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

[Syther101]

The AMD Radeon RX 580 & RX 570 Review: A Second Path to Polaris

www.anandtech.com

Very interesting article—scroll down to the part about external monitors. It seems like there is a basic Radeon limitation with respect to memory clock switching. The GPU can lower or increase the memory clocks based on load. Normally this happens during the screen refresh, to avoid tearing. When there are multiple monitors with different configurations, there is no predictable moment both monitors will be blanked. So the driver keeps the memory clock at full tilt at all times.

Damm yeah I reckon this is it you know ?. If you scroll further down interestingly the article even explains that if you match your monitor to "correct timings" you'll avoid the issue.

It could certainly explain why certain monitor configurations have been reported to work fine. It could even explain why Apple support is claiming this to be "as intended", if this is indeed a limitation of AMD's architecture.

I'm a recent purchaser in a unique position of still owning my 2017 15" model with a 480. I've been monitoring this thread and others like it for the past week. I've done direct comparisons with the exact same workload on both machines connected to my Dell monitors. One monitor for each machine, both the same monitor.

As expected the GPU is pegged at 18w constantly vs 9w of the 2017 model. Temps are also 20 degrees higher when idle.

This feels like a good itch to scratch. But I'm now past my 14 day return window and pretty miffed that in a machine with better thermal architecture, Apple would go and negate those improvements by using a shoddy GPU. Don't get my wrong I can see how they aren't left with much choice due to their relationship with Nvidia. But still!

 

[wegster]

Hi everyone, I read the past few 100 posts in this thread, and I still have a question. I'm hoping someone may post their experience.

I sold my 2016 15" MBP i7 with Radeon Pro 455 to get a new 2020 13" MBP i7. Unfortunately the iGPU is too slow for me, so I'm planning to return it to get a 16" (I could have planned that better from the beginning, but hey, mistakes happen).

I'm using my MBP with an external LG 4K Display scaled to 2560x1440 HiDPI and I leave the MBP open (no clamshell).

The 2016 MBP used to become quite hot in that configuration and fans always were at around 2500 rpm. The 2020 13" MBP is cooler and fans are around 1600 rpm (because of the lack of dGPU).

Now, I'm not really interested in power consumption. I know the 5500M draws too much power and people are complaining. What I'm interested in is fan noise.

What's your experience with this? Do fans get even higher than the 2500 rpm I used to get on my 2016 MBP in that screen configuration? I'm worried that it might be noisier than my old MBP. If it's about the same, I can manage. Louder would be too much.

Thank you for your insights!

More or less this is effectively 'idle' for me - fans will sit between 1800/min and ~3500RPM.
3200RPM isn't much noise on the current setup, more like a low background hum, but it does ramp up on noise as they go higher. 3500 starts to become noticeable..

Note - 'idle' in the case has a fair amount of things open, but I'm just waking it and not active in numerous apps at the moment - Teams, iTerm2 with a few tabs, Outlook, Mail, Safari, Chrome, OneNote, Preview, Painbrush, PowerPoint, Xcel, Messages.. I usually have at least one VM running and various other apps, so it's hard to say, but this is still effectively 'idle' in my case as I'm not doing any active or intensive work at the moment.

Lid open in defaults at the moment - looks like 1792x1120 on the MBP retina screen, and 3840x1600 on the external LG 38UC99 via MBP -> TB3/USB-C CalDigit dock -> display (tried other combos, no change to power draw).

While I was typing this and screenshooting, I quit a few apps (Teams is often a pig, and PPT), and fans dropped down to ~2.6K, which you can hear, but at least with it sitting on a stand next to my display, is nearly silent, for me at least.

Fan history over past week, using left side as the one that spins faster (power on left side. I know some have claimed move it to the right side, but I saw no appreciable change, just the right side would then have the left side #s..)

 

[matram]

The AMD Radeon RX 580 & RX 570 Review: A Second Path to Polaris

www.anandtech.com

Very interesting article—scroll down to the part about external monitors. It seems like there is a basic Radeon limitation with respect to memory clock switching. The GPU can lower or increase the memory clocks based on load. Normally this happens during the screen refresh, to avoid tearing. When there are multiple monitors with different configurations, there is no predictable moment both monitors will be blanked. So the driver keeps the memory clock at full tilt at all times.

This article is about the old GPU architecture used in the 15”. The 16” uses the new RDNA / Navi architecture.

 

[bill-p]

This article is about the old GPU architecture used in the 15”. The 16” uses the new RDNA / Navi architecture.

