Why is this news again now? It was exactly the same design limitation from the M1 generation. It's not a bug, it's a known limitation of the design feature(fanless operation).
Modders would work around this by putting thermal pads in between the heatsink and the back cover to transfer the heat to the aluminum under-chassis increasing the surface area and thermal mass and getting "free" performance capable of reaching the more expensive Pro models during extensive workloads like rendering.
Provided of course you never want to have the laptop on your lap ever again after that, but if you're someone like me who always uses their laptop on fixed surfaces like beds and tables and never on body parts, and you really really need that extra bit of performance, then it should be fine.
Does it? Is the performance per watt of the M3 SoC worse than the performance per watt of the M2? Is the average and peak performance of the M3 MacBook Air worse than that of the M2 MacBook Air? If not, how is this a step backward?
The gap is increasing because these chips get a higher performance (and power) ceiling. That's a good thing. It means that the MacBook Pro -- a machine with a much better cooling solution -- can get much better performance. The M1 SoC didn't have a high enough performance ceiling to really take advantage of the MacBook Pro cooling solution, so it wasn't much faster than the Air. That was a bad thing.
"M2 to M3 Air isn't as significant as it should be" might still be true though. How did you arrive at this conclusion? What benchmarks are you basing your opinion on? What is the actual improvement from M2 to M3 according to the metrics you care about, and what do you think the improvement should have been?
> The M1 and M2 generally performed the same for light and medium-sized tasks whether they had active cooling fans on them or not, and the M3 acts the same way.
Compared to the M1, a fully enabled M3 with eight CPU cores and 10 GPU cores running some typical benchmarks has roughly 25 to 30 percent faster single-core CPU performance, 35 to 40 percent faster multi-core performance, and between 45 and 65 percent better graphics performance, depending on the test.
The numbers speak for themselves, though—however fast it throttles, a passively cooled M3 runs faster than an M1 or M2, and it stands up well against modern actively cooled laptops with Intel and AMD CPUs inside. But if you regularly stress the CPU and GPU, you're probably leaving a bit of the M3's performance on the table relative to the same chip in the $1,599 MacBook Pro.
No? The linked article is saying that the M3 MacBook Air is less powerful than the M3 MacBook Pro, and yeah, you can sustain higher power consumption for longer with a fan than without, no shit. The M3 MacBook Air is significantly more powerful than the M2 MacBook Pro, which tells us that it's a step forwards.
Hacker News's title length limitation makes the title here incredibly misleading, the full title says "33% Slower Than MacBook Pro".
But if you don't spend even $200 more on the Macbook Pro to get 16GB, you are leaving a bit of the M3's performance on the table due to swapping. So, $1799. Or step up to the M3 Pro with 18GB at $1999.
And on the Air you could drop that same improvement in. So, $1,299.
Which says to me that from $1000 to $2000 there are a variety of ways to get satisfactory performance out of a long-battery M3 Macbook.
A corollary is that there is no way to get everything (unlimited multitasking, unthrottled renders, no swapping, no IO constraints) you could want out of a $999 Macbook Air M2/8GB/256GB.
I’m more surprised the CPU is allowed to peak at 114°C at all.
x86 CPUs and GPUs usually have a recommended load temperature of ~85°C and start throttling heavily past ~100°C, with emergency shut-off at ~120°C
I know Apple has always preferred running MacBooks hot and silent instead of cool and loud, but are they using a different solder or PCB for the M-series?
Perhaps it’s a new way to reintroduce planned obsolescence. If consumers can detect throttling, then why not just make hardware run hotter so eventually solder points come off and bake PCBs to randomly introduce errors to get consumers to buy new computers?
Not just the solder but x86 CPU’s start to break down at that point. The overclocking groups have information about this but as I understand it the actual physical transistors on the die start to break down within a period of months leading to degraded performance, errors, and eventually non-operation.
Granted, the chip may be designed for these temperatures, but I would assume that apple does what other manufacturers do and just assume the cases of users where high heat is not intermittent and causes premature failure will be rare enough to be inconsequential.
Speaking Intel for a moment, since I don't know all the details for AMD and Apple, their CPUs are nowadays designed to boost until they reach 100C if they weren't already power or frequency limited by then. Intel also provides a 3 year warranty when the CPU is operated to spec, which includes the aforementioned.
AMD does likewise, though their CPUs' thermal limit is 95C from what I know; I assume they have a similar warranty period to Intel.
Apple also seems fit to operate their CPUs like that.
Given all that, I'm going to presume that advances in fabbing tech mean CPUs today tolerate high temperatures better than ones from yesteryear.
Ryzen 7000 series all have a TjMax of 95C. I was quite surprised to observe this getting one recently. The motherboard's fan controller seemed unconcerned until the chip hit 100 and even then didn't throttle the CPU, it just cranked the hell out of the fan.
I guess if they've determined it's not detrimental it's fine, but on a MacBook Air with no fan, I can't imagine you've left much performance on the table by heat-soaking those extra ~20 degrees. The way these chips heat up that's gotta be just a few more seconds at full speed.
Not sure if you’re aware of this (and if you’d be comfortable doing it, considering the risk), but delidding the 7000 Ryzen’s and applying Liquid Metal cooling paste drops average temperatures by ~20°C (!). Remember to apply to both the cooler and the CPU.
This is because the Internal Heat Spreader for 7000 series was made primarily with cooler height compatibility in mind. It’s compatible with all previous (aftermarket) AMD coolers, but because the chips are lower height they had to make the IHS thicker which lowers the velocity of heat transfer.
Too risky for my taste! What I ended up doing is lowering the TDP in the BIOS from it's default 105 to 85 watts and now temps are great. I have a 7700X.
Before doing that I tried lowering the max temp instead but I think 100+ watts is just too much for the cooler I have and I couldn't tolerate the noise. I'm sure I've lost some performance but I can't tell the difference.
