Why do people keep talking about "Thermal Throttling?"

[macdudesir]

Okay, hear me out. I know the MBPs (especially nTB ones) get hot at full load...that is obvious and something you can't escape. But from everything I've seen, the CPU's hold their own pretty well even running at full tilt with the fan(s) at full speed. In every video or review where they talk about thermal throttling, they say how after 20 minutes or whatever at full load, the CPU "thermal throttles" down from whatever crazy turbo boost it has (say, 4.5GHz) to a slightly slower speed, like 3.9GHz for example, and then proceed to talk about how Apple skimped on cooling in the machine. Now, I'm using my processor for reference...the base clock on this processor is 3.1GHz...So, it might be a technicality of the word, but I don't understand how someone can say that the CPU is thermal throttling when it is still running well above its base clock speed.

If you ask me, I would consider thermal throttling to be any time the CPU has to lower its clock speed below its base clock speed in order to maintain thermal operating specifications.

Point is, turbo boost is just that, it is a turbo system designed to increase the processing performance of your CPU for short term intensive tasks, in order to make the system run a little snappier. It was never designed to run at full turbo boost for hours on end. If it were, there would be no point in turbo boost, and its highest turbo speed would be considered to be its base clock. Apple designs the thermal system to be able to keep the CPU sufficiently cool running at full load at its base clock...not at full turbo boost.

So, if you ask me, Apple has actually done well considering that almost all reviews and videos I have seen stress testing MBPs and other Apple computers show that the CPU stays at least a few hundred MHz above its base clock during hours of testing.

Further, don't just take it from me, but if you look at how Intel defines turbo boost, they say "Maximum turbo frequency indicates the highest possible frequency achievable when conditions allow the processor to enter turbo mode...Availability and frequency upside of Intel® Turbo Boost Technology 2.0 state depends upon a number of factors including, but not limited to, the following:

• Type of workload
• Number of active cores
• Estimated current consumption
• Estimated power consumption
• Processor temperature

Note: Intel® Turbo Boost Technology 2.0 allows the processor to operate at a power level that is higher than its TDP configuration and data sheet specified power for short durations to maximize performance."

With that said, I will end my rant. I'm sorry, that is just something that has been grinding my gears in a large number of reviews and videos I am watching about the new MBP's. I think people have unrealistic expectations. If turbo boost stuck at full speed all the time, what point is there in getting a 3.1ghz i5 vs a 1.2ghz i5 if they both boost to 4.whatever...

Interested to hear you all's thoughts...if some models are in fact thermal throttling below base clock, id be interested to know.

 

[Shamgar]

A different laptop with the same CPU but sufficient cooling to maintain the boost clocks would outperform the MBP. Apple doesn't promise more than the base clock, so they're not being dishonest, but it is still relevant for a reviewer or commenter to point out what performance is being left on the table because of cooling choices.

 

[jerryk]

Other systems do just fine without throttling and running for hours and hours. Especially any workstation level systems or gaming systems.

 

[Queen6]

My own primary notebook is capable of running at full boost indefinitely as do many others. Not being capable of this is simply an indication of an inadequate cooling system.

Over powered silicon in a overly thin chassis is little more than a sales gimmick...

Q-6

 

[Maxx Power]

The way Intel specifies and have always specified their processors is such that the top performance level is "p0" (as per Intel). "p0" is historically the top speed for Intel processors long before Turbo Boost. When older Intel processors used the "Speed Step" technology, it would change from "p0" level to "p1", "p2", etc. The predecessor of Turbo Boost was Intel Dynamic Acceleration (IDA) on Core 2 processors, which also jumped between performance levels in response to demand and thermals. In modern, complex CPUs, the nature of power consumption has become very bursty under load, such that its instantaneous power consumption under load greatly exceeds the rated TDP. Because Intel specifies the TDP of the processor, any OEM would have to design their power delivery and cooling circuitry to those numbers. Previously, if an OEM designed their systems for the rated TDP, they can be sure that "p0" level can be sustained. From Nehalem onward, this is no longer the case.

