am seeing 9W with my 5500 in clamshell for single 4k Monitor.....wanted to see the power consumption of 5600....