M1 Pro and M1 Max CPU Temperature Forewarning

[oz_rkie]

If one change saves you 35 seconds every time, and you never need to change it back, then that's a win. If the details mean you're constantly manually tweaking things and spending 3 minutes every time to save time on the encode, then reduce noise for the editing, then extend battery when undocked, then burn the battery when undocked for this one important job but then save battery for the rest of the trip, then it's a serious loss.

Why would you keep changing settings? You find the best OC at the best power and fan curve for your situation and call it a day.

Having them does hurt me. It makes it harder to find the few things I actually care about. It makes it more likely that I forget to change them back to a different mode under different conditions.

That's a weird way to look at it. I have a high end gaming desktop that I spent a while to initially setup, dial in a good OC at a good voltage curve and tune the fans (maybe an hour or so to setup and a couple hours of prime 95 to validate) and then run it 24x7 with a good 5-10% free performance ALL the time. My friend who also has a gaming desktop does not know about overclocking or tuning and does not care about it which is completely fine. He does not bother ever going into the BIOS or installing and using intel XTU. He can still use his computer just fine for everything that he needs it for. But those avenues are still there for those that want to make use of them.

Its easy enough to hide 'advanced' options tucked away in the UI so casual users don't even see these options.

 

[mr_roboto]

I mean yeah, I just imagine these people acting this way If I brought this up in conversation in person. I'd be baffled. I guess it all depends on tone of voice too, which is hard here. Anyway, I come from the PC world and 100C is just insane. But if Apple says its okay to run it that hot, I guess that's okay...I wouldn't want to wear out the fans for no reason.

100C isn't insane in the PC world either. Common knowledge like this is often generated by internet echo chambers, and isn't necessarily real.

As an example, Intel's own Alder Lake i9 processors are rated to run at 100C. Look at the "Tjunction" spec:

Intel® Core™ i9-12900HK Processor (24M Cache, up to 5.00 GHz) Product Specifications

Intel® Core™ i9-12900HK Processor (24M Cache, up to 5.00 GHz) quick reference guide including specifications, features, pricing, compatibility, design documentation, ordering codes, spec codes and more.

ark.intel.com

Most of the noise about 100C being awful comes from PC overclocking enthusiasts. Few of them know as much about chips as they think they do, but they do observe things. One is that if you want to overclock a chip nominally rated to run at 100C, you might have to cool it to 80C or lower to make it work reliably. By the time the echo chamber is done with it, this observable true fact gets transmogrified into "OMG 100C is SUPER DANGEROUS!!!!".

But the real truth here is something far less scary: there are well known semiconductor physics phenomena which reduce the speed of transistors as they get hotter. Chip designers try to assure adequate timing safety margin in worst-case conditions, so whatever the top end of the temperature range is, that's where timing margins are worst. Now an overclocker comes along and wants to run the chip faster; they're blowing through whatever timing (note: not damage) margin was built into the chip and must do something about it or the OC won't be stable. The two easiest knobs for an overclocker to turn are reducing operating temperature below the manufacturer's spec, and increasing voltage above spec.

None of that applies to a M1 MacBook. Apple doesn't publish M1 chip specs, so technically we don't know if 100C is out of spec, but it's really common for chips like M1 to have a max operating temp spec somewhere in the 95C to 105C region. Combine that with the fact that Apple clearly designed the cooling system to allow the chips to get that hot, and it seems likely that it's in spec.

Well, I would not be so eager to trust manufacturers. I am not saying that in this case 100C is not fine or anything, it probably is within spec. But just to remind those who do not know or might have forgotten, just a couple years ago with their i9 based intel macbook pros, the same Apple engineers were fine to let the chip throttle just to keep the system quiet. Yes, that was a hot intel chip but the point is Apple has many times in its history favored form (noise in this case) over function. Again, not saying that in this instance with the ARM chip that 100C is not within spec, but just not to rush to 'trust' Apple

The i9 MBP throttling thing wasn't what you think it was.