What technologically changed with the introduction of the Turbo Boost is not transparent to users. With Turbo Boost, the processor can now be programmed to stay in each performance level on a time and power basis. This fine tuning was not available with earlier IDA and Speed Step technologies. However, Intel leaves the actual implementation of Turbo Boost entirely up to each OEM. In the firmware (EFI, UEFI, BIOS), an OEM can specify the amount of time allowed under a given TDP (you can read this information with an Intel utility). On most desktop PCs, you can even set your own in the BIOS/UEFI, as per your cooling setup. On laptops, the implementation of Turbo Boost behavior varies greatly from laptop to laptop, even under the same manufacturer and sometimes on the same model with different firmware revisions (a common subject of analysis for the review of laptops).

Now, given the information above, the main reasons why I surmise that people are often unhappy with Turbo Boost are:

(1) What Intel has really done historically was to guarantee "p0" performance under specified cooling and power conditions. With Turbo Boost, Intel has make "p0" a CONDITIONAL time- and temperature-based state. This changeover was sudden but the underlying need only gradually evolved, making the way Turbo Boost operate a total shocker to many people who are used to purchasing specifications.

(2) The fact that many computers (laptops predominantly) are programmed by the OEM to sustain "p0" performance levels and both the power delivery and cooling systems are designed from the ground-up to sustain "p0" states indefinitely makes for interesting and sometimes unfair comparisons between computers.

(3) The more cooling-deficient computers always leave "p0" performance states sooner. It doesn't take a genius to read a few review at Notebookcheck and find out that those laptops with lesser cooling systems are typically also programmed by their OEMs to only make use of the "p0" states infrequently.

So in my opinion then, (1) leads people to think that you pay for the performance, which is typically and historically advertised as "p0" (and many times benchmarked under unconstrained power and thermal conditions, hence "p0" also) and (2) and (3) suggest that sustained "p0" states are entirely feasible, but up to the OEM and their designs to maximize "p0" time.

Hence I think that most people who complain about Turbo Boost in Macbook Pros draw the conclusion that their laptops are unable to sustain "p0" states because Apple programmed the CPUs to back down from "p0" when too hot, or exceeded the long-term power limit. The reality is that the definition of performance (or the "p0" state) changed and this was not well-communicated to the end users. Additionally, because of inconsistencies in the design and firmware programming of different laptops, some can sustain "p0" and others can not. This latter point leads to confusion and competitive behaviors between some OEMs.

My personal thoughts on this matter for Macbooks and Macbook Pros is that - these are form factor-limited machines. All design considerations were centered around portability. Given that goal, I think Apple has done a good job traditionally of balancing raw performance and usability. If one reads a few review on Notebookcheck, one is bound to quickly find out that Apple is commonly the most generous in their long-term TDP tunings of their computers, frequently allowing the maximum TDP exceeding the nominal TDP values (for example, some 15W chips are allowed 28W long-term sustained TDP). A lot of the other laptop manufacturers especially from their previous generation stuff, were not nearly as generous. The most common constraining factor is then not the TDP programming (or power allowance) but thermal state.

Having said that, I would gladly welcome Apple to introduce something a bit thicker and can sustain "p0" more often or indefinitely. Upgrade-ability is obviously a beneficial addition in this case.

 

[leman]

Because people don’t have a clue There are barely any notebooks on the market that offer better sustained CPU performance than the current MBP. And it’s certainly the thinnest machine that can run for hours on full CPU without any throttling whatsoever.
[doublepost=1517996507][/doublepost]

Having said that, I would gladly welcome Apple to introduce something a bit thicker and can sustain "p0" more often or indefinitely. Upgrade-ability is obviously a beneficial addition in this case.

Both burst and sustained performance of 2016/2017 MacBook Pro'ss in on par with much heavier and thicker gaming laptops and doesn't show any dips over long periods of use even with upgraded CPUs (unlike some competitors like the Dell XPS 15). I am not really sure how Apple can do any better here if large dedicated gaming laptops can't do any better.
[doublepost=1517996548][/doublepost]

My own primary notebook is capable of running at full boost indefinitely as do many others.

I thought you have a Surface Book? It dips hard after few seconds of sustained load...

 

[Queen6]

Because people don’t have a clue There are barely any notebooks on the market that offer better sustained CPU performance than the current MBP. And it’s certainly the thinnest machine that can run for hours on full CPU without any throttling whatsoever.