From reviews I've seen, the M3 Air doesn't burn your lap as much as x86 laptops with active cooling, at least in sustained gaming workloads.
> During the game, the surface temperatures of the laptop certainly got warmer, though never uncomfortably hot. Many gaming laptops I’ve reviewed get much hotter (and louder, of course) during games. Looking at you, Razer Blade.
It seems like everything must be modified to make it news. The Air (like all the previous Airs including that Intel era short-lived Core M3 model) was always designed with burst capacity in mind, and midrange performance overall.
Edit: midrange being midrange web and productivity, not midrange gaming, midrange CAD, midrange 3D modelling or midrange video production. (but you could always do those for at bit.. or in bursts)
You buy it if you want to get something in the ecosystem, but don't need much, and that is okay. Not everything has to be maximum all the time.
But then these laptops are really expensive, if they’re basically just meant to be chrome books. The other day there was a thread about some pc laptop review with some commenter mentioning how they find it odd that all non Mac hardware is full of weird compromises.
But the MacBook Air with 16g of ram and 500gig hdd costs like 1500$, but is apparently still only „meant“ for light browsing, lacks ports, only comes with a mirror (glossy) screen, and the OS is missing a bunch of basic functionality.
The MacBook seems expensive and full of compromises.
> but is apparently still only „meant“ for light browsing
Where do you get that idea? The M1 Air was already more than enough for all my web dev needs. I have lots of docker containers running all the time and I don't notice anything.
Depending on your domain they can make decent developer machines at the 16+GB config.
I have been researching this lately. At the 24GB config it is looking like the M3 air puts down about the same real world CPU performance as my 2019 16” i9 MacBook Pro. For light docker and development use cases these look really compelling to me.
That is what we have been rolling out to cloud-native developers as well, they get to choose etc, but when you don't need to run a local Kubernetes cluster with heavy tools, but instead can do something with a container, some FaaS simulation etc. you're pretty much golden with the 24GB configuration. Even 16GB used to be plenty, but with the GPU zero-copy sharing and the runtimes eating a bit more memory, the extra couple of GB do really make a difference.
The amazing thing is that with this iteration, you're really leaving practically all Intels behind (at least Alder Lake and earlier) as you're either just as productive, or more productive, but without the heat, power draw or constant charging.
If you look closely, you might have noticed that I wrote that you might get this type of device when you want something in the ecosystem but don't need much.
To pre-empt people who want to qualify "much" as something else, I even included an edit so you don't get people with a different idea on what "much" might mean in this specific context (as it has for over 10 years, none of this is really new). It appears that did not reach you in time.
You specifically probably don't want something in the ecosystem, and you do need much. As such, you should probably not buy this. Other people do exist, and they might buy it if it fits them. It doesn't matter if the numbers on paper tickle you the wrong way.
The whole argument of the GP is that you shouldn’t expect much of these machines, meaning the kind of tests the article made are moot.
So which one is it? Is it a device with great performance, or not? It’slike people say „these things have great performance“, then somebody comes along to run some test saying „these machines throttle like crazy if you tax them, they have poor performance“, and then the the first group comes along saying „you expect too much“.
I guess they have great performance, as long as you don’t need it?
Performance in a vacuum doesn't mean much. How something performs depends on the task. If you take all the laptops in the world and ask them to dive to the bottom of the ocean, they all perform badly (because they will not reach the bottom and they also won't survive, double failure).
This is also why a lot of 'tests' doen't really tell you much on the actual use of the machine. Even if we can profile all users, and bucket their tasks in terms of system load you don't get a perfect testing system to compare machines.
How 'good' something is, is almost a quality factor where the only real thing you can describe is "to what degree did the device do what you expected it to do". Good quality would be "meets expectation", bad quality would be "did not do what was expected" (and you can extrapolate from there). This is also where consumer satisfaction is generally measured, people get satisfied with their stuff if they can do what they wanted to do, not because of numbers on paper or tests on a website. Ironically, this is also where building your own PC with some cool GPU, Ryzen, and a bunch of M.2 SSDs tends to be satisfying, not because those component's numbers enabled your day to be better, but because you built something and you can use it, you feel ownership and accomplishment, and that's a quality you can be satisfied about.
For an Air, the best you could probably do is describe what a user wants ("I would like to create some content, consume some media, play the odd game here and there, do some software engineering, some 2D work") and then after a while ask them how close the machine performed to their expectation. And even that will simply vary from user to user.
Even work fleets have this problem, IT departments love to standardise, but unless the work all employees perform is specified to the degree where it could be completely automated you won't be able to construct a homogeneous fleet that covers everyone's needs. So you'll always end up with cases where it's an expensive solution, a cheap solution, a bad fit, a good fit, and that just gets averaged out on a quarterly report so we can call it 'good enough'.
I guess my point is "The Mac mini starts at $600 and even though it has 8 GB RAM you aren't going to find another computer anywhere near as performant in that price range."
It depends on the spec - everyone dumps on the base specs but they are still a crazy powerful Chromebook, and in the higher specs they are legitimately competitive as ultrabooks and developer workstation spec laptops.
In neither configuration are they particularly cheap, but you’re also holistically paying for a first-party supported Unix OS with great third party support, a bandwidth level 2-4x greater than x86 laptops currently deliver (with unified memory and a good gpu for the power), very good screens/keyboards/trackpads and an aluminum chassis, a huge amount of thunderbolt with all ports at full spec, incredible battery life, etc. No, apple doesn’t compete dollar for dollar with an atom based Chromebook with a plastic body or whatever, but you also do get a much nicer machine.
