rMBP 15: Worth to choose Low-End CPU to minimize heat?

[GoodOne]

Hey all,

I'm currently working with a rMBP 13 2013 and I'm very satisfied with the heat and noise Levels.

As I'm looking Forward to upgrade to a rMBP 15 2016 once they are released, I'm wondering if it's worth to pick the "slowest" CPU available as well as the iGPU in order to minimize heat and fan noise?

Judging from the current 15 models, the low end 2.2 and 2.5 GHZ CPUs have the same TDP but my guess is that the 2.2 GHZ model should run a little cooler potentially?

Or am I wrong?

I don't really Need much CPU power, I'm just looking to get a bigger screen and cool temperatures are important to me. And coming from a 2012 CPU I guess it's much faster anyway so I don't really care about the high end CPUs although I'm looking Forward to the massive increase in SSD Speed the newer models offer.

So, should I get the low-end CPU with integrated iGPU in my case? Or is there no difference between the low-end the and the mid-level CPU in Terms of heat and noise?

 

[leman]

Its really the matter of chance. Because of random factors in the tech process, supposedly identical chips are not made equal. You can have a higher-clocked CPU that actually uses less power than a owed-clocked CPU of the very same design. In fact, the CPUs/GPUs used for mobile parts are usually the same chips as the desktop parts, but they are chosen specifically because they happen to be more efficient. Bottomline: what you say sounds logical, but it really doesn't have to be this way. It might, but not necessarily.

However, a model without the dGPU will probably be cooler — not because of the dGPU vs iGPU per se, but because how the driver is operating. Usually, when the dGPU kicks in, the system tries to optimise for performance. With the iGPU, the system usually tries to optimise for power use. On its own, iGPU is probably hotter, because it has negative impact on the CPU performance in scenarios that require high GPU utilisation. Because dGPU is a separate component, it can be cooled separately, increasing the thermal window.

 

[OS X Dude]

The overall thinness and lack of adequate cooling means they'll get toasty after prolonged or heavy use regardless. If you're really concerned about thermal-throttling, the iMac processors offer higher TDP as they're desktop-class rather than the notebook-class ones used in the MacBook Pro. Granted, that may not be an option if you're dead-set on a desktop, but it is just physics. Plus, the post-2012 21.5" iMacs are actually pretty portable and you can even get transport bags for them.

In more direct answer to your question, a lower-specced system ought to run cooler so long as it's being used for less power-hungry tasks. Depends on your usage.

 

[zhenya]

I think you are likely to see the opposite effect in practice. The faster processors are generally sorted as the best quality from the same batches, and so are a bit more energy efficient by nature, plus the fact that a faster processor can do the same amount of work in less time, allowing it to return to its lowest power state more quickly.

 

[robvas]

As I'm looking Forward to upgrade to a rMBP 15 2016 once they are released, I'm wondering if it's worth to pick the "slowest" CPU available as well as the iGPU in order to minimize heat and fan noise?

Yes. I can tell when the discrete graphics in mine turns on - it heats up way, way more.

 

[GoodOne]

I think you are likely to see the opposite effect in practice. The faster processors are generally sorted as the best quality from the same batches, and so are a bit more energy efficient by nature, plus the fact that a faster processor can do the same amount of work in less time, allowing it to return to its lowest power state more quickly.

Sounds logic, too. From that perspective I should get the high-end CPU? I always that the high end CPUs would run hotter?

It seems Logical to get the iGPU but on the CPU I'm still on the fence.

 

[maratus]

It ultimately depends on Intel, but given the situation with the 2012/2013 rMBPs, it's still likely that the future Intel CPUs will show the same behaviour.

As you may know, 2.3/2.4GHz versions ran significantly cooler (up to 10-12C less under 100% load) than 2.6/2.7GHz due to lower voltage boost. 2.3/2.4GHz processor were the only ones that didn't throttle under continuous 100% load with 20-25C ambient. We don't know how exactly Intel is testing the chips and whether it'd better to have a higher clocked SKU downvolted vs. stock lower end SKU though.

The best thing you can do once the models are out is to find a good store which will allow you to install Intel Power Gadget for monitoring and do a stress test by running eight yes> /dev/null commands in terminal.