In modern chips, the amount of power used by the processor is under software/firmware control. The first few weeks of i9 MBPs had a wrong or bad firmware image loaded by Apple's factory, and it was misconfiguring the i9's power limit registers with values more appropriate to a desktop i9 than a laptop 45W TDP chip. Since it was a 'K' model (fully unlocked) chip, it honored the misconfigured power limits.

The resulting problem didn't even have a chance to be a thermal issue. Instead, it was about the VRMs. Intel provides a special VRM overcurrent alarm input on their CPUs; if it gets asserted the CPU throttles itself to save the VRMs. Apple had naturally designed their MBP VRMs to supply about 45W sustained, but with the misconfigured i9 chip the VRMs were being asked to deliver something like 120-130W sustained, which caused them to assert the overcurrent alarm.

Once Apple shipped a fixed firmware image with power limits appropriate to the chassis and motherboard, i9 MBPs didn't throttle any more.

 

[Reggaenald]

Hey, I am not sure who you are referring to, but I don’t see anything wrong in the replies you have received.
I certainly stand by mine, as I am pretty certain Apple’s engineers will have thought long and hard about when the fans should kick in, and you seem to agree after all.

To put it simply, my advice is that we should trust the manufacturer.
After all, if something were to go wrong with these machines, they’ll be the ones paying the price for their mistakes.

…
What about MBP‘s from 2016-2020? The engineers clearly made some mistakes with those models that Apple never intended to stand up for.
And thermals were and are, for everyone still using machines from that time, often terrible.
Im not an engineer, but I also don’t need to be to know after years of use that custom fan rules better the performance of the machine considerably.
Apple engineers chose form (noise) over function, they must have known that. Apple also at any point could have pushed a software update fixing this, but they never did and never will.
Don’t just trust a company, that’s silly and dangerous. Apple‘s been caught cheating often enough the last 6 years don’t you think?

 

[EmotionalSnow]

Don’t just trust a company, that’s silly and dangerous.

Trust the engineers that are definitely more qualified than the average person to make judgements about stuff like this. Unless you have concrete proof that Apple's factory settings are bad all you are doing is creating panic among people who just shouldn't worry about stuff like this.

Choosing noise over performance is not cheating neither does it apply to Apple Silicon MacBooks at all. With M1 drastic compromises like these are simply not necessary anymore.

 

[Wando64]

…
What about MBP‘s from 2016-2020? The engineers clearly made some mistakes with those models that Apple never intended to stand up for.
And thermals were and are, for everyone still using machines from that time, often terrible.
Im not an engineer, but I also don’t need to be to know after years of use that custom fan rules better the performance of the machine considerably.
Apple engineers chose form (noise) over function, they must have known that. Apple also at any point could have pushed a software update fixing this, but they never did and never will.
Don’t just trust a company, that’s silly and dangerous. Apple‘s been caught cheating often enough the last 6 years don’t you think?

If I didn’t trust Apple I wouldn’t buy their hardware and use their software and services.
Usually mistakes are acknowledged and corrected, within reason, with replacement programs (Butterfly Keyboard, Delaminating Screen, etc…)

As for cheating, I literally have no idea what you are referring to.

 

[weckart]

I have a feeling Apple engineers might have spent some time deciding when the fans should or should not kick in.

They've never got it right in the past, though. Dead GPUs from overheating and damaging the substrate. Warped logic boards that GPUs rise from or have broken solder connections. The CPU might survive but the chips and components around certainly wont appreciate the heat.

Apple always prioritises quiet running over longevity.

 

[Force11111]

They've never got it right in the past, though. Dead GPUs from overheating and damaging the substrate. Warped logic boards that GPUs rise from or have broken solder connections. The CPU might survive but the chips and components around certainly wont appreciate the heat.

Apple always prioritises quiet running over longevity.

This.

I also wanted to bring up the i9 mbp that ran so hot it perma throttled. pretending apple engineers gaf about cooling is laughable.

They keep the fan profile conservative for battery life and noise. Those factors sell laptops.

 

[Wando64]

This.

I also wanted to bring up the i9 mbp that ran so hot it perma throttled. pretending apple engineers gaf about cooling is laughable.