This should go without saying (but seldom does) but different people have different value judgements, and different products exist for these different people. If you want to min/max your life with a $300 chromebook that's fine. Other people prefer to look at it in the "I spend 16 waking hours staring at Screens every day, I want it to be a good one". And if you are otherwise in the market for a well-specced x86 laptop, the apple stuff is not that bad, and while it does have its downsides and weaknesses it also has strengths and unique points over x86 too. You're not going to go run a LLM on that 6840U or 12700H. Have fun getting sleep to work right/waking up to a cooked laptop in your bag. Remember bluetooth problems? Me neither, they disappeared when I got issued a macbook (i9 even but the apple silicon models are better).
You see the same thing in GPUs a lot too. People fucking obsess about some numbers that largely come down to 10-15% one way or another, and you know what? If you want the NVIDIA features or (perceived or real) driver stability/herd immunity, it's fine. If you want the AMD VRAM and perf/$ advantage it's fine. It ultimately doesn't really matter all that much.
People get super wrapped into "but this one is 15% better perf/$!!! that's like half a segment!!!" but that means we might be talking about a whole $50 at the register, and if it doesn't do what you want it to, if it's going to nag and chafe at you, then just spend the money. And for some people, especially with larger purchases, there is also the psychology of "I'm spending $2500 on this Thinkpad anyway, I'm keeping it for years, I don't mind spending a little more to check boxes". Especially for what amounts to craftsmen choosing tools - we are going to stare at these machines for a long time, right?
Anyway like, there's zero question in my mind that the apple stuff smokes the latitudes or thinkpads you'd otherwise get. They're great.
I'm helping my mom pick an M2 or M3 MacBook Air, and your comment keeps me on the fence. If it's just burst capacity for web browsing and simple home office activities, then why not save $100 and get the 2022 M2 machine?
Whereas most compelling reason I've seen so far for M3 over M2 is Ars Technica's review yesterday--get the M3 for the extra 2 years of supported macOS updates. Well, except that shouldn't be a compelling reason.
I think that long-term included manufacturer support is always neat. If you have a machine that is supported for 7 years or 9 years (or 5 years) that is all a factor that a general user might be interested in.
> … getting "free" performance capable of reaching the more expensive Pro models during extensive workloads like rendering
Really need to make sure you do not keep that thing on your lap or the cost of fertily treatments from your baked testicles will greatly exceed the cost of a laptop stand or even a Pro.
This is true, although in my case the Linux box is a VM running on Ivy Bridge-era Xeons, so my M1 Air handily outpaces it for single core performance.
If I ever get new hardware in my rack, this is absolutely the path I’ll take. I love the battery life and lightness of the Air, and honestly its performance is nearly always adequate for me. If it isn’t, a U worth of Epyc would solve that nicely.
Talk about diminishing returns on time invested in applying a thermal pad(10 minutes?) but then continue talking about spending time building a linux box in a closet instead. :SMH:
Some people choose laptops over building linux boxes in closets for a reason. I can't take my closet PC with me on a train or a coffee shop.
Then there's the issue with space. If you live in a small European apartment like I do, then I have no closet, so no extra space just for a separate PC out of sight.
Then I would still need a laptop for SSH into it anyway, that closet PC won't replace it.
It's a dumb feature, though. The only benefit is fashion. There is absolutely a market for people wanting a reasonably-priced machine with reasonable (16G+) memory and reasonable performance (which requires a fan) for typical workaday jobs. That's right in the middle of the "boring windows enterprise" segment, and it sells a ton of hardware.
But Apple won't serve that market for segmentation reasons, leading to this weirdly crippled Air[1] that for a lot of geeks is sort of the worst of all worlds. Really you might as well just buy a Chromebook for half the price
[1] Whose only real selling point vs. the putative 16G+fan device we all actually want is that it's so thin you can mount it on a pole as a halberd blade for 1d10 damage.
> That's right in the middle of the "boring windows enterprise" segment, and it sells a ton of hardware.
The devices that are sold in this market are as expensive as MacBooks or sometimes even more expensive. So I don't think this is an issue.
Also if the MacBook Air ran Windows it would be the perfect device for most workloads in this market. Because the workload common there is running a Browser, Outlook, Word, Excel and SAP (or any other ERP and CRM of your choice).
In what world is the Air's performance not reasonable? I have an M1 Air with 16GB of RAM and I've never felt like it was lagging behind me. It's my main computer which means it's used for work and every nerdy experiment I might run.
The best part is I can do all of that shit all day long on the battery and the laptop is super easy to move/carry around.
With no modifications or extra effort any thermal throttling I've ever encountered hasn't been noticeable.
When the M1 Air first came out there were the breathless articles about thermal throttling. It's only something that will come up with extended periods of high CPU usage.
I've tried to get my Air to thermally throttle and have only been able to do so doing 3D renders or running local LLMs for long periods (over an hour in my testing). Even long compiles don't really stress the CPU so much it throttles noticeably.
The simple solution in that case was a $20 laptop stand with a fan off Amazon.
If your use case is 100% CPU usage all the time either don't get an Air or get a cheap laptop stand with a built-in fan. If your Air isn't your CI/CD build machine or part of a render farm it'll be just fine.
I would love to have the option of running the machine at full speed with the tradeoff of sizzling hot surfaces. Cool when on the go, full power when laptop is sitting at desk well ventilated.
From a performance point of view there's almost no point in the article. You want to do long running intensive work, get something with active cooling.
Except... isn't 114 C a bit much for something that sits in your lap? How about on bed sheets? Could it melt some plastic-y table cover?
114C is the CPU temperature, we have no idea what the surface temperature of the MBA is.
These types of safety considerations are normally bread and butter considerations to any electronics manufacturer - there are plenty of guidelines and standards for them to reference and test to.
That said, the safety limits are that - safety limits. They are not necessarily comfortable. I think the safety limits for continuously held metal parts (so a MBA laptop) is like 55C. I think the rough guideline on "comfortable" is usually considered sub 40C.