As far as iGPU goes, it definitely helps to reduce heat when you have external screens connected because dGPU will be forced on dGPU equipped models. At least that's how it is now. Oh, I'd strongly recommend to use GfxCardStatus, because only you should decide if you want iGPU or dGPU for any particular task. That little application makes a big difference in user experience.

I think you are likely to see the opposite effect in practice. The faster processors are generally sorted as the best quality from the same batches, and so are a bit more energy efficient by nature, plus the fact that a faster processor can do the same amount of work in less time, allowing it to return to its lowest power state more quickly.

The will be no opposite effect due to speed in most scenarios. Heat is proportional to power, not total energy. If a slower processor takes longer to complete the task, so what? It'll just runs longer under load, but it'll be cooler while doing so. I dare you to find any reasonable use case that'll translate "same amount of work in less time" into lower power density.

And your cherry-pick argument is also invalid, because it's only useful when you are able to fine-tune your voltage boost right next to its stability threshold (i.e. when overclocking or downclocking/downvolting your processor). In real life default values for each CPU will be closer to average and won't take advantage of better cherry-picked crystals as much as we want. On the other hand, you'll still have higher voltage for higher frequency, which means higher end CPUs with higher base and TB frequency will run hotter.

Because dGPU is a separate component, it can be cooled separately, increasing the thermal window.

No, even though it's the same component, it's cooled by the same heatsink and the bottleneck isn't the overall interface area between heatsink / silicon but rather the heatsink to air + fan part. rMBP heatsink is too thin and fans are too weak, that's why your main concern here is total power. That's also only true due to iGPU being much slower and consuming much less power, of course.

The overall thinness and lack of adequate cooling means they'll get toasty after prolonged or heavy use regardless.

Toasty isn't a scientific term. If one model goes to 95C under prolonged heavy load and another one goes to 103C and starts throttling under the same load, it's far from being the same. Yes, both rMPBs will be "hot" in that scenario, which is kinda obvious, but that's it.

 

[Shazamadoo]

The will be no opposite effect due to speed in most scenarios. Heat is proportional to power, not total energy. If a slower processor takes longer to complete the task, so what? It'll just runs longer under load, but it'll be cooler while doing so. I dare you to find any reasonable use case that'll translate "same amount of work in less time" into lower power density.

And your cherry-pick argument is also invalid, because it's only useful when you are able to fine-tune your voltage boost right next to its stability threshold (i.e. when overclocking or downclocking/downvolting your processor). In real life default values for each CPU will be closer to average and won't take advantage of better cherry-picked crystals as much as we want. On the other hand, you'll still have higher voltage for higher frequency, which means higher end CPUs with higher base and TB frequency will run hotter.

most confusing paragraph that honestly doesn't explain anything. "Heat is proportional to power, not total energy".. um what does that mean?

a CPU that is equal in all aspects, except the one that has a higher clock speed, will use more energy thus producing more waste (heat)..

Now to the op, if you don't do any CPU intensive tasks on ur computer, having the top of the line CPU vs the entry level option, your not going to notice any heat issues.. Now if your doing heavy duty renderings, the higher clocked CPU will most likely get hotter, but it will also be more effective for the task at hand.. so you decide what you need.

 

[maratus]

most confusing paragraph that honestly doesn't explain anything. "Heat is proportional to power, not total energy".. um what does that mean?

Energy = Power * Time

but it will also be more effective for the task at hand.. so you decide what you need.

It won't if thermal throttling occurs

 

[leman]

a CPU that is equal in all aspects

The problem is that CPUs are not equal. Its literally a lottery. You can have a higher-clocked chip that runs cooler than a lower-clocked one.

 

[Shazamadoo]

Energy = Power * Time


It won't if thermal throttling occurs

not asking for the most basic of equations for "Power" you learn in elementary school.. I meant your statement "Heat is proportional to power, not total energy".. which makes no conventional/unconventional sense at all..

The problem is that CPUs are not equal. Its literally a lottery. You can have a higher-clocked chip that runs cooler than a lower-clocked one.

I believe you read the first 8 words in the sentence I posted and jumped to reply or something? I'm not getting the point of your post..

 

[maflynn]

So, should I get the low-end CPU with integrated iGPU in my case? Or is there no difference between the low-end the and the mid-level CPU in Terms of heat and noise?

Yep, why spend money when you posted that you don't need the power. Why bother buying components that you'll really not take advantage.