They keep the fan profile conservative for battery life and noise. Those factors sell laptops.

The i9 throttled, it didn't melt.
There are two ways of achieving cooling on a laptop, Fans (or other forms of heat dissipation) OR throttling.

As for your last comment, long battery and low noise certainly sell laptops to me.

 

[oz_rkie]

The i9 throttled, it didn't melt.
There are two ways of achieving cooling on a laptop, Fans (or other forms of heat dissipation) OR throttling.

Or you know, the 3rd way of not putting in an i9 when you are sure it would always get throttled down to an i7 or an i5 level. But Apple knows that they can get away with it, marking up that i9 pricing, their loyal fanbase will always find a way to justify Apple's decisions like 'throttling as a novel cooling strategy'. lol. First time I've ever seen someone justify throttling as a legit way of cooling a chip and not just a thermal protection.

 

[Sharewaredemon]

Oof, hopefully they’re manufacturing them to last longer than their warranty period! Isn’t that like.. 1 year? Haha. I mean I have Apple Care for the next 3 years, but I had to pay for that. That’s where my caution comes from. This is a $4200 laptop. Not trying to just fry it in two years.

tl;dr - most Macs last 10 or so years.

I would say @quarkysg means (or at least agrees with me that it's) 3 years instead of 1 (at least for Apple).
The way I see it, Apple is looking to make a lot of money on AppleCare, so they are designing their computers to last more than 3 years, which increases their odds of making bank on AppleCare. As well, I've worked in tech education for over 10 years and it's always amazing how long lasting the bulk of the Apple computers we buy are. Yes things break, and things are also broken by users, but generally speaking, most computers are retired/recycled when they aren't useful anymore, as opposed to breaking after 3 (or before) the 3 year mark.

 

[Wando64]

First time I've ever seen someone justify throttling as a legit way of cooling a chip and not just a thermal protection.

I think you are quoting me out of context (easily done by reading just a comment).
My comment was about fears of “CPU about to melt” and accusations of “warping motherboards”.
Throttling cools the system (yes, as a form of thermal protection) and prevents those outcomes.

 

[TracerAnalog]

For clarification -- I wasn't intending to offend, if it seemed so. My point is simply that this has been a contentious subject on these forums, one that industry experts (e.g., @cmaier) have weighed in on with a resounding "no, this isn't an issue --- they're meant to run that way."

A primary issue is that there is a lot of confusion around the topic. Mostly, you'll see hardcore gamers or tech reviewers spewing that 100C is too much, but what they're really commenting on is that once you get to this point, it's likely that thermal throttling is soon to follow. Silicon is not going to melt away at 100C, so whether you're on a Mac or PC, you're not in danger of destroying your CPU. What you will be in danger of (if your cooling solution is not good enough), is thermal throttling -- which, as @Technerd108 pointed out, doesn't seem to be an issue with these machines, as they can very easily maintain optimal max, non-detrimental temps (of around 100C). Previous generations were unable to do this (both due to the silicon designs in them being more power hungry, and the thermal designs being more restrained), so if you hit 100C on an older Mac, you'd be seeing thermal throttling very soon after (as frequency would necessarily need to be reduced to remove heat at an appropriate rate).

I have played with the cooling settings on many of my PCs, and always ended up setting everything back to default. In the end I cannot be bothered and hope engineers know what they’re doing ?

 

[now i see it]

While 100°C may seem hot, I’ve never heard of a Mac ever burning up in 30 years - and the Core2Duos regularly hit over 100°C when rendering anything. No problems.

 

[LinkRS]

I always purchase AppleCare (AppleCare+ for my 2021 16" ), and have trusted that my systems won't fail due to overheating. My 2012, 2015, and 2019 (16") MacbookPros often hit 100C, then the fans would kick on, and drop it down to 80C or so, and continue to run. With the 2015 and 2109 models, as soon as I employed the AMD GPU, the fans would kick on, and stay on. The strip of metal between the display and the function keys (or TouchBar for the 2019) would get too hot to touch, but they never noticeably throttled. Conversely, I have yet to hear the fans in my 2021 16", but that strip of metal still gets pretty warm on this one too. I trust that if there actually is a problem caused by Apple not cooling adequately, they will have to replace my system under AppleCare. . Apple seems to favor temperature over noise. Despite this, I don't think they will let it get too hot, to the point of damage to keep it quiet.

