Severe thermal throttling discovered in Apple's M2 MacBook Pro
twitter.com
twitter.com
The SSD performance problem on the 256Gb models is a massive problem though.
Although after all this I'm still not sure why the M2 MBP exists. By the time it's useful it costs nearly as much as the MBP 14" for the same spec which is far superior. And if you want to cheap it out, the M1 MBA is probably a better deal because it's far cheaper and perfectly adequate.
> Export time: 21 min 40sec compared to M2: 19 min 40 sec.
So, the M2 is still faster, even throttling?
(Also no mention of ambient temperature, location of the device, etc)
German reviewer group, didn't find any of such problems: https://www.notebookcheck.net/Apple-MacBook-Pro-13-2022-M2-L...
> We can see a maximum package power of 35W for the M2 SoC during combined CPU/GPU load. However, this value can only be maintained for a couple of seconds before there is a drop to around 25W while the fans speed is slowly increasing. The package power will stabilize at 28-30W once the fan has reached its maximum speed, but the chip temperature is still very high at 98 °C. This means the cooling solution cannot utilize the full potential of the M2 chip, despite the fan.
This suggests OP is correct.
Care to explain why? I recall the speeds are still in the order of Gigabytes a second. Who is this actually going to have a significant impact on in the real world?
We don't have enough data to see if the problem scales to things like latency and fragmented read/write throughput (yet). I'll draw my conclusions then because that does have a massive impact on every day productivity.
So parallelization of read/write isn't possible. You have to wait on that one chip. Which kills random I/O speeds, which are the more real world model of usage than blasting a 500mb file all at once.
So while it is reasonable to worry about reduced parallelism in small SSDs when moving to newer, larger NAND dies, it is wildly incorrect to say that a 256GB drive has no capability for parallelization.
For any individual buyer, the upcoming Air or the existing 14” MBP would be better choices depending on your needs and budget.
With the first generation or two of MacBook Air, it was light but it was WAY less performant than even the entry level MacBook Pro. With Apple Silicon, the case for the entry Pro is very slim, but inertia keeps it going.
I think it's like how Apple kept making the eMac for a while, even after the iMac had eaten its lunch.
Looking at this guy's youtube channel, he's cranked out over six "gotcha" videos about the M2 alone. His channel is chock full of "WHAT THEY DIDN'T TELL YOU!" and "DON'T BUY UNTIL YOU WATCH THIS" clickbait nonsense.
Until someone who is reputable confirms this - like Linus Tech Tips - it's a non-story.
Eventually what we end up with is YouTube video lists that look like a competition for retarded peacocks trying to attract a mate.
Although Linus annoys the fuck out of me anyway.
makes browsing YT sort of bearable
What would a dev want that a gamer wouldn't? Battery life? Linux support?
GPU: depends on what you do, but pointless for many devs, unless you are into data science, useful for both gamers and creatives, but has different aspects
Display: devs - high res & reflection free, creatives - colour accuracy, gamers - refresh rate
CPU: not much difference, tho devs would likely prefer faster cores over more cores
Memory: dev's are most likely to run out of it, unsure of rest
Storage: creatives are most likely to run out of it and most likely to benefit of fast storage
Keyboard: dev - something that works more than 1 month (looking at you, 2017 MBP), creative - whatever, gamers - seemingly rgb and mechanical
Mice: opinionated, but you can't game on a trackpad
Footprint: creatives and devs are likely to prefer slimmer machines, but only to a point
Ports: as a dev, I really don't care, gamers aren't likely to care either, but for a creative I can imagine it's absolute dongle life
I typically just use my laptop docked except for short stints at the coffee shop, but if you're the type who brings it everywhere and uses it as intended (built in keyboard, mouse, monitor) it does make a big difference.
In some cases they change to a less clickbaity title after the initial surge of viewers has passed.
So they do like it.
From all appearances, this laptop exists solely to hit a price point with the name "Pro", primarily to appease enterprise customers that won't buy Macs that don't have the "Pro" badging. I would't be surprised if the M2-based Airs, which move the Air to the new design, actually have better thermals despite being fanless.
I think it exists because they have a ton of those outdated bodies and components that they would not have anything to do with.
They probably have a day's supply.