That section is not just about GPU architecture, but about a conscious decision made by AMD to employ a memory clock strategy to "reduce power consumption". In practice, the strategy didn't really help reduce power consumption at all, and we may be seeing the more dramatic side effect of it with the 16" MacBook.

But it also confirms that perhaps the 2018 - 2019 15" MacBooks are also affected by this issue. The 16" is not alone.

 

[Syther101]

Don't get me wrong. I've certainly not got a leg to stand on, but it would certainly explain things. All I know for certain is that there are glaring differences between my 2017 15" and this new model and the information in the article sounds awfully similar to experiences in this thread and my own.

That section is not just about GPU architecture, but about a conscious decision made by AMD to employ a memory clock strategy to "reduce power consumption". In practice, the strategy didn't really help reduce power consumption at all, and we may be seeing the more dramatic side effect of it with the 16" MacBook.

But it also confirms that perhaps the 2018 - 2019 15" MacBooks are also affected by this issue. The 16" is not alone.

As for the 2018/2019 models. This is something I'd be very interested in trying to work out. I didn't post earlier but here are my results directly comparing my 2017 vs 2019 models. One with the 560 and one with the 5500m. Both connected to 1440p Dell ultrasharps. Both with internal displays active.

These machines have the exact same applications, as the 16" is replacing the 15" once sold and I've just finished migrating.

The workload pictured is Safari with youtube video playing. Dropbox, Slack and Docker all open in the background.

The 2 year our model is smashing the 16" on the idle temps. You can see the wattage increase and also a 10 degree difference in the temp recorded for the GPU memory which is interesting to see:

(left 15", right 16")

 

[bill-p]

Just an observation: on the 16", your CPU computing high side is roughly 12W higher than on the 15". That means something is taxing your CPU and causing it to not be "idle".

These are my idle temps after I've opened the lid for 15 minutes. My room is at 27 Celsius right now. Temperatures are rising in San Jose, CA recently.

 

[impulse462]

Thank you.

Can anyone tell me what typical fan rpm values are for an idle 16" MBP drawing the infamous 18W for GPU?

It’s around 2300-2600 connected to an external with the laptop lid open with Radeon voltage at 18-20W

 

[silvetti]

Thank you.

Can anyone tell me what typical fan rpm values are for an idle 16" MBP drawing the infamous 18W for GPU?

Here, see (with evidence, and not just some random letters )

 

[Syther101]

Take a read through some of these posts on reddit:

🔎 RDNA multiple monitor memory clock - Google Search

www.google.com

These are desktop users, mentioning the same things. This post here and a few others I've seen mention "monitor timings" and something the AMD GPUs do to stop display flickering occuring.

With current AMD Radeon GPUs if you connect two monitors with different timings/resolution/refresh rate/bit depth, memory frequency will be increased to the maximum speed. There's nothing that can be done with it.

If your monitor can run at a compatible timing, like 60Hz, 120hz, etc. The memory is not required to be running at full clock speed and thus the wattage goes down to normal.

I've actually been able to confirm this. At full resolution my Dell Ultrasharp refresh rate is locked at 59.88 Hz. If I set the monitor temporarily to 1080p, it allows for the 60 Hz option to be picked. Now take a look at the Radeon High Side. It's only 5w!

This is with only a single monitor to add more info. Two monitors at 1080p 60Hz and I get the 17/18w of usage. Which is kinda strange. These cases don't directly translate and I can only hope this is a software thing. But again, it all sounds awfully similar.

 

[matram]

Very interesting, thanks for sharing!

 

[Alexandros Mystridis]

Only way I can get the Radeon High side down to 5W is by using 3440x1440@100Hz in clamshell mode. Display is AW3418DW connected with displayport(HDMI can only raise it up to 50Hz)
All other combinations draw ~18W and overheat my Macbook Pro 16 (32GB, i9, 5500M)

 

[lukelol]

Latest (and final) response from Apple:

Hello Luke,

I’ve received an answer from my engineers with that additional data that we sent up and they’ve stuck to their initial response [expected behavior]. They’ve also informed me that they are not aware of any optimization issues present in those models or the OS at this time.

I apologize that we didn’t get the answer that we had hoped for.

Let me know if you have any additional questions.