Might want to check your facts, there are plenty of notebook's that can run at fullbore including significantly more powerful dGPU's. Without doubt the MBP is the thinnest.

Q-6

 

[leman]

Might want to check your facts, there are plenty of notebook's that can run at fullbore including significantly more powerful dGPU's. Without doubt the MBP is the thinnest.

Please read my post again. I said: "There are barely any notebooks on the market that offer BETTER sustained CPU performance than the MBP", not "the MBP outperforms all other notebooks". Notebookcheck for instance offers good estimates of long-term sustained performance by running Cinebench 60 times in a row. I haven't seen any laptop that would deliver more systematic scores than the MBP. This is also consistent with my own testing — you can push the machienfor hours without any dips in performance.

 

[Queen6]

I thought you have a Surface Book? It dips hard after few seconds of sustained load...

Well things change, and I do still have a 1st Gen Surface Book nor does it's CPU throttle in "seconds" I am unsure of running the system at 100% for hour upon hour, equally stating a few seconds is simply inaccurate at best. My own i7 remains capable of holding full boost for periods exceeding 30-40 minutes, beyond that I don't know as I've generally better things to be getting on with it. As for Surface Book 2 personally I have no experience.

This notebook I'm responding on is significantly more powerful than the MBP, nor does it throttle, as it's designed for heavy sustained use. Hence why I purchased and tested this one, as it will be required to run heavy load for sustained periods. The notebook it replaced was the same, however being a thinner chassis proved to generate more noise than I cared for. This one I'm really impressed with the performance and scalability I may opt for the version with OC CPU and GTX 1080, although the 7700HQ, GTX 1070, 32Gb is currently no slouch.

Apple's not the only company to figure out thermal's, Apple is just master at getting away with it in a thinner chassis, at times to the cost of it's customer's. Prior to 2016/2017 the 15" MBP with dGPU made for a pretty good toaster.

Q-6

 

[leman]

This notebook I'm responding on is significantly more powerful than the MBP, nor does it throttle, as it's designed for heavy sustained use.

What brand/model is that?

P.S. And yes, all Surface Book CPUs throttle in seconds simply because thats what the 15W CPUs are designed to do. I do not mean to display these CPUs as inferior. Its simply the tradeoff price one has to pay — you get short-term performance of its "bigger" brothers but the performance will fall off when you push it for longer. All 15W parts "throttle", by design.

P.P.S. I have an issue with the liberal use of term "throttle" to be honest. It should be reserved to situations where the cooling system is not adequate in dissipating the TDP. Instead, people use it for basically everything. Looking at the 8-gen 15W for example — these chips are designed to surpass their TDP for brief periods of time (thats where they get their excellent burst performance from). So even if you have adequately designed (15W) cooling system, it simply can't support the CPU running on its max power. Is it even fair to call this throttling? Now, if the cooling system could only handle 10W, yes, we could talk about throttling. In the end, the issue is that people fundamentally misunderstand how stuff works and then don't even bother to check the information they are mindlessly repeating.

 

[Maxx Power]

Both burst and sustained performance of 2016/2017 MacBook Pro'ss in on par with much heavier and thicker gaming laptops and doesn't show any dips over long periods of use even with upgraded CPUs (unlike some competitors like the Dell XPS 15). I am not really sure how Apple can do any better here if large dedicated gaming laptops can't do any better.

Yeah, I agree that the "p0"-sustained performance is generally good on moderm Macbook Pros with 2 fans compared to all competitions of the same form factor. Lenovo, for example has been quite bad at sustaining "p0" states.

Anyway, I was referring to sustained "p0" states under both CPU and GPU loading. I think really only some gaming laptops and workstations can sustain higher clocks, but they are categorically larger than the Macbook Pros. The smaller non-TB 13 inch Macbook Pro also drops out of "p0" under sustained CPU load.