You know, I expect this kind of remark on a random twitter thread, but I'd expect folks around here to know better.
The point of this test is: if you push the hardware to it's spec'd limits the cooling system is inadequate and the result is massive pulsed throttling.
If your claim is "well just don't use the hardware to it's full spec'd limits", then I don't know what to say other than, that's absurd. I buy equipment expecting it to meet the specs on the tin. If it can't do that, it's defective, or Apple is lying to consumers. One or the other.
The spec on the tin under what circumstances? Do you except a computer to not throttle if run in 80˚C weather? Do you except a cheap car to be able to drive at its max speed non-stop (minus fuel up) for days/weeks?
See the discussion elsewhere in the comments about intended usecase. Nowhere on the tin does it say that either of these computers will run at 100% power without any slowdown. Nowhere on the Air does it say that it's designed for 8K rendering.
There isn't anything absurd about designing a laptop to allow for bursts of power, or even heavy sustained load on the cpu OR the gpu but not BOTH.
Are you sure you’re not getting them confused?
8600M had a design defect, bad composition of glue securing microbumps connecting die to package. Thermal cycling in high stress scenarios softened improperly selected underfill and broke microbumps connecting die to package. Repeated heating up can again soften the compound to release stresses and temporarily reconnect broken traces. It never fixes the main issue of broken chip.
https://en-academic.com/dic.nsf/enwiki/2011837#Problems
https://semiaccurate.com/2009/08/21/nvidia-finally-understan...
>On July 2, 2009, the date being ironically a year after the notorious 8-K that publicly kicked off bumpgate, the company put up a job listing for a “DIRECTOR OF PACKAGE TECHNOLOGY”.
https://notebooks.com/2008/10/10/apple-to-replace-faulty-nvi...
2010 saw 820-2850 with N8 (330uF) Tantalum caps frying under the GPU. For 2011 Apple "wisely" switched to AMD and we got 820-2915 with fried AMD GPUs :) Apple is simply BAD at cooling.
The person posted pictures of the original paste job from the factory. The amount of paste is not unreasonable, and what was instead evident was that the paste had dried out, which, again, is normal.
That didn't stop people from overreacting and declaring a corporation (one famous for obsessively controlling its production chain) to be idiots. It did not have "10x the thermal paste needed."
GamersNexus has solidly disproven the "too much paste causes overheating" myth. Excess paste just squirts out the sides, which is not a problem unless it is conductive: https://www.youtube.com/watch?v=EUWVVTY63hc
What caused people to really lose their minds was a screengrab that was claimed (with no evidence whatsoever) to be from the repair manual showing wildly inappropriate levels of thermal paste being applied. The screenshot was never confirmed and given the source - again, SomethingAwful - it was almost certainly not the repair manual or factory instructions (the factory instructions wouldn't be in English, for starters...)
https://www.techpowerup.com/11403/overheating-macbook-pro-my...
Seems Pepperidge Farm's memory isn't as good as you thought, or you're pushing some sort of agenda.
I've seen way too much thermal paste on Mac laptop boards.
Not just Apple; for a while I think 2007-2010, I was seeing a number of very inexpensive desktop computers with cooked thermal compound on AMD CPUs.
These observations aren't worth much, but it doesn't surprise me if a new Mac has thermal issues in the field. Although my M1 MacBook Air is the best, coolest-running machine I have ever had.
The claim was for the service manual, not repair manual:
http://forums.macnn.com/showthread.php?t=295925
Apple used to produce detailed service manuals in English:
https://www.applerepairmanuals.com/the_manuals_are_in_here/i...
More on the overheating issue:
https://paulstamatiou.com/macbook-too-hot-to-handle-literall...
> Remember that time about 15 years ago when MacBook Pros were all overheating and the root cause was Apple was adding 10x the amount of thermal paste needed at the factory?
Many people saw the images from the repair (oh, I'm sorry, service) manuals and directly claimed that was how systems were being prepared in the factory.
> Apple used to produce detailed service manuals in English:
I'm aware of that. The factory does not use repair (or "service") manuals, and they would not be written in English.
> More on the overheating issue
That post does not show a Macbook Pro, does not show over-application of thermal paste from the factory, and does not contain any evidence of impacted thermals.
Further, it claims: "anything more is overkill and can hinder the cooling properties of the heatsink."
