M4 MacBook Air after a month | developer's machine — Transcript
Full transcript
- 0:00By now, you've probably already seen a
- 0:02bunch of M4 MacBook Air unboxings. It
- 0:04looks almost the same as last year. A 10
- 0:06core M4 chip, up to 32 GB of unified
- 0:10memory, wafer thin body, two lonely USBC
- 0:13ports, and of course, a slick new color.
- 0:16But the real story this year is under
- 0:17the hood. Proper Thunderbolt 4 lanes
- 0:20this time, and a surprise ability to run
- 0:23two external 4K screens, plus the
- 0:26built-in screen. I've used every single
- 0:28Apple Silicon Air machine ever since.
- 0:32Well, Apple Silicon M1. And after a
- 0:34month with the M4, I'm here to show you
- 0:36the real dev results. Docker code
- 0:38builds, local LLM tests, the stuff that
- 0:41matters to software developers. And with
- 0:43Apple giving us more RAM options than
- 0:45actual ports, which config should you
- 0:47get? I got some suggestions as we go
- 0:49through the video.
- 0:52The short version is that the M4 keeps
- 0:54boosting single core power while also
- 0:56speeding up multi-core tasks cuz each
- 0:58core is now faster. So everything's
- 1:00going to be faster. Geekbench single
- 1:02core reflects everyday snappiness,
- 1:03JavaScript applications in your browser.
- 1:05Multi-core measures how fast you can
- 1:08handle actual parallel jobs like Docker
- 1:10images or heavy code compiles. We've
- 1:12come a long way since M1 to M2 was about
- 1:15a 15% jump. M2 to M3 about 17%. And now
- 1:19the M4 sees around 23% gains over the
- 1:22M3. That's pretty good news for your dev
- 1:24environment. This means you can expect
- 1:25slightly shorter compile times, smoother
- 1:27multi-container docker spins, and fast
- 1:30hot reloads. And remember, all these
- 1:31MacBook Airs, the Apple silicon
- 1:33varieties, are fanless. The Intel one
- 1:36had a fan and still has a it's pretty
- 1:37loud, actually. You've probably seen it
- 1:39in my videos. Well, that means that
- 1:40under really heavy marathon loads, you
- 1:43can lose a bit of that top performance.
- 1:45When the machine gets really hot, it
- 1:47starts to throttle. And that means about
- 1:485 to 10% less speed than the peak
- 1:51benchmarks would
- 1:53suggest. All right, let's get practical.
- 1:55Real developer work is build, run,
- 1:57break, rebuild. And I did actually
- 1:59stress this M4 Air for the last month or
- 2:02so with every coding task that I could
- 2:04muster. It's actually been more than a
- 2:05month, but a month looks nicer in a
- 2:07title. I've actually gone through some
- 2:08of these in more detail in some of the
- 2:10videos, but here's the recap. Even
- 2:11though I'm not a C++ developer, I did
- 2:13some C++ tests. Quick sort on 10 million
- 2:16integers took about 2 minutes 34
- 2:17seconds, and that's 26 seconds faster
- 2:19than the M3. Makes sense. Single core,
- 2:22but full core merge sort took about 2
- 2:24minutes 31 seconds, almost a minute
- 2:26faster than the M3. Huge gains there.
- 2:29So, obviously, the more cores you use,
- 2:30the better gains you're going to see.
- 2:32Python matter test that I usually do.
- 2:34It's an interpretive bit of code that
- 2:35uses all the CPU cores as well. Finished
- 2:37in 34 seconds, beating last year's time
- 2:40by 15 seconds. That's crazy. And of
- 2:42course, speedometer 3.0 single core
- 2:44operations there. Chrome hit 49.3, about
- 2:48four points higher than the M3. So,
- 2:50pretty good. Now, that's not the only
- 2:51story here. Since this comes with 16 GB
- 2:54to start and 32 at the maximum, I tested
- 2:57those machines against each other in a
- 2:58video, separate video with more details.