To the OP: If it makes you feel better running the custom fan curves, go right ahead and ignore the advice to "leave it alone." One thing is for sure, is you will not cause any damage if your MacBook Pro runs cooler , and something can be said for piece of mind

 

[mi7chy]

This is all very true. Maybe I’m just too accustomed to the never ending tinkering in the PC world. It just makes me nervous about the longevity if temps of the CPU, GPU and RAM are often in the 80-90C range.

I like to tune CPUs and GPUs and target for 70C for low fan noise and max ~80C. Better to err on the side of caution.

https://www.reddit.com/r/mac/comments/sm6bhg

 

[januarydrive7]

Or you know, the 3rd way of not putting in an i9 when you are sure it would always get throttled down to an i7 or an i5 level. But Apple knows that they can get away with it, marking up that i9 pricing, their loyal fanbase will always find a way to justify Apple's decisions like 'throttling as a novel cooling strategy'. lol. First time I've ever seen someone justify throttling as a legit way of cooling a chip and not just a thermal protection.

@mr_roboto already addressed this.

I'm using my i9 16" mbp right now, training a parameterized hyperband deep neural network tuner in the background, and it seems temps have stabilized to just under 100C, with fans at ~70%. Mind you, I'm in the southern CA desert, with AC off and ambient temp is about 25C.

To be clear, throttling is an effective way of protecting a chip from overheating, but it's not something that M1 Pro/Max machines suffer from in any way, as the safe 100C is able to be maintained entirely from the thermal designs.

 

[mr_roboto]

I like to tune CPUs and GPUs and target for 70C for low fan noise and max ~80C. Better to err on the side of caution.

https://www.reddit.com/r/mac/comments/sm6bhg

LOL. To an actual engineer (I am one, if that's not clear), that reddit post can be translated from the original naive thusly: "A defective electrolytic cap short circuited itself and did the things you'd expect - made a popping noise, maybe arced a bit, released the magic smoke. Computer wouldn't turn on again because that's a bit difficult with a short on a power rail. Apple replaced the motherboard because they don't do component level repair."

Despite your attempt to troll with the incident, it obviously doesn't have anything to do with how hot they're letting the CPU run.

 

[UBS28]

You are fine. The cpu is meant to run up to 100c without throttling or having any thermal problems. The fans come on once it reaches that temperature to make sure it doesn't get any hotter and throttle. You really can't push these cpu's to a point where they throttle or the cooling solution can't keep up. The thermals on these new MBP are really good!

It is better not to hit 100c on your cpu constantly.

It's like driving your car in the red constantly, which will blow up the engine at some point.

 

[UBS28]

Well, I would not be so eager to trust manufacturers. I am not saying that in this case 100C is not fine or anything, it probably is within spec. But just to remind those who do not know or might have forgotten, just a couple years ago with their i9 based intel macbook pros, the same Apple engineers were fine to let the chip throttle just to keep the system quiet. Yes, that was a hot intel chip but the point is Apple has many times in its history favored form (noise in this case) over function. Again, not saying that in this instance with the ARM chip that 100C is not within spec, but just not to rush to 'trust' Apple

Apple could have clocked the M1 Max higher if they actually allowed the cooling to do it's job. But Apple was clearly aiming for a "quiet" laptop, which they succeeded.

Will be interesting to see what they will do with the Mac Pro, if they favour performance over noise.

 

[bobcomer]

It is better not to hit 100c on your cpu constantly.

It's like driving your car in the red constantly, which will blow up the engine at some point.

I don't think you can call 100C red, more like yellow to me.

My M1 MBA is over 100C quite often, and has gone as high as 118C. 118C was red, machine slowed down quite a bit and you don't want to touch the machine right over the SoC... No damage long term from it though.