As for "larger gaming laptops that can't do any better" - their engineering is not anywhere near Apple's. I have torn down a lot of those laptops and I can say without a doubt that their hardware design is utter crap. Some (many MSIs for example) puts battery packs that are meant for external batteries on the inside, right next to the heat pipes. Some use very questionable quality components (like capacitors). A lot of them use 1x1 wifi. Power management is universally poor on these laptops. Many of these laptops feature very high TDP GPUs that when gaming, the GPU eats up most of the cooling budget. So I think the reason that they can't do any better has nothing to do with the size, but has everything to do with poor design and engineering. Which means that Apple can do better with more room to work with. That more room could mean, although not necessarily - more sustained performance, higher burst performance, more environmental tolerances, more thermal margins, etc.

 

[PBG4 Dude]

I run my 2017 15” MBP under full load for hours 5-7 at a time rendering. It is not slowing down as time progresses which would infer CPU throttling. The last frames are taking roughly the same amount of time to render as the first frames, and I have the metadata to prove it.

 

[macdudesir]

Because people don’t have a clue There are barely any notebooks on the market that offer better sustained CPU performance than the current MBP. And it’s certainly the thinnest machine that can run for hours on full CPU without any throttling whatsoever.
[doublepost=1517996507][/doublepost]

Both burst and sustained performance of 2016/2017 MacBook Pro'ss in on par with much heavier and thicker gaming laptops and doesn't show any dips over long periods of use even with upgraded CPUs (unlike some competitors like the Dell XPS 15). I am not really sure how Apple can do any better here if large dedicated gaming laptops can't do any better.
[doublepost=1517996548][/doublepost]

I thought you have a Surface Book? It dips hard after few seconds of sustained load...

I am glad you brought up XPS 15 and Surface Book, as IMHO those are the biggest competitors to the MBP in the MBP's category (which is not big, bulky, heavy gaming computers). My roommate has a gen 1 surface book, and it does in fact drop back to its base clock very quickly. I have also seen my previous Surface Pro 4 throttle below its base clock speed. Intel power gadget at the time said "power limit throttling: yes...thermal throttling: yes."

As for XPS 13 and 15...I have read tons of articles about them not maintaining any turbo under full load. They will throttle up for a few minutes, then quickly retreat to their base clock with normal tasks, not even benchmarking. Further, they are also power limit throttled back down to their TDP quickly...whereas apple has allowed its intel CPU's to maintain a power level above TDP for hours.

Please read my post again. I said: "There are barely any notebooks on the market that offer BETTER sustained CPU performance than the MBP", not "the MBP outperforms all other notebooks". Notebookcheck for instance offers good estimates of long-term sustained performance by running Cinebench 60 times in a row. I haven't seen any laptop that would deliver more systematic scores than the MBP. This is also consistent with my own testing — you can push the machienfor hours without any dips in performance.

Just to agree with you and add a bit...Yes, this is also what I have noticed. As I said in my original post, in most of the reports I've seen, the new MBPs to go up to full turbo and then only drop a few hundred MHz before continuing at a still heightened clock speed for hours. There are few other notebooks doing that.

P.P.S. I have an issue with the liberal use of term "throttle" to be honest. It should be reserved to situations where the cooling system is not adequate in dissipating the TDP. Instead, people use it for basically everything. Looking at the 8-gen 15W for example — these chips are designed to surpass their TDP for brief periods of time (thats where they get their excellent burst performance from). So even if you have adequately designed (15W) cooling system, it simply can't support the CPU running on its max power. Is it even fair to call this throttling? Now, if the cooling system could only handle 10W, yes, we could talk about throttling. In the end, the issue is that people fundamentally misunderstand how stuff works and then don't even bother to check the information they are mindlessly repeating.

I think this gets to the core of my angst with these reports...the term thermal throttling should be reserved to when a computer can't handle its specified TDP...which MBPs are doing, and then some at full load.

My own primary notebook is capable of running at full boost indefinitely as do many others. Not being capable of this is simply an indication of an inadequate cooling system.

Over powered silicon in a overly thin chassis is little more than a sales gimmick...