Here's a review that tested sustained gaming workloads on the M3 Air.
> During the game, the surface temperatures of the laptop certainly got warmer, though never uncomfortably hot. Many gaming laptops I’ve reviewed get much hotter (and louder, of course) during games. Looking at you, Razer Blade.
Fearmongering aside, from what I've seen, the reviews say it gets warm but not hot.
Tom's Guide broke out the thermal gun and said the hot spot on the bottom of the 13 inch model was about four degrees F above normal body temps while running Cinebench on a loop. The 15 inch one got about 9 degrees F above body temp.
You certainly wouldn't need to worry about melting your table.
Because not everyone is a nerd immersed in this particular niche 24/7, and as a casual consumer for example, it's good to know not to accidentally buy a laptop you can cook an egg on.
What an utterly disappointing article. 33% Slower, after how long? At start? After it fully throttles? The 2 bar graph showing difference is first run? n-th run? How long is the run? You can mod with thermal pads to keep it running at full speed much longer, but how much longer?
To be fair it’s not 33% slower at opening vs code, and having 30 browser tabs open with slack while using a hot reload compiler on a big code base… It’s 33% slower while running some synthetic 3d video processing benchmark. And even the other 66% is pretty good
Right, so don’t buy an Air if this is a regular part of your workload. You paid less for an air and you are getting less performance under heavy extended load. Sounds about right.
Depends on the video chat service being used, but in some cases you can force them to use a codec with GPU accelerated encode/decode by running the web version in Safari, which reduces CPU load and heat a lot.
It might’ve changed since then but the difference was pretty pronounced with Google Meet a few years ago. In Chrome it roasted laps but in Safari it barely moved the needle.
Yeah that's what happens when you make it thinner and without a fan, you run it at lower power and it can also get hotter...
This is such clickbait it's no wonder why there are people that have a seething hate for Apple. Nothing has competed with Apple laptops since the M1 dropped.
It might be exaggerated, and the Apple Silicon MacBooks still fall behind as soon as gaming or other GPU-centered tasks come up.
But having a super thin laptop that does 4K Video editing and has an really good battery life doing it, is amazing.
What they can do with 30W from the wall is simply amazing. Not even talking about the Pro/Max Chips that can decode and encode several ProRes Streams at once.
But again, some people don't just edit hi res video.
Give me at least one example of a fanless laptop with performance, weight and battery life of M1 Air, I’ll wait (probably forever, because I haven’t found one).
It isn't hard to create an arbitrary set of constraints to single out a single laptop. Just because some laptop has a fan, which is often a complete non-issue in non-Apple laptops (Apple's laptop fans in the models that have them are some of the loudest I've experienced), doesn't mean that it doesn't compete and that there aren't tradeffs. For example, Apple laptops are extremely uncomfortable on your actual laptop due to how hot they get.
It's a tradeoff, not a problem. My XPS has fans and is silent except when doing something very intensive, like gaming, which macOS is terrible at if a game is even available for it. If the XPS gets dusty, which it hasn't so far, then I can easily clean it by removing the back panel since the XPS is made to be repaired and even upgraded. So if you truly want a fanless laptop, then it looks like the Air is the main contender, but then you have to use macOS and the other downsides of a Macbook. Microsoft makes fanless laptops, but those are the Surface Pros, with their own set of tradeoffs. The Surface Laptops do have fans. For me personally, whether a laptop has a fan or not is pretty immaterial. So like I said, it's a collection of tradeoffs and personal preferences.
And please note the original quote that I responded to. They referenced all Apple laptops with the M1 and later chips.
Same. In comparison, my last job I joined right when the last of the Intel Macs were going away, so my work machine was a fully loaded 16” i9 MBP. Aside from being enormous, and having the battery life of a gaming laptop, it would get absurdly hot unless I disabled Turbo. While the heat wasn’t a physical problem when docked, it would spin the fans up, and sometimes I don’t want headphones on.
I don’t know who thought shoving an i9 into a laptop designed to be sleek and quiet was a good idea, but they were mistaken.
Oh, also, my base (8 GB RAM, 256 GB SSD) M1 Air destroyed the i9 in actual workloads.
Isn't that by design? I mean, it's their target market.
People who buy air are not supposed to run heavy applications; they are mostly folks in suits running spreadsheets or presentations; at least, that's my observation.
Also for people who travel a lot (obviously there's some overlap there). You really start to notice the difference around the third or fourth airport TSA line. :-)
I've been buying MBAs since they first came out because I want my portable computer to be portable and am willing to give up a certain amount in furtherance of that goal.
At home/office you can plug it into a full-sized monitor and keyboard (if you want) without losing the portability factor when you're on the road.
the 15" MBA really needs the M3 Pro chip so it can support multiple displays. That's a kink even with "suits clicking on spreadsheets", it is definitely true that everyone is used to at least dual external monitors by now. Let alone actual dev work.
C'mon guys, you sized the M3 Pro down and everything, I know it'll fit. Yeah, it'll throttle under heavy load but that's what MBAs do. The 15" chassis is overkill for the M3 base-tier and it'll be fine, and that's the buy-in for multiple displays. M3 is a great tablet/ultraportable but it's still ultimately not an ultrabook chip.
Personally I happened to have ultrawide already so it was not the end of the world using a MBA for a while with my single external screen... but it's a bad limitation and there's no real solution except either putting the 2nd display controller on M3 (probably not going to happen for cost reasons) or putting the M3 Pro in the 15" chassis.
Even if you think it's part of some strategy to price-ladder you into the MBP... surely having another rung on the ladder is part of the game too? ;)
Hopefully we see it with the N3E variants that should be releasing sometime this year...
