Yes, your MacBook Pro is running too hot, and yes, it is Apple's fault.

[dblissmn]

Yes with SB2011, not with rMBP late 2013

My headline pretty much says it all; even in a warm, muggy office, running a secondary monitor, I don't push my rMBP "late 2013" 15-inch past 50 degrees Celsius unless I'm pushing it with stats and graphics. That's a 15 to 20 degree improvement on the Sandy Bridge 2011 15 inch.

Something big has happened here, and I don't think it's just the better venting on the Retina models. But two questions. 1) is this just Intel's better silicon, or also down to Apple improving the QC on the cooling system; and 2) is this just the 15 with discrete graphics, or are other Haswell machines including the 13 showing similar improvements even with external displays on Iris integrated graphics?

 

[Daoud45]

is there anything I can do?

... The main issue is NOT the compound used, nor the quantity. The main issue is the substandard machining of the heatsink's copper base. The over zealous use of thermal compound exacerbates the issue, but does not create it....

Doward – Is there any way to fix the heatsink copper base issue? If so, would you recommend fixing it or use existing measures such as fan control, cooling pads, etc. And finally, if you did recommend fixing it, who can do it and how much (approximately) would it cost?

And thanks for your detailed description of the problem.

 

[SWAN808]

interesting I might check that video...

 

[Doward]

Doward – Is there any way to fix the heatsink copper base issue? If so, would you recommend fixing it or use existing measures such as fan control, cooling pads, etc. And finally, if you did recommend fixing it, who can do it and how much (approximately) would it cost?

And thanks for your detailed description of the problem.

Recommendation would be to micropolish the heatsink - if you've got the tooling for it. Otherwise, a lapping and heatsink re-paste will go a very long way!

For what it's worth, just did a 15" Arrendale setup - now that's a HOT running CPU!

 

[barnettgs]

...to micropolish the heatsink - if you've got the tooling for it.

Just curious of what tool to use, I have a dremel tool with the full kit. I am sure there is a polisher bit there somewhere, would that do?

 

[Doward]

Just curious of what tool to use, I have a dremel tool with the full kit. I am sure there is a polisher bit there somewhere, would that do?

Oh God, no!! You will not have a flat surface with a dremel, and will likely make things worse!

You can use a lapping stone, something like marble or granite that is perfectly flat, or a machine straight edge to lap the heatsink.

Electropolishing requires special tooling, and would be outside most people's capabilities.

Lapping, however, is relatively easy.

 

[barnettgs]

Oh God, no!! You will not have a flat surface with a dremel, and will likely make things worse!

You can use a lapping stone, something like marble or granite that is perfectly flat, or a machine straight edge to lap the heatsink.

Electropolishing requires special tooling, and would be outside most people's capabilities.

Lapping, however, is relatively easy.

Just asking

Fair enough but my heatsink isn't exactly flat anyhow. Could you show me if you have already posted of how you have lapped the heatsink?

Thanks

 

[Doward]

I don't think I've done a video on lapping the heatsink, but there are many on YouTube

 

[barnettgs]

I don't think I've done a video on lapping the heatsink, but there are many on YouTube

Got mixed results on youtube, some no-show, some lapping by hands and some not related to Macbook Pro so how am I supposed to know the good from the bad?

 

[ha1o2surfer]

Some people just do not seem to understand that an overheating Macbook actually hurts their performance that they payed top dollar for. This is very evident in benchmarks that are NOT using geekbench. LuxMark is a great way to test this. Your CPU should be reaching full turbo clocks 24/7 and not base speeds if your heatsink is working properly.

Non Turbo Base Clock Score:

WITH full turbo:

My 2012 Macbook Pro running this test with all OpenCL CPUs and GPUs participating in the benchmark gets a lower score than just my CPU alone because of all the thermal throttling. NO excuse, Apple.

