JOIN: youtube.com/@azisk/join

I understand now why my M1 Pro 32GB RAM Macbook uses around 12GB of RAM in rest mode. Thanks for the information!

VS Code + X code + docker and 8GB is still not using swap! That's impressive memory management.

Really surprised just how well it handled with 8gb of ram. Windows struggles with that much these days from what I've observed.

I went with M1 Air with 16GB and 500GB. Great little laptop, the only thing lacking is the screen brightness. It’s not great writing outdoors but for the price its a great deal.

Thanks for the detailed comparison! The tables showing the differences really helped. Subscribed!

Just what I needed to know. Thank you for posting this

A memory allocator only "reserves" memory, but this memory is only reserved in the virtual address space of the program. It is only written to real memory after the first write. Therefore your program won't actually use 8 GB, but it allocates it (you can see it in htop for example in the "VIRT" column).

great videos man, very informative and delivered in an organized way.

Thank you for your videos, they have helped me a lot picking out my macbook and getting it much sooner.

I have been waiting for this comparison since M2 came out - no idea why nobody else did it 🤷🏻‍♂ You have a new subscriber✌🏻Thanks!

Excellent video. You’re the first one I’ve seen who did this type of test. Thank you

It would be interesting to compare the SSD's TBW status before and after using all that swap.

I write os kernels. I don't exactly know how macOS handles things, but I'm fairly certain that just allocating memory is going to fill the processes' page tables with pointers to just one (zeroed out) memory page. Usually memory is actually allocated only when you try to write to the memory you asked the OS for.

Always prioritize the ram upgrade cause you cannot add more ram down the line whereas you could always buy more external SSDs.

I randomly stumbled on this and took time to watch until it ended. Thank you so much. I have the 14”M2 pro chip 2023 16gb ram 256 ssd and didn’t feel it was enough not like I actually have to work with so much power but I crave the 32gb 1T. No plans on getting or upgrading soon. But i think I’ll just get the M3 Max 32gb 1T n live with that forever. This machines are exceptional…

🎯 Key Takeaways for quick navigation: 00:00 📊 Intro: RAM vs. SSD for MacBook Air software development. 01:19 💻 Basic Usage: RAM usage comparison, SSD matters with larger files. 02:28 🧠 macOS Memory Management: Efficient RAM use and caching. 04:06 ⚙️ RAM Testing: 16GB outperforms 8GB in Xcode builds. 05:05 🪟 Heavy Workloads: 8GB models struggle with memory pressure. 06:27 🐳 Docker Test: 8GB models face memory issues, 16GB remains stable. 07:48 🔥 Stress Test: Base model handles heavy usage well. 08:56 ⏱️ App Startup: 16GB excels, base model surprises in speed. 10:29 🏁 Final Assessment: 16GB RAM shines for developer tasks. Made with HARPA AI

i was using a thinkpad e14 gen 2 with ryzen 7 4600u 24gb model, it wasn't able to handle my development apps, but after i got m2 air base model, i am amazed how it is able to handle everything that i used on thinkpad very easily with just 8gb of ram. Thanks for your mac testing videos. You got a new subscriber from Nepal

Currently learning to code on a 2012 MacBook Air so…. I am definitely keen to upgrade and weighing options

At last I found this video...this is what I wanted to know...thanks Alex...keep up your super practical videos.

this video helps me a lot with the dilemma. Thanks, Alex

Crazy how it took this long to get this comparison! I remember when these MBAs came out a tonne of reviewers were recommending the 8gb 512gb model and it made literally no sense to me. Glad we've confirmed that the 16gb ram model is the one to choose.

mom wake up, alex ziskind finally uploaded again!

Thanks for this awesome compare! I'd bought a MBP 16" 2023 with 16gb ram and 1tb ssd. You think its fine for an computer sience student and how long i can go with it ? I dont use vms

Interesting video, thanks a lot. But may I ask a question which is out of this topic. Do you have any info about MacBook Air Thunderbolt output resolutions to external monitors or TVs that support a 120hz refresh rate? Everyone says that only the MacBook Pro supports a high refresh rate via Thunderbolt, and on the MacBook Air, this function is purposely limited to 4k 60hz. Is that true? I am not talking about connecting to multiplay monitors at the same time. I mean only one laptop connected to a single 4k 120hz monitor or 4k TV via Thunderbolt with a special Thunderbold 120hz supported cable

The problem I have with swap is it will wear out the NAND in the MacBook faster because it constatly has to write to it.