The amount of people running any kind of workload that stresses M3 this much, are not choosing a Mini or MBP instead, and doing this on their laps, must be infinitesimal…
Normal office workers giving presentations while sharing video in a conference room over HDMI while simultaneously streaming the meeting over Zoom will absolutely push the thermal limits within five minutes.
An M3 won’t even come close to breaking a sweat doing all of those things.
Hell, add in multiple intensive browser tabs, a few large Excel sheets, and VSCode open to a large Rust project with rust-analyzer running and still, the processor will hardly even get hot. I’ve done all of this on an M2 with zero problem whatsoever.
I mean it probably depends on a bunch of details, you probably wanna use a video conferencing tool which uses hardware accelerated video encoding and decoding.
This has been my experience. My 15" M2 Air has never gone off the deep end on these tasks. Cool and quiet during streaming and heavy multitasking.
To be fair, I felt compelled to spend hundreds of dollars extra on 24GB and 1TB to address the swapping and IO constraints that exist at the 8G/256GB price point.
And most people also aren’t running synthetic benchmarks, which is practically the only way you’re going to hit these limits short of parallel compilation of a large project. If that is you, by all means, buy the Pro (or a desktop); as the article states, this is about the Air throttling, not the Pro.
Oh now I see you edited your post. Do you have any reference that this is affecting the pro series? I just got an M3 Max-based MBP for work and everything seems to work great so far.
That's not a "problem", it's by design. As long as cooling extracts enough heat to keep the system running at full speed, it is doing its job well and the heat is the result of that. Until we start using photonic processors or something, there is no way around physics. It's a wonder we can even do that on a tiny laptop in the first place.
My partner is getting one of those computers. I can tell you that for her, the computer will not run anything but Messages, WhatsApp, Spotify and Safari at the same time. It's a perfect computer for most users. I'd recommend Pro models for better display and ports not the compute power. Percentage of users that need lots of power is very small based on my observation
As a satisfied M1 user since nearly launch (my previous laptop was a 2013-era Air, and before that, various Dells), I can confidently state that an Air is also quite capable of doing development work. Perhaps not for game dev, or if you insist on an IDE that needs GiB of RAM just to launch, but for nvim it is more than adequate. Runs Docker (I prefer colima due to the bloat of Docker Desktop) just fine, limited of course to the resources you can devote to the VM it spawns.
Would you recommend the M2 or M3 though? My mom uses Gmail and shopping and the like. The M2 is perfectly sufficient, if not for the 2 year difference in future macOS upgrade support.
I had a 13" M1 MacBook Air. It was an OK machine to move to from my 2012 MacBook Air, but the thing was super fragile.
Let's just say it's been retired while my 2012 is still going strong (my not-so-technically-inclined partner uses it for basic browsing).
I now have a 15" M2 MacBook Air and I have to say: this machine is the laptop of my dreams. It could be one of the best computers I've ever owned.
Now I'm just holding out for a 27" or larger iMac. I have a 24" M1 iMac and while I have gotten used to the screen being smaller, I still really could use the extra real-estate.
I am a web developer and media producer. These machines handle Premiere Pro and Blender just fine for my purposes. I have zero interest in getting the Studio.
> Now I'm just holding out for a 27" or larger iMac.
I don’t think that’s going to happen, unfortunately. It was a great machine with great price/performance ratio. You could even upgrade its RAM without an engineering degree. But Apple seems to think the solution is Studio + external display, for this decade.
Yeah, I'm not holding out much hope, but I'm an AIO guy and I have zero interest in putting a tall Mac Mini on my desk. I ended up adding a 32" TV as a second monitor and it's good enough.
Must be specific to your circumstances ('media producer') since my M1 Air performs brilliant still and will for long time, much better than the 2012 Air it replaced (that still works, but not well enough). A 2012 Air is doing better than an M1 Air? I am sorry, but I do not believe this...
No, sorry, I meant it was so fragile, it no longer exists— I bumped it off my coffee table and the screen shattered (fell 18" to a hardwood floor).
I balked at spending over $700 to replace the screen. It definitely was leaps and bounds more performant than my 2012. I also hated the lack of MagSafe.
I usually baby my computers, but I did accidentally step on my 2012 MacBook Air when I was in college (I had placed it on stairs while doing a photo shoot with some classmates). It was fine.
My 2012 can't receive macOS and Adobe upgrades, but my partner is happily using it for basic browsing and photo editing.
>> Apple continues to employ a fan-less cooling design for 13-inch and 15-inch M3 MacBook Air models, and while that means users will be pleased about the completely silent operation, they will lose out on a ton of performance. <<
Without scientific data I only guess that overwhelming amount of users will not give a damn loosing out on tons of performance in an ultraportable consumer device having thousands of tons of performance already much below the peak performance lost, unused.
Of course, most want to buy a cheaper model when they are unable to use the performance of supercomputers decades ago in browsing or text editing, or just reading. Which is true for decades concerning products aimed at everyday people. The fallacy of performance and other specification battle is meaningless for lots of consumers and chased by specialty users mostly, apart from the attention seeking marketing folks.
Can a moderator please change the title? As it stands, it looks like the M3 MacBook Air is 33% slower than the M2 MacBook Pro, since that's the natural point of comparison when nothing else is specified.
The HN title should be changed to reflect the fact that it's 33% slower than the MacBook Pro.
As someone that has used M1 Macbook Air for a few years, I can confirm that this is absolutely non-issue unless when you compile for an extended duration.
106 comments
[ 6.4 ms ] story [ 166 ms ] threadModders would work around this by putting thermal pads in between the heatsink and the back cover to transfer the heat to the aluminum under-chassis increasing the surface area and thermal mass and getting "free" performance capable of reaching the more expensive Pro models during extensive workloads like rendering.
Provided of course you never want to have the laptop on your lap ever again after that, but if you're someone like me who always uses their laptop on fixed surfaces like beds and tables and never on body parts, and you really really need that extra bit of performance, then it should be fine.