Read this slowly: over-application of non-electrically-conducting thermal paste is not harmful. The heatsink clamp force is designed with the viscosity of the thermal paste in mind to generate the desired film thickness. It will just ooze out the sides.
Someone who over-applies thermal paste, or uses thermal paste that is too viscous, might need to wait some period of time for the excess paste to migrate out. Until it does, thermals might be impacted a bit. It is well known in the OC'ing PC community that you need to wait a day or two for thermal readings to be meaningful.
https://gizmodo.com/thermal-greasy-apple-sics-lawyers-on-som...
"After a Something Awful denizen took apart his MacBook Pro and discovered that Apple had slathered on far too much thermal grease, he found that using a more modest amount dropped his MacBook Pro's temperatures by several degrees. Now the forum has recieved a threatening letter from Apple's legal staff, requesting a link to this image [pictured above] be removed because 'The Service Source manual for the MacBook Pro is Apple's intellectual property and is protected by U.S. copyright law.'"
Like many others [1,2,3,4,5], I saw a lot of these MacBook Pro overheating issues resolved by removing the profusion of thermal paste and replacing it with a sane amount of Arctic Silver.
[1] https://www.teamfortress.tv/8399/macbook-pros-and-overheatin... "But what really fixed everything, was replacing the factory thermal paste with my own application of Arctic mx-4. My core temps are now 15-20 degrees celsius cooler. No joke [TL;DR]Apple did a really bad job of applying the thermal paste on my 2011 Macbook Pro. So bad that the excess of paste around the diodes were causing overheating. I cleaned up their shitty job and put on Arctic Mx-4 and now my fps is higher than it ever was. Here are pictures of what I found. This is embarrassing because it makes Apple look pretty bad."
[2] https://old.reddit.com/r/apple/comments/2xfzzi/if_you_havent... "I own a Late 2008 Macbook unibody, and lately the temps of the macbook were insane, until today. I was reading up on how Apple applies way to much thermal paste from the factory and boy were they right. I got around to opening the case and pulled the logic board earlier today. Looked at the thermal paste and it just seemed like they put it there. No real care for it working or not. Here: http://imgur.com/9z0cZwU Anyways, I cleaned up the thermal paste, and reapplied Arctic Silver 5. It made a huge difference! Temp at Idle before were roughly 55-60 degrees celsius. Temp at Idle after are at 40 degrees celsius."
[3] https://forums.macrumors.com/threads/its-official-macbooks-h... "It's official: Macbooks have bad thermal paste applied"
[4] https://www.pistonheads.com/gassing/topic.asp?h=0&f=95&t=496... "Early Macbook Pro - Overheating, too much thermal paste!"
[5] https://free-info-pages.com/my2011macbookpro/replacing-therm... "Why reapply thermal paste? Too much paste reduces the rate at which the CPU transfers heat to the heatsink. Subsequently, temperatures (both idle and load) increase. You can see the gobs that were used on my 15″ MacBook Pro here, on both the i7 2.2 Ghz CPU and the AMD 6750m GPU, as well as the heatsink"
But those high temps look brutal, as does the oscillating throttle, also nobody wants a fan running full speed non-stop if possible.
The M1 did end up hitting 100% on the fans, but capped out at 94˚C instead of 108˚C. It also did end up throttling slightly (dropping to 21.87W from a peek of 24W). The M2 meanwhile throttles from 29.46W to 7.3W.
The time comparison was M1: 21 min 40sec, M2: 19 min 40 sec. Much less than you would except given core and spec difference.
So yeah, technically the one fan isn't enough for all circumstances.
Is exporting 8K footage for 20 minutes of their entry-level laptop the target use-case? And if so, how much would fixing this be worth to the average end user?
Blaming bad thermal design on "use case" does not make sense to me.
If you'd like to run some intensive numerical programming on your laptop just because, I would find it outrageous for it to shutdown due to thermals. Can other hardware designers use that excuse too? "Oh your software use case is not covered by the cpu voltage regulators", "the power supply was not meant to be on all the time", "the power cable should not be connected for that long", "you should not use 90% of your ram for 4 hours straight, it was not designed for that use case"
"Oh the average end user only uses the car 1h/day. This 5h long trip was not the expected use case, that's why it overheated"
Cars can only rev the engine in the red zone for a limited duration. Lasers on CNC machines, welders, engine hoists, etc. all have specific capacities that you shouldn't exceed.