After watching your video, I felt confident with my decision and I bought m1 macbook air base model for my data science course. Though I just wanted it to run all my applications and some light projects., it works like a breeze. I was a intel i3 5th gen windows user so you can imagine the bump i felt in performance ❤ Thanks, I love your videos. Very informative.

Super useful video!! Nice work

Heh, I was waiting for the compressed memory to show up in this video and I like that you saw it in action. macOS is actually impressive in this regard tbh since yeah, even 8GB models can perform significantly well even though we've recommended 16GB as the default for Windows setups for years now. Sidenote: I know this was a test featuring real world dev stuff but I honestly would've given the machines a Code LLM or even just an LLM with lots of parameters to truly see if you could push the memory pressure to red.

Great stuff! This is an antidote to those swayed by a particular channel with an agenda on the 256GB SSDs. I was a pain in those comment sections and was blocked. I insisted that you upgrade the RAM, not the SSD, against the claims of those videos. Funny enough, that channel never did the test you posted despite my insistence and persistence (before I was cancelled). I wonder why? Buy as much RAM as you can afford, and do not worry about filling the RAM, as Apple uses ML to manage it. Currently, on my 24GB MB Air M2, I have 7.10GB cached, and I only have five (in total) Safari and Firefox tabs open (an aside, do not use Chrome, as it is not only a RAM eater, but it chews through battery more than Safari). I use an external drive for storage; I use under 40GB on the internal drive).

This test definitely pushed these computers well past even a heavy workload 99% of MacBook users. For anyone on a budget, such as a university student, considering the baseline 8/256 M1 or M2 MacBook Air, I think this shows that you can confidently buy that baseline model and not have to worry. And don't fall for the "future-proof" argument, either. If you take care of it, that baseline MBA will give you 5+ years of great service. No reason to blow up your budget to fix a problem you almost certainly will never face.

I'm sorry for out off topic question.) What kind of microphone(s) do you use?

The zig test didn’t do anything because you didn’t assign to the memory. Only if you assign to it will the virtual memory pages be marked “dirty” and be resident in memory otherwise they are merely reserved. But good video. 😊

Welcome back we really did miss you

I used 8 GB with Intel processor and 16 GB with M1 MBP. As an iOS develop, especially with SwiftUI, 16 GB are barely enough. I will sure go for 24 GB or 32 in the future.

The MacBook came with 8Gb and 256Gb as base configuration in 2015. It’s crazy that Apple still has that as the RAM and SSD base configuration in MacBooks.

All modern OS use virtual memory. You can allocate as much memory as you want, until you actually write to each memory page nothing will happen. That's why when you use zig to allocate memory it didn't work. Try write random data to that memory and it will quickly blow up.

I'm considering an upgrade from my old trusty Macbook Pro from 2018. Its battery health ain't what it used to anymore and the screen seems to flicker from time to time. As a Java/C# developer (still a student), would it be better for me to go for the all new Macbook Pro M3/M3 Pro or would the the Macbook Air M2 be sufficient?

Nice comparison, I did both upgrades (16GB RAM & 512GB SSD) "just in case.."

Thanks fully I bought 16GB and 512GB model. Very excited to collect from the store in the next few days.

Great video looking at important real-world differences. However, to be clear, neither the RAM nor SSDs in current macbooks are user upgradeable. Make sure you get what you will need for the lifetime of the machine when you make your initial purchase.

I long for the days when RAM could be increased after the initial purchase. These days, I would go for as much RAM as needed upfront. No second chances except on a new MacBook purchase.

I am looking to get a mac mini m2 pro and this is exactly what I wanted to know before I buy one. Thanks!!!

Finally found a video I was looking for ❤️

I have a 2021 M1 macbook pro - 16GB/512 GB and my brother has a 16GB/256 GB of same model. Without any other apps running or any paging, i have noticed that Android studio builds considerably faster on the 512 GB model. Maybe a 30-40 percent faster. That really surprises me.

Hello Alex, to make upscalling video AI in 1080p or 4k, wich one do you recommand please ?

What do you think about graphic design needs? I think it is the same case right? Should I upgrade the RAM first?