- 3:00But just to give you a taste, this is
- 3:01where more memory really counts. I built
- 3:04a big NX monor repo in 3 minutes and 8
- 3:06seconds on the 16 GB model, but on the
- 3:0932 it dipped to 2 minutes and 12
- 3:11seconds. And I did a huge.net project
- 3:14compilation with 100,000 name spaces and
- 3:17classes. Compilation took 94 seconds on
- 3:19the 16 GB version and 69 seconds on the
- 3:2232 GB version. Pretty big difference
- 3:24there. Finally, Docker Compose spinning
- 3:27up 100 containers took about 20 seconds
- 3:29on the 16 GB machine and 19 on 32. Not
- 3:33that much of a difference in that case.
- 3:35Big difference on local LLMs. Pretty
- 3:37popular topic these days, especially on
- 3:40this channel. You all keep asking me
- 3:41about that all the time. I ran Deep Sea
- 3:43Coder 1.5 billion. Small model, but
- 3:45gives us an idea of what happens. It
- 3:47breezed at 55 tokens per second on the
- 3:4916 GB model and 63 on the
- 3:5332. Last year I showed how an 8 gigabyte
- 3:56machine runs pretty much everything.
- 3:58There's a few things that didn't run
- 4:00very well. I have a video on that, but
- 4:02most of the time you're going to be okay
- 4:03with 8 gigabytes as long as you're not
- 4:05running too many things at the same time
- 4:06obviously. But it's surprising what Mac
- 4:09OS can actually handle in those cases.
- 4:11But 16, you've got a lot more room to
- 4:13move around. But once you push bigger
- 4:15toNet builds, for example, with giant
- 4:17monor repos, the extra RAM shaves off
- 4:19big chunks of time. And of course, the
- 4:2132 GB machine, which I nicknamed the
- 4:23MacBook Air Pro. We're getting more and
- 4:25more Pro every year. Actually, that
- 4:27one's going to give you a lot more room
- 4:28to spread out. You can have a lot more
- 4:30things running at the same time and not
- 4:32feel the effects of things slowing down.
- 4:34For example, in my normal dev day is
- 4:36four terminal panes open running git
- 4:38commands and npm scripts, a couple of
- 4:39instances of VS Code. I have Xcode with
- 4:42an iOS simulator running, and
- 4:43occasionally I run the Android emulator,
- 4:46which is a beast. But this year actually
- 4:48they've done a good job with uh
- 4:50optimizing it a little bit more and it
- 4:52runs a little bit smoother than the
- 4:53previous years. Docker desktop sleeps
- 4:54until I really need it. But Chrome stays
- 4:56open with at least 25 taps. And during
- 4:59this whole time, the OS stays responsive
- 5:01even on a 16 GB machine. App switching
- 5:03is instant except those times when I
- 5:05fire up Android Studio and Docker
- 5:06containers all at the same time. That's
- 5:08when I feel a little bit of a half
- 5:09second hiccup. It's kind of like
- 5:11cramming too many people into a small
- 5:12car. Somebody's going to have to wait
- 5:14for a seat. Come on, let the old lady
- 5:15sit down, you rude person, you. Now, if
- 5:17you're expecting marathon loads or
- 5:19multiple virtual machines, maybe even
- 5:21the MacBook Air is not enough for you if
- 5:23you're doing that kind of stuff. In
- 5:24which case, you're probably going to
- 5:25want to dive into a MacBook Pro. And I
- 5:28actually tested the 32 GB MacBook Air
- 5:30versus the 16 GB MacBook Pro because
- 5:33they were the same price when they came
- 5:34out. Same M4 chip. The Pro, of course,
- 5:36has a few more features like it's a
- 5:39little thicker, more heat is spreading
- 5:41around. It has a fan, so it's gonna suck
- 5:43that air out or
- 5:45blow, whatever you want to call it. So,
- 5:47it's got more thermal headroom. It also
- 5:48has a better screen, of course, with
- 5:50more hertz. But here's a summary on how
- 5:52it did on some tests that I ran. C++
- 5:54quick sort single core. The Pro finished
- 5:56in 2 minutes 33 seconds, while the 32 GB
- 5:59Air took 2 minutes and 36 seconds.
- 6:01Obviously, they have the same cores,
- 6:03same chip, so very close. But maybe that
- 6:05fan and the Pro helped the clock speeds
- 6:07just a little bit. C++ merge sort all
- 6:10the cores all at once. 2 minutes and 18
- 6:12seconds on the Pro versus 2 minutes and
- 6:1326 seconds on the Air. Here we saw the
- 6:16Pro's cooling actually give an edge to a
- 6:18sustained load. That huge net
- 6:19compilation with 100,000 classes, 69
- 6:22seconds on the 32 GB Air and 87 seconds
- 6:25on the 16 GB Pro. Wow, that extra RAM
- 6:28saves you from swapping. So, the Air
- 6:30crushed that test. The Air beat the Pro.