 

[laurihoefs]

Some people just do not seem to understand that an overheating Macbook actually hurts their performance that they payed top dollar for. This is very evident in benchmarks that are NOT using geekbench. LuxMark is a great way to test this. Your CPU should be reaching full turbo clocks 24/7 and not base speeds if your heatsink is working properly.

Non Turbo Base Clock Score:
Image

WITH full turbo:
Image


My 2012 Macbook Pro running this test with all OpenCL CPUs and GPUs participating in the benchmark gets a lower score than just my CPU alone because of all the thermal throttling. NO excuse, Apple.

Intel® Turbo Boost Technology 2.0 Quick overview

"Intel® Turbo Boost Technology 2.01 automatically allows processor cores to run faster than the rated operating frequency if they’re operating below power, current, and temperature specification limits."

Size matters. You make some compromises for portability, and should not expect a laptop to be able to dissipate as much heat as a desktop system.

 

[ha1o2surfer]

Intel® Turbo Boost Technology 2.0 Quick overview

"Intel® Turbo Boost Technology 2.01 automatically allows processor cores to run faster than the rated operating frequency if they’re operating below power, current, and temperature specification limits."

Size matters. You make some compromises for portability, and should not expect a laptop to be able to dissipate as much heat as a desktop system.

Then why pay for a more powerful CPU when it's being so gimped by the cooling system?

I do NOT make compromises and I get full CPU and GPU performance out of my 14 inch gaming laptop that crushes every Macbook Pro to date. I own a Macbook, and it's collecting dust because of the compromises that Apple has made.

 

[Doward]

Got mixed results on youtube, some no-show, some lapping by hands and some not related to Macbook Pro so how am I supposed to know the good from the bad?

Perhaps I should do a video on lapping, then

 

[laurihoefs]

Then why pay for a more powerful CPU when it's being so gimped by the cooling system?

I do NOT make compromises and I get full CPU and GPU performance out of my 14 inch gaming laptop that crushes every Macbook Pro to date. I own a Macbook, and it's collecting dust because of the compromises that Apple has made.

Gaming laptops usually don't need to be slim or quiet, both which are expected of the rMBP.

You pay more to get more processing power. The higher base clock speed is maintained, despite of throttling. The turbo clock speeds are not meant to be maintained 24/7.

A very good read are the Intel application guides. They are documents meant for people who design the system around the processor, including the cooling system, and include interesting information, like expected realistic workloads, acceptable temperatures, etc.

Edit: here's the collection of documents for all Core CPUs: http://www.intel.eu/content/www/eu/en/processors/core/CoreTechnicalResources.html
And here's the document that specifies thermal management (section 5) of Haswell CPUs: http://www.intel.eu/content/dam/www/public/us/en/documents/datasheets/4th-gen-core-family-mobile-m-h-processor-lines-vol-1-datasheet.pdf

Edit 2: it's actually interesting how well figure 15 provided as an example on page 71 of the last linked document seems to correspond to the results in this thread: Automated tool to reveal throttling and overheating - GitHub (see results of higher end CPUs)

Perhaps I should do a video on lapping, then

This is actually a very good idea, especially since there is a lot of conflicting information about lapping and the results it yields floating around.

It would be nice to see some results from someone with a systematic approach and much experience with laptop cooling.

 

[ha1o2surfer]

Gaming laptops usually don't need to be slim or quiet, both which are expected of the rMBP.

You pay more to get more processing power. The higher base clock speed is maintained, despite of throttling. The turbo clock speeds are not meant to be maintained 24/7.

A very good read are the Intel application guides. They are documents meant for people who design the system around the processor, including the cooling system, and include interesting information, like expected realistic workloads, acceptable temperatures, etc.

Actually the laptop I own (14 in G46VW) is much quieter than my Macbook Pro with the fans full bore. I get full turbo all day long with temps never reading higher than 80c. I understand that the Macbook isn't made to run full bore for 24/7 but I thought it was a considered to be a mobile workstation?