"M2 to M3 Air isn't as significant as it should be" might still be true though. How did you arrive at this conclusion? What benchmarks are you basing your opinion on? What is the actual improvement from M2 to M3 according to the metrics you care about, and what do you think the improvement should have been?
The reviews certainly don't say so.
> The M1 and M2 generally performed the same for light and medium-sized tasks whether they had active cooling fans on them or not, and the M3 acts the same way.
Compared to the M1, a fully enabled M3 with eight CPU cores and 10 GPU cores running some typical benchmarks has roughly 25 to 30 percent faster single-core CPU performance, 35 to 40 percent faster multi-core performance, and between 45 and 65 percent better graphics performance, depending on the test.
The numbers speak for themselves, though—however fast it throttles, a passively cooled M3 runs faster than an M1 or M2, and it stands up well against modern actively cooled laptops with Intel and AMD CPUs inside. But if you regularly stress the CPU and GPU, you're probably leaving a bit of the M3's performance on the table relative to the same chip in the $1,599 MacBook Pro.
https://arstechnica.com/gadgets/2024/03/review-apples-effici...
Linked article is literally a review that says so.
Hacker News's title length limitation makes the title here incredibly misleading, the full title says "33% Slower Than MacBook Pro".
But if you don't spend even $200 more on the Macbook Pro to get 16GB, you are leaving a bit of the M3's performance on the table due to swapping. So, $1799. Or step up to the M3 Pro with 18GB at $1999.
And on the Air you could drop that same improvement in. So, $1,299.
Which says to me that from $1000 to $2000 there are a variety of ways to get satisfactory performance out of a long-battery M3 Macbook.
A corollary is that there is no way to get everything (unlimited multitasking, unthrottled renders, no swapping, no IO constraints) you could want out of a $999 Macbook Air M2/8GB/256GB.
x86 CPUs and GPUs usually have a recommended load temperature of ~85°C and start throttling heavily past ~100°C, with emergency shut-off at ~120°C
I know Apple has always preferred running MacBooks hot and silent instead of cool and loud, but are they using a different solder or PCB for the M-series?
Granted, the chip may be designed for these temperatures, but I would assume that apple does what other manufacturers do and just assume the cases of users where high heat is not intermittent and causes premature failure will be rare enough to be inconsequential.
AMD does likewise, though their CPUs' thermal limit is 95C from what I know; I assume they have a similar warranty period to Intel.
Apple also seems fit to operate their CPUs like that.
Given all that, I'm going to presume that advances in fabbing tech mean CPUs today tolerate high temperatures better than ones from yesteryear.
I guess if they've determined it's not detrimental it's fine, but on a MacBook Air with no fan, I can't imagine you've left much performance on the table by heat-soaking those extra ~20 degrees. The way these chips heat up that's gotta be just a few more seconds at full speed.
This is because the Internal Heat Spreader for 7000 series was made primarily with cooler height compatibility in mind. It’s compatible with all previous (aftermarket) AMD coolers, but because the chips are lower height they had to make the IHS thicker which lowers the velocity of heat transfer.
Before doing that I tried lowering the max temp instead but I think 100+ watts is just too much for the cooler I have and I couldn't tolerate the noise. I'm sure I've lost some performance but I can't tell the difference.
> During the game, the surface temperatures of the laptop certainly got warmer, though never uncomfortably hot. Many gaming laptops I’ve reviewed get much hotter (and louder, of course) during games. Looking at you, Razer Blade.
https://www.digitaltrends.com/computing/macbook-air-m3-gamin...
Edit: midrange being midrange web and productivity, not midrange gaming, midrange CAD, midrange 3D modelling or midrange video production. (but you could always do those for at bit.. or in bursts)
You buy it if you want to get something in the ecosystem, but don't need much, and that is okay. Not everything has to be maximum all the time.
But the MacBook Air with 16g of ram and 500gig hdd costs like 1500$, but is apparently still only „meant“ for light browsing, lacks ports, only comes with a mirror (glossy) screen, and the OS is missing a bunch of basic functionality.
The MacBook seems expensive and full of compromises.
Where do you get that idea? The M1 Air was already more than enough for all my web dev needs. I have lots of docker containers running all the time and I don't notice anything.
I have been researching this lately. At the 24GB config it is looking like the M3 air puts down about the same real world CPU performance as my 2019 16” i9 MacBook Pro. For light docker and development use cases these look really compelling to me.
The amazing thing is that with this iteration, you're really leaving practically all Intels behind (at least Alder Lake and earlier) as you're either just as productive, or more productive, but without the heat, power draw or constant charging.
To pre-empt people who want to qualify "much" as something else, I even included an edit so you don't get people with a different idea on what "much" might mean in this specific context (as it has for over 10 years, none of this is really new). It appears that did not reach you in time.
You specifically probably don't want something in the ecosystem, and you do need much. As such, you should probably not buy this. Other people do exist, and they might buy it if it fits them. It doesn't matter if the numbers on paper tickle you the wrong way.
So which one is it? Is it a device with great performance, or not? It’slike people say „these things have great performance“, then somebody comes along to run some test saying „these machines throttle like crazy if you tax them, they have poor performance“, and then the the first group comes along saying „you expect too much“.
I guess they have great performance, as long as you don’t need it?
This is also why a lot of 'tests' doen't really tell you much on the actual use of the machine. Even if we can profile all users, and bucket their tasks in terms of system load you don't get a perfect testing system to compare machines.