I'll be going with an eGPU as I'm outside of the return window and I really like the machine aside from the noisy heat. Apple was unable to offer any discount on the eGPU they have on apple.com

 

[jc_9]

Really is a shame, the Macbook Pro 16" would have been perfect aside from this issue. Apparently people have the same problem with external monitors with AMD desktop GPUs also. Really sucks for us that nVidia is no longer a competitor for Apple products...

 

[AFPBoy]

Apparently people have the same problem with external monitors with AMD desktop GPUs also. Really sucks for us that nVidia is no longer a competitor for Apple products...

Plot twist: The MacBook Pro 16" released this July will have an AMD CPU and an Nvidia GPU. ?

 

[k2so]

that's a shame for product,i had many searches about overheating problem and no one knows official solution,apple have to recall these devices.

 

[Syther101]

Is there anybody here who when running their Macbook in clamshell, using just one monitor, is able to select 60 Hz and confirm no high wattage?

I've today purchased a monitor that I know truly supports 60 Hz unlike the Dell Ultrasharp. All the ultrasharp series, afaik will show locked to 58.88 Hz on modern Macs.

 

[maxsquared]

Is there anybody here who when running their Macbook in clamshell, using just one monitor, is able to select 60 Hz and confirm no high wattage?

I've today purchased a monitor that I know truly supports 60 Hz unlike the Dell Ultrasharp. All the ultrasharp series, afaik will show locked to 58.88 Hz on modern Macs.

I am not sure if this is what you are referring to.

I am using Dell Ultrasharp U2720Q which connected via USB-C and 60Hz is present. And also, not sure if it's related, my 16 inch runs normal 90% of the time, would USB-C connection anything to do with it? I used to use Ultrafine 5K (via Thunderbolt) and my 16 inch was relatively fine too.

 

[HarryPot]

Just updated to 10.15.5, and something changed.
I don't have iStatsMenu, but my computer lasted longer before fans going up. And they haven't gone above 3500RPM.

But not sure if this Fan RPM and temperature makes sense.
It seems 57C is quite cool, but fans are still going up to 3500RPM.

This is with my MBP with the lid open, and connected to an external 27" 1080p Samsung monitor.

 

[davidbend]

Just updated to 10.15.5, and something changed.
I don't have iStatsMenu, but my computer lasted longer before fans going up. And they haven't gone above 3500RPM.

But not sure if this Fan RPM and temperature makes sense.
It seems 57C is quite cool, but fans are still going up to 3500RPM.

This is with my MBP with the lid open, and connected to an external 27" 1080p Samsung monitor.

View attachment 918996

I don't think there is any improvement in this version. They didn't mention anything about it in the release notes, moreover, 3450 rpm is too high unless you did some heavy tasks before you captured those stats.

 

[AFPBoy]

I've today purchased a monitor that I know truly supports 60 Hz unlike the Dell Ultrasharp. All the ultrasharp series, afaik will show locked to 58.88 Hz on modern Macs.

I have an older model (Dell U2713HM), and it's locked at 59.88 Hz. Is that close enough to 60 Hz, or do you think it makes a difference (i.e. cause >7 W power draw in clamshell)?

 

[Tnanman]

Plot twist: The MacBook Pro 16" released this July will have an AMD CPU and an Nvidia GPU. ?

Honestly at this point, that would be the dream Macbook Pro... a man can dream!

 

[stevenzli]

With current AMD Radeon GPUs if you connect two monitors with different timings/resolution/refresh rate/bit depth, memory frequency will be increased to the maximum speed. There's nothing that can be done with it.

I've actually been able to confirm this. At full resolution my Dell Ultrasharp refresh rate is locked at 59.88 Hz. If I set the monitor temporarily to 1080p, it allows for the 60 Hz option to be picked. Now take a look at the Radeon High Side. It's only 5w!

Thanks for sharing your findings. Just want to report my testing result using 16in MBP (base model, 5300m, lid open) and LG Ultrafine 4k (21.5in) connected directly using USB-C cable.

Setting the MPB and external monitor to the same fresh rates (60hz, 48hz) at various resolutions could not get the Radeon High side down to 5W. I tried exactly as you mentioned and set LG Ultrafine 4k to 1080p at 60Hz, with MBP display at default but still no luck.

 

[mrmachine79]

I got a 16" 2.4GHz 5300M to replace my 2019 15" 2.3GHz 560X, which was suffering from near constant max fans (5900RPM left side) with an external display and occasional kernel_task spikes to 1000% freezing up the whole system for 10+ seconds at a time.