I'm sorry for going off topic but I guess the Macbook just isn't made for anything intensive for long periods of time such as video editing or even light gaming.

Intel specifies that a heatsink should be able to dissipate heat at a CPU's rated TDP which is usually 35,45,47 watts in current macbooks.

"Intel defines TDP as follows: The upper point of the thermal profile consists of the Thermal Design
Power (TDP) and the associated Tcase value. Thermal Design Power (TDP) should be used for
processor thermal solution design targets"

"If you are a thermal engineer, the processor TDP specification is very important because your thermal
solution (fans, heat sink, etc.) must be able to be able to dissipate the rated TDP value. Intel and AMD
both agree on this point."

I mean, this sounds pretty self explanatory to me.

Link: http://www.intel.com/content/dam/doc/white-paper/resources-xeon-measuring-processor-power-paper.pdf (The document is states it's for server CPUs but the information is the same for all CPU's)

a CPU at it's base clock speed is not consuming it's TDP, it's actually 10-15 watts under (which makes sense to allow for turbo boost). This means if the heatsink is being saturated while drawing 36 watts at 2.9GHz and the CPU's base clock is 2.7ghz, then by definition it's throttling. This can be seen in Intel's Tuning utility under windows for more proof.

 

[laurihoefs]

Actually the laptop I own (14 in G46VW) is much quieter than my Macbook Pro with the fans full bore. I get full turbo all day long with temps never reading higher than 80c. I understand that the Macbook isn't made to run full bore for 24/7 but I thought it was a considered to be a mobile workstation?

I'm sorry for going off topic but I guess the Macbook just isn't made for anything intensive for long periods of time such as video editing or even light gaming.

Is it considered to be a mobile workstation? All the mobile workstations I know of (e.g. Lenovo W series, high-end EliteBooks, Dell Precision, Fujitsu Celsius, etc.) are much thicker and heavier than the rMBP. I did not expect to get the same thermal performance from a rMBP, than I got from a W5**. Nor would I expect it to perform better than an Asus gaming laptop.

Intel specifies that a heatsink should be able to dissipate heat at a CPU's rated TDP which is usually 35,45,47 watts in current macbooks.

"Intel defines TDP as follows: The upper point of the thermal profile consists of the Thermal Design
Power (TDP) and the associated Tcase value. Thermal Design Power (TDP) should be used for
processor thermal solution design targets"

Link: http://www.intel.com/content/dam/doc/white-paper/resources-xeon-measuring-processor-power-paper.pdf (The document is states it's for server CPUs but the information is the same for all CPU's)

a CPU at it's base clock speed is not consuming it's TDP, it's actually 10-15 watts under (which makes sense to allow for turbo boost). This means if the heatsink is being saturated while drawing 36 watts at 2.9GHz and the CPU's base clock is 2.7ghz, then by definition it's throttling. This can be seen in Intel's Tuning utility under windows for more proof.

Yes, it is throttling. Just as it should, per specification.

 

[ha1o2surfer]

Is it considered to be a mobile workstation? All the mobile workstations I know of (e.g. Lenovo W series, high-end EliteBooks, Dell Precision, Fujitsu Celsius, etc.) are much thicker and heavier than the rMBP. I did not expect to get the same thermal performance from a rMBP, than I got from a W5**. Nor would I expect it to perform better than an Asus gaming laptop.

Then why stick a quad core and a 650m in a macbook in which the performance can be half of what they would normally run at ( in my laptop and a mobile workstation)? I guess that is what I don't quite understand.

 

[laurihoefs]

Then why stick a quad core and a 650m in a macbook in which the performance can be half of what they would normally run at ( in my laptop and a mobile workstation)? I guess that is what I don't quite understand.

Because throttling only occurs under very limited circumstances. In most applications you get the full benefit of the four cores and turbo boost.

It would be nice, if the cooling performance would allow to maintain higher clock speeds, like it does on designs that exceed Intel's guidelines (like some workstation and gaming laptops)... But it's a limit brought by the design. You gain a very portable computer, but lose some performance in extreme cases.