How 'good' something is, is almost a quality factor where the only real thing you can describe is "to what degree did the device do what you expected it to do". Good quality would be "meets expectation", bad quality would be "did not do what was expected" (and you can extrapolate from there). This is also where consumer satisfaction is generally measured, people get satisfied with their stuff if they can do what they wanted to do, not because of numbers on paper or tests on a website. Ironically, this is also where building your own PC with some cool GPU, Ryzen, and a bunch of M.2 SSDs tends to be satisfying, not because those component's numbers enabled your day to be better, but because you built something and you can use it, you feel ownership and accomplishment, and that's a quality you can be satisfied about.
For an Air, the best you could probably do is describe what a user wants ("I would like to create some content, consume some media, play the odd game here and there, do some software engineering, some 2D work") and then after a while ask them how close the machine performed to their expectation. And even that will simply vary from user to user.
Even work fleets have this problem, IT departments love to standardise, but unless the work all employees perform is specified to the degree where it could be completely automated you won't be able to construct a homogeneous fleet that covers everyone's needs. So you'll always end up with cases where it's an expensive solution, a cheap solution, a bad fit, a good fit, and that just gets averaged out on a quarterly report so we can call it 'good enough'.
In neither configuration are they particularly cheap, but you’re also holistically paying for a first-party supported Unix OS with great third party support, a bandwidth level 2-4x greater than x86 laptops currently deliver (with unified memory and a good gpu for the power), very good screens/keyboards/trackpads and an aluminum chassis, a huge amount of thunderbolt with all ports at full spec, incredible battery life, etc. No, apple doesn’t compete dollar for dollar with an atom based Chromebook with a plastic body or whatever, but you also do get a much nicer machine.
This should go without saying (but seldom does) but different people have different value judgements, and different products exist for these different people. If you want to min/max your life with a $300 chromebook that's fine. Other people prefer to look at it in the "I spend 16 waking hours staring at Screens every day, I want it to be a good one". And if you are otherwise in the market for a well-specced x86 laptop, the apple stuff is not that bad, and while it does have its downsides and weaknesses it also has strengths and unique points over x86 too. You're not going to go run a LLM on that 6840U or 12700H. Have fun getting sleep to work right/waking up to a cooked laptop in your bag. Remember bluetooth problems? Me neither, they disappeared when I got issued a macbook (i9 even but the apple silicon models are better).
You see the same thing in GPUs a lot too. People fucking obsess about some numbers that largely come down to 10-15% one way or another, and you know what? If you want the NVIDIA features or (perceived or real) driver stability/herd immunity, it's fine. If you want the AMD VRAM and perf/$ advantage it's fine. It ultimately doesn't really matter all that much.
People get super wrapped into "but this one is 15% better perf/$!!! that's like half a segment!!!" but that means we might be talking about a whole $50 at the register, and if it doesn't do what you want it to, if it's going to nag and chafe at you, then just spend the money. And for some people, especially with larger purchases, there is also the psychology of "I'm spending $2500 on this Thinkpad anyway, I'm keeping it for years, I don't mind spending a little more to check boxes". Especially for what amounts to craftsmen choosing tools - we are going to stare at these machines for a long time, right?
Anyway like, there's zero question in my mind that the apple stuff smokes the latitudes or thinkpads you'd otherwise get. They're great.
Whereas most compelling reason I've seen so far for M3 over M2 is Ars Technica's review yesterday--get the M3 for the extra 2 years of supported macOS updates. Well, except that shouldn't be a compelling reason.
Really need to make sure you do not keep that thing on your lap or the cost of fertily treatments from your baked testicles will greatly exceed the cost of a laptop stand or even a Pro.
Build a Linux box, put it in a closet, ssh into a dev container. It’s way easier than you think.
If I ever get new hardware in my rack, this is absolutely the path I’ll take. I love the battery life and lightness of the Air, and honestly its performance is nearly always adequate for me. If it isn’t, a U worth of Epyc would solve that nicely.
Some people choose laptops over building linux boxes in closets for a reason. I can't take my closet PC with me on a train or a coffee shop.
Then there's the issue with space. If you live in a small European apartment like I do, then I have no closet, so no extra space just for a separate PC out of sight.
Then I would still need a laptop for SSH into it anyway, that closet PC won't replace it.
It's a dumb feature, though. The only benefit is fashion. There is absolutely a market for people wanting a reasonably-priced machine with reasonable (16G+) memory and reasonable performance (which requires a fan) for typical workaday jobs. That's right in the middle of the "boring windows enterprise" segment, and it sells a ton of hardware.
But Apple won't serve that market for segmentation reasons, leading to this weirdly crippled Air[1] that for a lot of geeks is sort of the worst of all worlds. Really you might as well just buy a Chromebook for half the price
[1] Whose only real selling point vs. the putative 16G+fan device we all actually want is that it's so thin you can mount it on a pole as a halberd blade for 1d10 damage.
The devices that are sold in this market are as expensive as MacBooks or sometimes even more expensive. So I don't think this is an issue.
Also if the MacBook Air ran Windows it would be the perfect device for most workloads in this market. Because the workload common there is running a Browser, Outlook, Word, Excel and SAP (or any other ERP and CRM of your choice).
The best part is I can do all of that shit all day long on the battery and the laptop is super easy to move/carry around.
With no modifications or extra effort any thermal throttling I've ever encountered hasn't been noticeable.
When the M1 Air first came out there were the breathless articles about thermal throttling. It's only something that will come up with extended periods of high CPU usage.
I've tried to get my Air to thermally throttle and have only been able to do so doing 3D renders or running local LLMs for long periods (over an hour in my testing). Even long compiles don't really stress the CPU so much it throttles noticeably.
The simple solution in that case was a $20 laptop stand with a fan off Amazon.
If your use case is 100% CPU usage all the time either don't get an Air or get a cheap laptop stand with a built-in fan. If your Air isn't your CI/CD build machine or part of a render farm it'll be just fine.
Except... isn't 114 C a bit much for something that sits in your lap? How about on bed sheets? Could it melt some plastic-y table cover?