I ran some tests with both machines side by side. Both plugged into their respective wall chargers. I ran tests with the internal display only and with identical LG 27UK850-W monitors connected (one each) via the USB-C cable supplied with the monitors. Both were running a fresh install of Catalina with just Intel Power Gadget, iStat menus, Cinebench and Luxmark installed.

The numbers quoted below are the lowest sustained values from the Intel Power Gadget graphs, to give an indication of minimum performance.

After ~10 Cinebench passes, the 16" CPU can sustain about 18W more power (60W vs 42W) and 500MHz higher frequency (3.3GHz vs 2.7GHz) under load. Cinebench results were about 25% higher (3450 vs 2759).

Both take a dramatic dive in power to 29W / 1.8GHz (16") and 26W / 1.7GHz (15") when the CPU and dGPU are concurrently under high load (Cinebench and Luxmark).

Both are similarly noisy at max fan speed. The 16" is maybe a *tiny* bit louder, but really not much in it. 54dB vs 52dB at 2 inches, according to my iPhone.

Just attaching the external display at "looks like 2560x1440" scaled resolution, the 16" dGPU uses 19W and the 15" dGPU uses 12W. Sustained CPU performance drops to 49W / 3GHz (16") and 39W / 2.5GHz (15").

When running a 1-1 video conference in Safari (Discord) and exercising the dGPU with the mission control animation, the 16" CPU remains stable and unchanged at 49W / 3GHz, but the 15" CPU drops further to 32W / 2.2GHz (below base).

Adding a Razer Core X Chroma eGPU and Sapphire Pulse RX 5500 XT drops the dGPU to ~5W, unless you're running apps that still use it. I configured Unity to use the dGPU even with the eGPU connected and it was using ~15W.

The stock Razer Core X Chroma is a little louder than both MBPs with min fans, but much more quiet than both at max fans. I replaced the eGPU case and power supply fans with Noctua fans (an easy job, no soldering or other permanent modifications required) and now it is inaudible.

I also tested a normal work morning connected directly the monitor while working during a long conference call, which might have triggered 1000% kernel_task on the old 15", with no issues.

Here are some screenshots to illustrate fan speed during a typical non-intensive work session (Chrome, Slack, Terminal, Unity, VS Code, no compiling):

And the corresponding dGPU power draw:

I was had the MBP connected directly to the external display until ~10:41. Then I switched to the eGPU but still ran Unity on the dGPU. Then I configured Unity to run on the eGPU, since which time the fan speed has been a flat line at ~1800 RPM.

Overall, I'm happy with the 16" so far (it's only been 1 day). The eGPU was an expensive work around, but it allows the CPU to run at its best, at the cost of having to disconnect it and quit some apps before I unplug.

Even without the eGPU, fan speed appears to be lower than max for a typical work, and I have not seen any kernel_task issues, yet.

Now that I have the and typically use eGPU though, I may just never know if that was truly resolved with the 16" alone.

 

[maik_is_here]

Now that I have the and typically use eGPU though, I may just never know if that was truly resolved with the 16" alone.

Thanks for testing and documenting. That took you a while! Nice to see the progress between 15 and 16. However, shows we still have to go a while.


question to all:

is there a verified, reproducible combination of resolution and refresh rate that results in 5W use, independent of monitor type?

Or is all this a stroke of luck that involves monitor brand, adapter type, zodiac and sock color of the day?

Thanks for any advice! Been trying to connect a full HD monitor to my mbp16 but am unable to get the promised land of 5W.

 

[joelhinch]

Thanks for testing and documenting. That took you a while! Nice to see the progress between 15 and 16. However, shows we still have to go a while.


question to all:

is there a verified, reproducible combination of resolution and refresh rate that results in 5W use, independent of monitor type?

Or is all this a stroke of luck that involves monitor brand, adapter type, zodiac and sock color of the day?

Thanks for any advice! Been trying to connect a full HD monitor to my mbp16 but am unable to get the promised land of 5W.

In my experience you can get 5w by meeting the following.

5w usage
- MBP screen closed, clamshell mode
- Single monitor
- Good quality USB-C to DisplayPort cable (that supports 4k@60hz in spec)
- at 4k 60hz or 30hz (any scaled res)

7-9w usage
- MBP screen closed, clamshell mode
- 2x 4k monitor
- Both with good quality USB-C to DisplayPort cable (that supports 4k@60hz in spec)
- Both at 4k 60hz or 30hz (any scaled res)

18w
- Exact scenario as above, with MBP screen open will result in 18w

18w
- The same display with USB-C to HDMI will result in 18w