Save your money and get the base version, I did the same thing with a SurfaceBook. Its a secondary device, so I opted for the base model and saved myself nearly a grand between the base model and the dGPU equipped version

 

[leman]

I believe you read the first 8 words in the sentence I posted and jumped to reply or something? I'm not getting the point of your post..

Fair enough I might have misunderstood your message.

 

[Shazamadoo]

Also regarding the thermal throttling... I am guessing you are implying about the Turbo Boost speeds?

The #1 misinterpretation people have about Turbo Boost is that it is a powerhouse "overclocking" function, when it is actually nothing more than an efficiency feature..

Turbo boost will kick in when certain criteria's are met, IE: power/temp/OS demand/etc. Essentially Turbo Boost is opportunistic, and it does not mean that when it kicks in your at the max rated Turbo Boost speed.. Turbo Boost scales up to speed in increments and can kick in when your using 1 core or all. Meaning you don't have to be maxing out your CPU for Turbo Boost to kick in..

What you want to look at is the base frequency of the cpu, and yes the maxed out CPU for the rMBP will outperform the entry level CPU when compressing/3D rendering/exporting and such.

 

[Fishrrman]

I've only owned one MacBook (2010 13" MacBook Pro, which is still doing just fine), but from my observations of the Mac boards, it seems like the MacBooks -MOST LIKELY- to have problems over their service lives are the ones with the higher-end dedicated GPUs.

Could be a run of bad luck on Apple's part, or could be that high-end GPUs are just less reliable in general. Can't say...

 

[maratus]

Also regarding the thermal throttling... I am guessing you are implying about the Turbo Boost speeds?

The #1 misinterpretation people have about Turbo Boost is that it is a powerhouse "overclocking" function, when it is actually nothing more than an efficiency feature..

Turbo boost will kick in when certain criteria's are met, IE: power/temp/OS demand/etc. Essentially Turbo Boost is opportunistic, and it does not mean that when it kicks in your at the max rated Turbo Boost speed.. Turbo Boost scales up to speed in increments and can kick in when your using 1 core or all. Meaning you don't have to be maxing out your CPU for Turbo Boost to kick in..

What you want to look at is the base frequency of the cpu, and yes the maxed out CPU for the rMBP will outperform the entry level CPU when compressing/3D rendering/exporting and such.

I'd suggest just quit speculating and go check the TB multipliers for all possible loads (1/2/4/8) on an actual machine. It's as easy as that. You can call it "overclocking" or "efficiency" feature, the name is completely irrelevant here.

And what if I told you that a "maxed out CPU" on a 2013 rMBP (2.7Ghz model) often stayed below 3.2GHz with 8 threads load due to overheating while the 2.4 model managed to max @ 3.2Ghz all the time with a little bit headroom left? Guess which one was faster?

As far as my previous post goes, feel free to re-read it if you didn't get it.

 

[Samuelsan2001]

Hey all,

I'm currently working with a rMBP 13 2013 and I'm very satisfied with the heat and noise Levels.

As I'm looking Forward to upgrade to a rMBP 15 2016 once they are released, I'm wondering if it's worth to pick the "slowest" CPU available as well as the iGPU in order to minimize heat and fan noise?

Judging from the current 15 models, the low end 2.2 and 2.5 GHZ CPUs have the same TDP but my guess is that the 2.2 GHZ model should run a little cooler potentially?

Or am I wrong?

I don't really Need much CPU power, I'm just looking to get a bigger screen and cool temperatures are important to me. And coming from a 2012 CPU I guess it's much faster anyway so I don't really care about the high end CPUs although I'm looking Forward to the massive increase in SSD Speed the newer models offer.

So, should I get the low-end CPU with integrated iGPU in my case? Or is there no difference between the low-end the and the mid-level CPU in Terms of heat and noise?

If you don't need the performance then you won't be causing much heat anyway buy whatever you like.

 

[snaky69]

Hey all,

I'm currently working with a rMBP 13 2013 and I'm very satisfied with the heat and noise Levels.

As I'm looking Forward to upgrade to a rMBP 15 2016 once they are released, I'm wondering if it's worth to pick the "slowest" CPU available as well as the iGPU in order to minimize heat and fan noise?

Judging from the current 15 models, the low end 2.2 and 2.5 GHZ CPUs have the same TDP but my guess is that the 2.2 GHZ model should run a little cooler potentially?