These types of safety considerations are normally bread and butter considerations to any electronics manufacturer - there are plenty of guidelines and standards for them to reference and test to.
That said, the safety limits are that - safety limits. They are not necessarily comfortable. I think the safety limits for continuously held metal parts (so a MBA laptop) is like 55C. I think the rough guideline on "comfortable" is usually considered sub 40C.
Here's a review that tested sustained gaming workloads on the M3 Air.
> During the game, the surface temperatures of the laptop certainly got warmer, though never uncomfortably hot. Many gaming laptops I’ve reviewed get much hotter (and louder, of course) during games. Looking at you, Razer Blade.
https://www.digitaltrends.com/computing/macbook-air-m3-gamin...
Well, that's like "never uncomfortably loud" for video card fans. In the eye of the beholder.
Tom's Guide broke out the thermal gun and said the hot spot on the bottom of the 13 inch model was about four degrees F above normal body temps while running Cinebench on a loop. The 15 inch one got about 9 degrees F above body temp.
You certainly wouldn't need to worry about melting your table.
Waste of bytes
It might’ve changed since then but the difference was pretty pronounced with Google Meet a few years ago. In Chrome it roasted laps but in Safari it barely moved the needle.
This is such clickbait it's no wonder why there are people that have a seething hate for Apple. Nothing has competed with Apple laptops since the M1 dropped.
That's not remotely true.
But having a super thin laptop that does 4K Video editing and has an really good battery life doing it, is amazing.
What they can do with 30W from the wall is simply amazing. Not even talking about the Pro/Max Chips that can decode and encode several ProRes Streams at once.
But again, some people don't just edit hi res video.
I want completely silent laptop that doesn’t get dust inside of it.
And please note the original quote that I responded to. They referenced all Apple laptops with the M1 and later chips.
I highly doubt M3 macs are any worse.
I don’t know who thought shoving an i9 into a laptop designed to be sleek and quiet was a good idea, but they were mistaken.
Oh, also, my base (8 GB RAM, 256 GB SSD) M1 Air destroyed the i9 in actual workloads.
So yours are probably MBPros not Airs. Those have a fan.
People who buy air are not supposed to run heavy applications; they are mostly folks in suits running spreadsheets or presentations; at least, that's my observation.
I've been buying MBAs since they first came out because I want my portable computer to be portable and am willing to give up a certain amount in furtherance of that goal.
At home/office you can plug it into a full-sized monitor and keyboard (if you want) without losing the portability factor when you're on the road.
C'mon guys, you sized the M3 Pro down and everything, I know it'll fit. Yeah, it'll throttle under heavy load but that's what MBAs do. The 15" chassis is overkill for the M3 base-tier and it'll be fine, and that's the buy-in for multiple displays. M3 is a great tablet/ultraportable but it's still ultimately not an ultrabook chip.
Personally I happened to have ultrawide already so it was not the end of the world using a MBA for a while with my single external screen... but it's a bad limitation and there's no real solution except either putting the 2nd display controller on M3 (probably not going to happen for cost reasons) or putting the M3 Pro in the 15" chassis.
Even if you think it's part of some strategy to price-ladder you into the MBP... surely having another rung on the ladder is part of the game too? ;)
Hopefully we see it with the N3E variants that should be releasing sometime this year...
Hell, add in multiple intensive browser tabs, a few large Excel sheets, and VSCode open to a large Rust project with rust-analyzer running and still, the processor will hardly even get hot. I’ve done all of this on an M2 with zero problem whatsoever.
To be fair, I felt compelled to spend hundreds of dollars extra on 24GB and 1TB to address the swapping and IO constraints that exist at the 8G/256GB price point.
Oh now I see you edited your post. Do you have any reference that this is affecting the pro series? I just got an M3 Max-based MBP for work and everything seems to work great so far.
Care to give any examples? My work M1 Pro gets moderately hot only when IntelliJ starts indexing multimillion line codebase.
Let's just say it's been retired while my 2012 is still going strong (my not-so-technically-inclined partner uses it for basic browsing).
I now have a 15" M2 MacBook Air and I have to say: this machine is the laptop of my dreams. It could be one of the best computers I've ever owned.
Now I'm just holding out for a 27" or larger iMac. I have a 24" M1 iMac and while I have gotten used to the screen being smaller, I still really could use the extra real-estate.
I am a web developer and media producer. These machines handle Premiere Pro and Blender just fine for my purposes. I have zero interest in getting the Studio.
I don’t think that’s going to happen, unfortunately. It was a great machine with great price/performance ratio. You could even upgrade its RAM without an engineering degree. But Apple seems to think the solution is Studio + external display, for this decade.
I balked at spending over $700 to replace the screen. It definitely was leaps and bounds more performant than my 2012. I also hated the lack of MagSafe.
I usually baby my computers, but I did accidentally step on my 2012 MacBook Air when I was in college (I had placed it on stairs while doing a photo shoot with some classmates). It was fine.
My 2012 can't receive macOS and Adobe upgrades, but my partner is happily using it for basic browsing and photo editing.
Without scientific data I only guess that overwhelming amount of users will not give a damn loosing out on tons of performance in an ultraportable consumer device having thousands of tons of performance already much below the peak performance lost, unused.
Of course, most want to buy a cheaper model when they are unable to use the performance of supercomputers decades ago in browsing or text editing, or just reading. Which is true for decades concerning products aimed at everyday people. The fallacy of performance and other specification battle is meaningless for lots of consumers and chased by specialty users mostly, apart from the attention seeking marketing folks.
For example, putting a metal plate below it while working, would that increase the peak performance noticeably?
Edit: the article refers to adding thermal pads as in https://wccftech.com/m2-macbook-air-overheating-problems-sol...
Still open to other suggestions :)
The HN title should be changed to reflect the fact that it's 33% slower than the MacBook Pro.