- 6:32The Manroad Python test, 29 seconds on
- 6:35the Pro, 34 on the Air. Again, this is
- 6:37using all the cores and it's pretty
- 6:39intense. So, I think the fan helped out
- 6:42here again. So, if your job is mostly
- 6:43huge CPU sprints, the Pro's fans keeps
- 6:46higher speeds under load. If you need
- 6:48more than 16 GB of memory and do mostly
- 6:50short builds, the 32 GB Air can outrun a
- 6:53Pro on RAM heavy tasks and it stays
- 6:55silent. Okay, real talk. Dropping 2
- 6:58grand on a MacBook up front, that's a
- 7:00whole car repair or a vacation or half a
- 7:03Thunderbolt dock. So, here's a smarter
- 7:05move. upgraded. They let you pick the
- 7:07exact MacBook you want. Custom specs,
- 7:10more RAM, extra storage, and instead of
- 7:12paying everything up front, you spread
- 7:14it out over 36 months. Interest starts
- 7:17from 0% depending on your eligibility.
- 7:19Even better, after 24 months, you just
- 7:21send the MacBook back in the box that
- 7:23they give you, and boom, new MacBook
- 7:26time. No sketchy marketplace listings.
- 7:29No, is this still available messages at
- 7:312 a.m. They work with Apple Premiere
- 7:33Partners. They have a 4.8 in a trust
- 7:35pilot rating and yeah, even folks from
- 7:38OpenAI back then. The best part is you
- 7:40can keep the cash you didn't drop up
- 7:42front earning 4% interest in a savings
- 7:44account. So go to getupgraded.com link
- 7:47in the description. Use code alex40 and
- 7:49get 40 bucks off your next MacBook.
- 7:51That's
- 7:52getupgraded.com. Code
- 7:56alex40. There was a few days when I went
- 7:58on a trip and I spent full work days
- 8:00relying only on the 16 GB M4 Air. Even
- 8:03with VS Code, Xcode, Chrome, Notion,
- 8:06Docker, Compose running, I saw no real
- 8:08lag. At worst, Mac OS swapped out about
- 8:101 GB. That's when it's using the SSD
- 8:12instead of the RAM when everything
- 8:14doesn't fit into RAM. Not a good thing
- 8:15to have happen. So, keep an eye on that
- 8:17if you have a low memory machine and
- 8:19you're using a lot of applications at
- 8:20the same time. But nonetheless, Windows
- 8:22switching felt smooth. Programs pop back
- 8:24in pretty easily. Now, I did notice that
- 8:26after 10-minute full core compile, the
- 8:28palm rest warmed up to around 46°. Not
- 8:31painful, but definitely toasty compared
- 8:33to older models. In fact, coming back to
- 8:35the M4 Air and putting your palms on it,
- 8:37it's always a little bit surprising
- 8:39because the palm rests are warm. Whereas
- 8:41in the previous models, they're kind of
- 8:43cool. So, you expect that. So, that's
- 8:45something new. Not great, but it's new.
- 8:47As far as battery goes, it seems like I
- 8:49could leave this thing on for days and
- 8:51it'll just keep working and not running
- 8:52out of battery. But, I wanted to be a
- 8:54little bit more scientific about it. So,
- 8:57I said a little bit more scientific, not
- 8:58a lot. Okay, just calm down. calm down.
- 9:00I did my developer loop, which I call it
- 9:03now. It's a dev test that I run that
- 9:05simulates what a dev would do, what I
- 9:07would do actually. Not all devs do the
- 9:09same stuff, but it has a bunch of
- 9:11programs open like things that I use,
- 9:13notion. It's got a little bit of music
- 9:15playing. The screen is set to full
- 9:16brightness, to-do list, uh 15 browser
- 9:19tabs. I have a couple of instances of
- 9:21terminal open and VS code and it runs
- 9:23through a loop. A half an hour loop of
- 9:25things that we do like coding, building
- 9:27code, running interpreted code.