Q-6

I respectfully disagree with you calling it a sales gimmick, as I knew exactly what I was buying when I bought this computer, and I actually ended up with even more than I thought I was getting! Because from the introduction of Turbo Boost, I knew how it is designed to operate...short bursts of power to speed up tasks that cater to burst performance...It can only be considered a sales gimmick if you incorrectly assume that "3.1GHz, turbo boost up to 4.5GHz" means that you're going to get consistent 4.5GHz under full load. The turbo should be seen as a plus, and you should base your purchasing decision on the base clock...otherwise I would've gotten the 2.3GHz which actually boosts to 4.6GHz...

With my new MBP, im getting more than I paid for...apple over designed the cooling system which allows it to run in a slight turbo state indefinitely!! It will happily hum along 800MHz above the base designed clock speed for as long as I ask it to. So to call this an inadequate cooling system is reckless IMHO...you definitely can make the perfectly valid argument that if the cooling system were designed even more efficiently, then it would add some performance by allowing full turbo boost, but that is well above and beyond the call of duty for Apple...and I personally wouldn't want a thicker and heavier computer to allow for a few extra MHz as I carry my computer around all day every day in university.

An automotive example (because that is my passion ), the Ford Focus RS has an "overboost" function where the car's engine management computer temporarily increases the turbochargers boost output in increase in acceleration in short bursts. Ford makes sure to mention that overboost is only available under a certain number of conditions, one of which being engine temperature. If the engine is too hot (for example from already driving hard previously), the turbo will stay at its normal boost levels (base clock, if you will). Even when the conditions for overboost are met, the car will not stay in overboost mode for longer than 20 seconds, as the engine will quickly get too hot and it could cause damage. If you bought a Focus RS, would you call overboost a sales gimmick and the cooling system inadequate because it can't hold full overboost for an entire 3 hour racing session??? Further, its not really fair to say either than "oh well the ecoboost mustangs hold higher boost with the same engines just fine" because there are two issues with that statement, the Mustang is a completely different car with a completely different use case for completely different people. Because of this, the cooling system is designed to handle a higher boost level. Further, in the mustang it isn't advertised as overboost...its just advertised as the normal amount as thats what it is designed to handle.

 

[SteveJUAE]

Might want to check your facts, there are plenty of notebook's that can run at fullbore including significantly more powerful dGPU's. Without doubt the MBP is the thinnest.

Q-6

Agreed, runnng full pelt is not that meaningful comparison when it's coupled with a mediocre dGPU, considering these class of laptops ie it's already throttled by component selection

 

[Queen6]

Agreed, runnng full pelt is not that meaningful comparison when it's coupled with a mediocre dGPU, considering these class of laptops ie it's already throttled by component selection

True, reality is as long as the power supply is adequate and cooling solution equals or exceeds the systems TDP there will be no throttling, unless artificially set by the manufacturer.

Apple excels with it's engineering prowess allowing for thinner chassis. Others not so much as their design and focus is not solely focused on a single aspect. Some will conveniently omit that manufacturers can and do design cooling solutions that exceed the base TDP and encompassing the higher TDP of full Boost.

Way I see it is some manufacture's design and engineer their product to allow the CPU to run at full Boost, others have differing priorities. Where the OS allows I set my systems to be in Boost pretty much continuously. Plugged in this notebook rarely falls back below 3Ghz, under load a solid 3.4Ghz on all CPU Cores and potentially tweak it up to 3.8Ghz, add on the GPU same response with the GPU rarely exceeding 80C.

Q-6

 

[SteveJUAE]

True, reality is as long as the power supply is adequate and cooling solution equals or exceeds the systems TDP there will be no throttling, unless artificially set by the manufacture.

Apple excels with it's engineering prowess allowing for thinner chassis. Others not so much as their design and focus is not solely focused on a single aspect. Some will conveniently omit that manufacturers can and do design cooling solutions that exceed the base TDP and encompassing the higher TDP of full Boost.

Way I see it is some manufactures design and engineer their product to allow the CPU to run at full Boost, others have differing priorities. Where the OS allows I set my systems to be in Boost pretty much continuously. Plugged in this notebook rarely falls back below 3Ghz, under load a solid 3.4Ghz on all CPU Cores and potentially tweak it up to 3.8Ghz, add on the GPU same response with the GPU rarely exceeding 80C.