 

[NT1440]

It is better not to hit 100c on your cpu constantly.

It's like driving your car in the red constantly, which will blow up the engine at some point.

Is it though? Or is the 100c a “common wisdom” number that fell into place years (decades) ago and is no longer accurate for the materials in use today?

That’s an honest question. I’m coming at this from the frame of many users who insist that their computers HAVE to reboot every night for them to work properly. It’s not true…but as their IT guy I’m not going to correct them because there are other benefits to me letting them continue with that mindset ?

 

[Krevnik]

It is better not to hit 100c on your cpu constantly.

It's like driving your car in the red constantly, which will blow up the engine at some point.

There's also a huge difference in the physics occurring in solid state electronics and a combustion engine running at thousands of RPM. Silicon does have one annoying tendency, and that is its electrical properties change with temperature. So you need to keep the temp in the "good range" where glitches aren't introduced when your circuit is complex, but that is well below the point of damage to the substrate.

Back in the day when thermal protection wasn't a given in CPUs/SoCs, it was definitely possible to fry a chip by yanking off the heat sink and triggering thermal runaway. But Intel had started putting in thermal protection years ago. The Pentium III was able to simply crash, but not damage itself. Pentium IV being one of the first to throttle to protect itself. The AMD Athlon at the time could die, but that's because uncontrolled thermal runaway takes the temps well beyond what doped silicon is meant to handle (370C is a lot).

Apple could have clocked the M1 Max higher if they actually allowed the cooling to do it's job. But Apple was clearly aiming for a "quiet" laptop, which they succeeded.

I kinda have to parrot the chip engineers on this one: It's very likely that the component libraries Apple used in the design don't scale up much in clock speed from where they are already at. Cooling is not the main limitation here.

 

[januarydrive7]

Is it though? Or is the 100c a “common wisdom” number that fell into place years (decades) ago and is no longer accurate for the materials in use today?

That’s an honest question. I’m coming at this from the frame of many users who insist that their computers HAVE to reboot every night for them to work properly. It’s not true…but as their IT guy I’m not going to correct them because there are other benefits to me letting them continue with that mindset ?

AFAIK, max operating temp is something that is within a design spec, and that max temp is not a limit set for preventing damage, but for preventing logical issues -- i.e., you can safely operate up to temp X (safe both in terms of will not cause damage to your system, as well as will not invalidate what your CPU is computing). Also AFAIK, these max temps are well below what you would need to actually damage your CPU.

 

[bobcomer]

(370C is a lot).

Cool article!

 

[mr_roboto]

Or you know, the 3rd way of not putting in an i9 when you are sure it would always get throttled down to an i7 or an i5 level. But Apple knows that they can get away with it, marking up that i9 pricing, their loyal fanbase will always find a way to justify Apple's decisions like 'throttling as a novel cooling strategy'. lol. First time I've ever seen someone justify throttling as a legit way of cooling a chip and not just a thermal protection.

As I already explained to you, this is not what Apple did with the 2018 i9 16" MBP at all. It had a cooling system compliant with Intel's specs.

The 2018 16" MBP was available with three different CPU SKUs, which I've pulled up in Intel's ARK database to prove the point:

Intel product specifications

Intel® product specifications, features and compatibility quick reference guide and code name decoder. Compare products including processors, desktop boards, server products and networking products.

ark.intel.com

As you can see, all three have the same 45W TDP. Did the 2018 16" i7 MBPs throttle? Nope.

This isn't because of some convoluted theory about Apple exploiting fanboys, or some made-up story about people justifying use of throttling (who exactly is doing that???), it's because there was a bug which caused the i9 to run at desktop-like power levels. In a notebook. That was never going to go well.

And yes, it was a bad bug. If I recall correctly, under heavy loads the i9 MBP underperformed both i7 options. But also, please pay attention to what actually happened even if it's inconvenient for your pet theory about Apple. All these problems literally disappeared a few weeks after product launch! Apple fixed them completely with a software/firmware update. Once installed, throttling went away and the i9 machines benchmarked faster than the i7 models, as expected.