- 9:29Basically, things that I covered
- 9:30already, but in a loop. And of course,
- 9:32watching videos is one of the things
- 9:33that we do. I know you do that. I just
- 9:36know it. I I have a feeling. How do I
- 9:37know this? I don't know. You tell me. I
- 9:39actually ran this one twice and it got
- 9:41about the same results on both the uh 32
- 9:44GB version and the 16 GB version. Now,
- 9:47the M4 MacBook Air actually has a
- 9:49slightly bigger battery,
- 9:5153.8 watt hours as opposed to 52.6 from
- 9:55last year, the M3. And look at these
- 9:57two. The M3 MacBook Air and the M4
- 9:59MacBook Air are very close to the number
- 10:03of minutes that they lasted under my
- 10:05test. About 330 minutes for the M3. The
- 10:08M4 lasted just a little bit less. 315
- 10:11maybe. If you interpolate that line
- 10:13down, it's going to be like 328 or
- 10:16something like that, but very close. How
- 10:18does this compare with some of these
- 10:19other machines that I tested? There's
- 10:20the Galaxy Book with the X Elite. The
- 10:23MacBook Pro 14, the M4 version lasted a
- 10:26little bit longer than both of those.
- 10:28It's got a slightly bigger battery.
- 10:29Here's the M4 Pro lasted quite a bit
- 10:32longer. The M2 Max is over here, a
- 10:352-year-old machine that I've been daily
- 10:37driving. So, yeah, the battery is going
- 10:39to go down a little bit. And here's the
- 10:40MacBook Pro 16 M4 Max. That's my current
- 10:43daily driver. Just about the same as the
- 10:45M3 MacBook Air. If there's anything that
- 10:47lasted a significant amount longer,
- 10:49here's a Surface Laptop 7 with the X
- 10:52Elite chip inside of it. That one lasted
- 10:54the longest. But we're all in a
- 10:56territory now where we don't really need
- 10:58to care about charging our laptop
- 11:00throughout the day. If you're working
- 11:02normal days and you're not Elon Musk who
- 11:04works 18our days, then you may be okay.
- 11:06That's what he said. Anyway, I I don't
- 11:08know. Maybe that he said that. Don't
- 11:09quote me on that. So, let's talk about
- 11:11who benefits from this
- 11:14upgrade. The M4 Air is still a paper
- 11:17thin laptop that slips into a backpack
- 11:19easily. With Apple's latest chip, it can
- 11:21handle serious deaf tasks on the go, no
- 11:24problem. So, for students and general
- 11:26developers, if you're in class or
- 11:27meetings all day, coding, note-taking,
- 11:29or Zoom calling, you last 8 to 10 hours
- 11:31on battery, no problem. You also won't
- 11:33distract the room with fan noise. A
- 11:36second benefit, people that are
- 11:37traveling, this is really nice. I've
- 11:39been really enjoying carrying something
- 11:41like this as opposed to the large
- 11:43MacBook Pro, especially when I'm walking
- 11:46around with my backpack. It really makes
- 11:47a big difference. It also will fit on
- 11:49cramped airplane trays. I brought a
- 11:52prop. This is the size of a typical
- 11:55airplane tray. Here's the MacBook Pro
- 11:5716in. Where do you put your mouse? Where
- 11:58do you put your food? Where do you put
- 12:00your coffee? And here is the MacBook
- 12:02Air. M4, M2, M3. Same size. you have a
- 12:05little bit more room to work with. Candy
- 12:07bar goes here. I don't know, your
- 12:08chicken nuggets go here. You can also
- 12:10charge it with a smaller USBC adapter, a
- 12:13power brick that perhaps you're
- 12:14borrowing because you forgot
- 12:16[Music]
- 12:17yours. If you crave more ports, built-in
- 12:20HDMI, or need that 120 Hz display, the
- 12:23MacBook Pro is the one to go with, of
- 12:25course. But I'm going to tell you this.
- 12:26If you're used to 60
- 12:28Hz, don't spoil yourself. Don't try the
- 12:31120 Hz MacBook Pro display. It's going
- 12:33to ruin you. It's going to spoil your
- 12:35day cuz you're going to want it.
- 12:37However, after working on the MacBook
- 12:40Air for that trip that I did for a
- 12:42while, you kind of get used to 60 Hz.