Expecting your CPU to run flat-out at 100% indefinitely without reprocussion seems strange to me. Heat is the great killer of electronics, and unless your rig is water cooled I don't know that I've used a laptop before with the thermal management to withstand extended full load- either it throttles down or it damages itself.
Like maybe you could change the throttling curve from (percent "max") 120-120-100-100-80-80-80-100-100... to a more consistent 90-90-90-90-90... but as long as the CPU knows internally when to throttle down to avoid damage, does it really matter either way? Fast-fast-slow-slow vs medium-med-med-med?
Setting your "it damages itself" remark aside for the moment, every laptop I've ever personally owned runs flat-out indefinitely at its rated maximum speed. I make a point of choosing laptops that have properly-designed cooling systems.
Electromigration is real, but that's a process that happens over many, many years. It is (_strictly speaking_) damage... but the only real way to entirely prevent it is to never use the device in question.
> ...unless your rig is water cooled...
In my personal experience, for mainstream (and, yes, this includes "workstation" and "server") CPUs, water cooling isn't notably better than a big-ass Noctua heatsink with a couple of their 140mm fans strapped to it.
Phase change "heat pipes" are really fuckin good for rapidly transferring heat from the contact pad of these heat sinks out to their radiating fins, and the big fans that are usually paired with them are quite good at cycling enough air through the system to ensure that the heat is moved out of the radiator in a timely manner.
You must be using a very nuanced definition of "rated maximum speed", incorporating time-dependent effects like Intel Turbo Boost, and CPU power limits that vary depending on power source, OEM-specific software options, and possibly discrete GPU usage. Which makes your claim almost a tautology.
<https://techreport.com/review/27909/the-ssd-endurance-experi...>
This is totally, legitimately incomparable to a CPU throttling to a fraction of its rated steady-state speed within a couple of minutes of constantly-applied workload because the laptop manufacturer decided to prioritize chassis thinness over cooling system effectiveness.
> If you'd like to run some intensive numerical programming on your laptop just because, I would find it outrageous for it to shutdown due to thermals. If you get a laptop without a fan and expect to push it like this then don’t be surprised if it can’t keep up.
Again, is the thing you want to do with the laptop within the scope of that the designers meant it to be used for.
That being said, I find it strange that their performance oscillates rather than hits a steady state. That part they should fix in a software update.
Apple does not sell any entry level laptops and this one starts at 1600€.
Macbook Pro is not an entry-level laptop. But you're right, this is hardly a typical use-case, like benchmarks almost never are.
I'm sure my Asus laptop would launch itself to the moon or perhaps spontaneously invent nuclear fusion if I did a 20 minute rendering on it.
So taking Apple's marketing literally, it seems exporting 8K footage would be a main expected use-case.
It's just the classic situation Apple continuously puts itself in with building a laptop that is starved for airflow.
They market it as being able to “work with more streams of 4K and 8K ProRes video”.
And all of the benchmarks are with respect to transcoding, and only 4K media not 8K.
I would 100% expect to be able to benchmark this computer for longer than 20 minutes without thermal throttling. The previous iteration was more than capable of it. This is new.
Without the specs these measurement don’t mean much.
Also HN: dismisses a real problem of their beloved CPU hitting temperatures that would make Intel laugh
Don't worry, denial is the first step to acceptance.
Wait, then what's the problem? It has a poorly configured thermal management leading to oscillating power/temperature. It's pretty sloppy on Apple's part, they'll almost certainly fix it in a software update, and it's not that likely most users will run into it before then. This guy needs to hype up every issue to drive traffic for his YouTube channel, fair, but not really something I'd personally get worked up about.
Edit: This kind of thing is why I don't buy a new model or upgrade to a new major OS version until 3-6 months after release when the major issues are fixed or at least known.
The new model hits the manufacturer's advertised performance in general, and is about 10% faster than the old model in the specific benchmark he did. The thermal porpoising is interesting but I wouldn't run around shouting about how there's some grave flaw. He does because it's part of running his business, it's not a knock on him but when stuff like this shows up on HN I figure we should have the context.