 

[ha1o2surfer]

It does make sense I guess but I would consider editing/converting videos or playing Bioshock (from the Mac App Store) non rare cases..

Again I bring the attention back to this quite from the intel doc

"If you are a thermal engineer, the processor TDP specification is very important because your thermal
solution (fans, heat sink, etc.) must be able to be able to dissipate the rated TDP value. Intel and AMD
both agree on this point."

For some extra facts...

Non Turbo Base clock CPU power draw (estimated) 24 watts

Under Full Turbo: (70 watts)

 

[laurihoefs]

It does make sense I guess but I would consider editing/converting videos or playing Bioshock (from the Mac App Store) non rare cases..

Does editing/converting videos cause the CPU clock speeds to throttle below the base clock speed?

Again I bring the attention back to this quite from the intel doc

"If you are a thermal engineer, the processor TDP specification is very important because your thermal
solution (fans, heat sink, etc.) must be able to be able to dissipate the rated TDP value. Intel and AMD
both agree on this point."

For some extra facts...

Non Turbo Base clock CPU power draw (estimated) 24 watts

Image

Under Full Turbo: (70 watts)

Image

So Turbo Boost power exceeds the TDP, if the clock speed is maintained for a long time? This is to be expected, isn't it?

You should also take in account the discrete GPU. What happens, if a cooling system designed to dissipate the TDP of the CPU and GPU, is faced with a thermal load of both of them exceeding their TDP?

 

[ha1o2surfer]

Does editing/converting videos cause the CPU clock speeds to throttle below the base clock speed?



So Turbo Boost power exceeds the TDP, if the clock speed is maintained for a long time? This is to be expected, isn't it?

You should also take in account the discrete GPU. What happens, if a cooling system designed to dissipate the TDP of the CPU and GPU, is faced with a thermal load of both of them exceeding their TDP?

I just did something stupid and let it turbo to 4ghz when I should have let it turbo to 3.4ghz which is 45 watts. lol

It will if the heatsink can't keep up but I would hope the heatsink can keep the temp under control so the CPU can stay above it's base clocks..

Turbo boost MAY run above a CPU's rated turbo but once it drops back down to 45 watts (or it's rated TDP) The temps should be way under throttling. not flirting at 100c...

A heatsink should be deisigned properly to cool both. As explained in my post here, Apple uses a heatsink in a way it was not designed.

Link: https://forums.macrumors.com/posts/18942535/

 

[simonsi]

Any manufacturing process will have a specification range of output tolerance, without understanding what that range actually is you can't sensibly draw any conclusions from any individual machine.

All Ford cars would run better if all their output came out with blueprinted engines (i.e. all engine components selected or machined to VERY close tolerances as with a hand-built race engine), however moving to that is not cost-free for Ford, nor would tightening up any manufacturing process tolerance be cost-free for its end customers.

Apple, their actuaries and user base, accept the compromise of current cost vs usage vs warranty support.

No surprises here and this is typical if you try and extrapolate from a non-random sample.

 

[barnettgs]

Perhaps I should do a video on lapping, then

That would be great and very helpful.

 

[Doward]

That would be great and very helpful.

I'll work on it, then

 

[The-Reporter]

Fan speed

Hi,

From my rMBP 2014 (15", Core i7 2.5GHz, GT750m) : with two process at 100% (yes > /dev/null), the right fan goes up to 5700rpm, and the left fan goes up to 6200rpm (full speed), after about only 1 minute.

That seems to high. My old MBP 2012 (15" Anti-Glare, Core i7 2.6GHz, GT650m) maintains the two fans at idle speed (2000rpm) (during more than 5 minutes) in the same conditions (two process at 100%, same ambient temperature, same surface below MBP).

Can you confirm that's normal or is there a problem with the rMBP ? Could you test it with your rMBP ?

Thanks.