- 12:44Again, there is a little bit of a
- 12:45difference. You notice it, but it's
- 12:47something you get used to, especially if
- 12:49all you're doing is just looking at
- 12:50code. That's not all we're doing, is it?
- 12:52We're watching YouTube, too. Also, the
- 12:54MacBook Pro, the fan gives you better
- 12:56sustained speeds. So, if you got large
- 12:58compilations to do, it's going to be a
- 13:00little faster. If your tasks are small
- 13:02and fit inside two-minute bursts, the
- 13:04Air can match or beat the Pro,
- 13:05especially with that 32 GB version. So,
- 13:08if you are going with the Air, here are
- 13:09some recommended builds for everyday
- 13:11work. 16 GB of RAM, 1 TB SSD, that's
- 13:14fine. Apple charges a lot for those
- 13:17upgrades of the storage, and it just h
- 13:19you can get something like this easy.
- 13:22It's another thing to carry around, I
- 13:24know, but if you want to just throw it
- 13:25in your backpack, and if you already
- 13:26have a backpack, then it's not such a
- 13:28big deal to carry extra stuff with you.
- 13:30But data only, not programs of course.
- 13:32So if you have a lot of programs like
- 13:35Xcode is a honker, right? Android Studio
- 13:37is big, Docker, all those programs are
- 13:39going to add up. 256 GB that I have on
- 13:43the base model here, it's a headache.
- 13:44After Xcode, Android SDKs, Docker
- 13:47images, node modules, you're already
- 13:49almost out of space after installing all
- 13:51that stuff. Forget virtual machines. 512
- 13:54GB works if you move large assets to an
- 13:56external drive, but you'll also be
- 13:58worried about constant space management.
- 14:001 TB right now is pretty good. Gives you
- 14:01enough room to breathe. You can stash
- 14:03multiple virtual machines plus local git
- 14:06clones, maybe a couple of small LLMs.
- 14:08For heavy multitasking, 32 gigs of RAM
- 14:11is a really good option this year,
- 14:13especially on a MacBook Air. It's pretty
- 14:14amazing that Apple added that option.
- 14:16There's also the 24 GB option, and I'm
- 14:19not too sure about that one. I don't
- 14:20know. Yeah, it's there, but if you need
- 14:22RAM, get 32. If you don't need that much
- 14:25RAM, go for 16. 24 is like it's like you
- 14:28can't decide. But if you get more RAM,
- 14:29you're not going to be sorry. If you get
- 14:31less RAM, you might be sorry. So, the M4
- 14:34Air keeps its single core supremacy. If
- 14:36your compiles stay under 2 minutes, it
- 14:38actually rivals a 16 GB MacBook Pro. It
- 14:40runs warm enough to notice, but not too
- 14:42hot enough to harm performance. It will
- 14:44throttle if you go push it too hard. But
- 14:46also, in the winter time, you can think
- 14:47of it as a built-in hand warmer. The
- 14:50thing has silence and stamina. Eight
- 14:52coding hours, still near half the
- 14:54battery, and no fan noise is a beautiful
- 14:56thing to see at the end of the day. So,
- 14:58the sweet spot is 16 GB of RAM, 1 TB of
- 15:01storage. Go 32 if you can. You won't be
- 15:04sorry if you do, and avoid the 256 GB
- 15:07storage tier if you value your sanity.
- 15:10If that profile matches your workflow,
- 15:11the M4 Air might be the most portable
- 15:14dev laptop you can buy. Also, don't
- 15:16forget if you do need more ports, you
- 15:17can always get a little gadget like
- 15:20this. I have a couple of other docks.
- 15:22This one is easy cuz it doesn't have a
- 15:24power plug. It's not Thunderbolt, but it
- 15:26just gives you more stuff, even an HDMI
- 15:29port. I'll leave a link to that one down
- 15:30below. It's pretty nice. I think that's
- 15:32going to wrap it up, folks. Thanks for
- 15:33watching, and I will see you in the next
- 15:35video.
- 15:37[Applause]
- 15:38[Music]
About this transcript
This page contains the full transcript of M4 MacBook Air after a month | developer's machine by Alex Ziskind, generated from the public captions YouTube serves with the video. The transcript has 3,194 words across 444 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.
What you can do with it
Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.
Free YouTube transcript tool
YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.