Q-6

Funny but notebookcheck quoted Microsoft on a similar point

"More and more laptops are not capable of maintaining their advertised performance under sustained load — yet not many reviewers test for this. (Source: Microsoft)"

 

[macdudesir]

Funny but notebookcheck quoted Microsoft on a similar point

"More and more laptops are not capable of maintaining their advertised performance under sustained load — yet not many reviewers test for this. (Source: Microsoft)"

Well gee I hope nobody tests Microsoft’s products (looking at you surface lineup...) to see how they do under sustained load...they will be the exact laptops they’re referring to

 

[Queen6]

Well gee I hope nobody tests Microsoft’s products (looking at you surface lineup...) to see how they do under sustained load...they will be the exact laptops they’re referring to

I'm of the exact opposite school of thought and hope they do get tested and more. As it's currently far too easy for manufactures to misrepresent performance. They should clearly indicate maximum Core frequency, by number of Cores and the interval of sustained load. In actuality all manufactures do the same, quote the max single Core frequency with absolutely no indication of performance under sustained load, and that should encompass dGPU's as required.

It's nothing about Microsoft or Apple being better than the other it's about us being smart consumers, and demanding more...

Q-6

 

[leman]

Way I see it is some manufactures design and engineer their product to allow the CPU to run at full Boost, others have differing priorities. Where the OS allows I set my systems to be in Boost pretty much continuously. Plugged in this notebook rarely falls back below 3Ghz, under load a solid 3.4Ghz on all CPU Cores and potentially tweak it up to 3.8Ghz, add on the GPU same response with the GPU rarely exceeding 80C.

So could you finally share with us what mysterious notebook that is?

 

[Queen6]

So could you finally share with us what mysterious notebook that is?

Similar to many other notebook's it simply performs as designed, nor is it unique or particularly special. Rather have the that feeding you'll spend the rest of eternity trying to prove how inferior it is in comparison the MBP which always entertains as your have absolutely no understanding of the usage or workflows involved.

Let's face it, your never going to purchase one, as your far too hung up on Apple...

FWIW it's not a Surface Book 2

Q-6

 

[SteveJUAE]

Well gee I hope nobody tests Microsoft’s products (looking at you surface lineup...) to see how they do under sustained load...they will be the exact laptops they’re referring to

I think you are somehow missing the point being made by notebookcheck and Microsoft

Maintaining high clock rates for long periods is not meaningful measure without the correct context nor is the occurance of throttling a given failure with insufficient clarity

It's about us consumers having the clear data to make an informed choice

 

[macdudesir]

I think you are somehow missing the point being made by notebookcheck and Microsoft

Maintaining high clock rates for long periods is not meaningful measure without the correct context nor is the occurance of throttling a given failure with insufficient clarity

It's about us consumers having the clear data to make an informed choice

I guess I did misunderstand...I saw it as a dig at manufacturers that aren't designing their computers to hold performance at full tilt because they aren't being held accountable...

I do agree with you and Queen6 that it should be talked about and tested more so consumers are better informed on what they are buying.

 

[kudos212]

LOL "This notebook I'm responding on is significantly more powerful than the MBP, nor does it throttle, as it's designed for heavy sustained use." This guy is lying out of his a$$ which is why he won't say what computer he has.

just too funny.

 

[leman]

Similar to many other notebook's it simply performs as designed, nor is it unique or particularly special. Rather have the that feeding you'll spend the rest of eternity trying to prove how inferior it is in comparison the MBP which always entertains as your have absolutely no understanding of the usage or workflows involved.

Wait, you don't want to tell us what notebook you got just because you think that this will annoy me? You really are a weirdo, man

Let's face it, your never going to purchase one, as your far too hung up on Apple...

I'm ok with downgrading my GPU if it means that I can comfortably carry around my laptop everywhere I go. Practicality trumps absolute performance demands in my case. Not to mention that I don't have that much use for a GPU besides gaming...

FWIW it's not a Surface Book 2

I gathered that much from your description

 

[SteveJUAE]

LOL "This notebook I'm responding on is significantly more powerful than the MBP, nor does it throttle, as it's designed for heavy sustained use." This guy is lying out of his a$$ which is why he won't say what computer he has.

just too funny.

I doubt you will get a response out of Q-6

He is known for many things but lying is not one of them