Or am I wrong?

I don't really Need much CPU power, I'm just looking to get a bigger screen and cool temperatures are important to me. And coming from a 2012 CPU I guess it's much faster anyway so I don't really care about the high end CPUs although I'm looking Forward to the massive increase in SSD Speed the newer models offer.

So, should I get the low-end CPU with integrated iGPU in my case? Or is there no difference between the low-end the and the mid-level CPU in Terms of heat and noise?

If you're not pushing it (and it sounds like you don't even push your 13"), you won't cause any more heat than you currently do. Chances are the 15" will run a bit cooler because it has 2 fans, not 1.

 

[Shazamadoo]

I'd suggest just quit speculating and go check the TB multipliers for all possible loads (1/2/4/8) on an actual machine. It's as easy as that. You can call it "overclocking" or "efficiency" feature, the name is completely irrelevant here.

And what if I told you that a "maxed out CPU" on a 2013 rMBP (2.7Ghz model) often stayed below 3.2GHz with 8 threads load due to overheating while the 2.4 model managed to max @ 3.2Ghz all the time with a little bit headroom left? Guess which one was faster?

As far as my previous post goes, feel free to re-read it if you didn't get it.

it's quite obvious you are grasping at straws trying to have your statements be correct and relevant..

Look at all the benchmarks for Final Cut Pro export times/Photoshop renderings on the various MBP configs.. wait why am I even arguing basic facts... I don't even know why I am even replying to your uneducated/falsified/made up statements..("Heat is proportional to power, not total energy") are you trying to flunk a graduate dissertation with that statement?

You are obviously ill-informed..

 

[maratus]

it's quite obvious you are grasping at straws trying to have your statements be correct and relevant..

Look at all the benchmarks for Final Cut Pro export times/Photoshop renderings on the various MBP configs.. wait why am I even arguing basic facts... I don't even know why I am even replying to your uneducated/falsified/made up statements..("Heat is proportional to power, not total energy") are you trying to flunk a graduate dissertation with that statement?

You are obviously ill-informed..

if you don't get it, it doesn't mean that others are wrong.

And since you're struggling, I'll elaborate. If processor A has higher speed and power consumption than processor B, being able to rush and finish the task earlier will almost never result in lower temperatures (and if you actually read the thread, that exactly was posted as an argument). Processor B will take longer time Tb to complete the same task and it'll use Pb average power while doing so (total energy Eb = Tb*Pb). Processor A may even be more energy efficient (Ea < Eb), but as long as it doesn't stop computation periodically during the task, Ta will be always greater than Tb. And A will run hotter and its temperature will be higher during the continuous computation, even though practically it won't be for as long as for B. Since for us the last statement about time is irrelevant (i. e. being @ 105C for 18 min vs. 95C for 20 min is actually worse in terms of heat perception, fan noise etc.) you can say that in practice we only care about P but not E.

And one counter example would be a task which is equalised in computation time for A and B by external constraints like data availability. Only in such a case both Pa and Ea will be lower (provided that A is more energy efficient) resulting in processor A running cooler than its slower competitor.

 

[Shazamadoo]

if you don't get it, it doesn't mean that others are wrong.

And since you're struggling, I'll elaborate. If processor A has higher speed and power consumption than processor B, being able to rush and finish the task earlier will almost never result in lower temperatures (and if you actually read the thread, that exactly was posted as an argument). Processor B will take longer time Tb to complete the same task and it'll use Pb average power while doing so (total energy Eb = Tb*Pb). Processor A may even be more energy efficient (Ea < Eb), but as long as it doesn't stop computation periodically during the task, Ta will be always greater than Tb. And A will run hotter and its temperature will be higher during the continuous computation, even though practically it won't be for as long as for B. Since for us the last statement about time is irrelevant (i. e. being @ 105C for 18 min vs. 95C for 20 min is actually worse in terms of heat perception, fan noise etc.) you can say that in practice we only care about P but not E.

And one counter example would be a task which is equalised in computation time for A and B by external constraints like data availability. Only in such a case both Pa and Ea will be lower (provided that A is more energy efficient) resulting in processor A running cooler than its slower competitor.

Wait.. I understand.. So we only care about P but not E right? So WTF + LMAO are = to U...

You must of learned these.. what is it? Some kind of Physics+Rocket Surgery algorithm from some "genius" i presume..