Great compare. What about running a LLM locally? I think the 16 or 24 would be ok on the Air. What is your opinion? I realize the air is not the best solution but it is the one I have access to at the moment. Thanks

I have a m1 MacBook pro 13 inch. Still feels snappy. Any suggestions on ram maintenance? Have the base model from 2020

can you share the docker compose example, i would like to run it on my machine and learn how things are done.,

I've never used the programming language you talk about, but can't you just allocate randomly generated memory? Wouldn't that avoid compression?

This is exactly why I purchased the M3-8gb-256ssd base model which is just about enough for Teams meetings & Office & Xcode work on the go. I can wait those extra processing seconds compared to the 16gb model. (this is not my primary machine though, only for mobility purposes)

Genial video !!! Estaba buscando esta respuesta ahora que han salido los MacBook Air M3 !! Gracias 🎉🎉

Welcome back sir, good vibes!

I got my m1 MacBook Air for $750 a year ago and it handles video editing and multi-tasking like a champ. I make 10-15 mins videos on Final Cut Pro and I've never had an issue. 16gb ram would only be a convince thing for me, incase I want 50 tabs open while I edit. I think for 99% of people, 8gb is enough. I can't justify buying the new MacBooks for over $1,300- $2500. The m1 air can do everything I need. Just my two cents. Also great video.

i''ve macbook pro m2 13" 8/512, and my productivity usually take 150-600mb swap memory, the question is, it's possible for my macbook ssd still normal till 5 year later?

I found on my 14inch m2 pro 16gig to show the famous beach ball, when running a about 10 chrome tabs, docker, photoshop and adobe illustrator. For me, illustrator was the main hog when I have 2 or 3 files with each around 30 art boards that have a lot of vector illustrations. I guess it’s safe to say 32 gig of rams would be a better choice.

Very informative yet impressive. I'm in the market for these devices and deciding between configurations. I'm a professor who used windows and I want to believe the 8 gb SHOULD be sufficient for me (office apps, excessive web browsing, nothing crazy like programming). This video really helped me consider!

Great video! I subscribed!

Either Apple should allow you to upgrade the RAM in your MacBook (using proprietary connectors if they wan't to stay Apple) or provide the memory upgrades for cheaper.

That Zig program, you should fill the allocated memory with random data so it doesn't compress well.

Brother Can i buy this base model for my iPhone & go pro's 4k footage editing perpose? Actually i make Cinematic travel vlog which is 9-10 minutes. Or 2-3 layers timeline only. Not quite longer. Can i consider this base model? Give your opinion 🙏🙏take🖤🖤

Could you test the Microsoft Surface Laptop Studio 2 vs the MacBook Pro m2 pro and m2 max?

Why didn’t you run a test after opening parallels? That was the most important part. 8GB was deep into swap but 16GB was still swap free. You just skipped to both being in swap territory.

Which program does Alex use to monitor RAM usage please?

Depending on what you are doing, I've found that 32GB is a good starting point for a developer, especially if heavily using Docker with k8s.

Hi Alex, I want just to add some personal experience. Love your channel. The topic of virtual memory management and MacOS is so complicated that neither this video nor my additions can fully explain it. I am writing here to provide a few examples and explanations from a developer's point of view, hopefully helping users not to make a mistake at the choice of their Mac. In the following, I often speak of "rule of thump" because as a developer, 10 or 20% optimization can be decisive, but in general the decisive leaps to a higher level of RAM and cores are more connected to the respective use case. In general, you must first understand that MacOS generally works with virtual memory and this works differently than with other systems. The memory is always managed and not only when the the Ram lacks and swap is necessary. In order for to work properly, MacOS Virtual memory requires the a linear address space. I don't want to discuss the differences in memory allocation in detail for 32/64 bit Intel and 32/64 bit ARM, but please keep in mind that efficient management depends on a continuous 64 bit ARM address space on AS. MacOS on AS trades part of the computing time for intelligent memory management. I think it's not difficult to understand that if you use all cores, e.g. in exports and benchmarks, this limits the abilities for memory management. Another important part that is to be understood is that the scheduler retains part of the ram and the computing power in order to keep the system still responsive and stable even under full load. Additional rule of thump here is, that apart from edge cases, the system itself and proportionally for the running apps, retains about 1/10 of the RAM and computing time resources and only invests these if the workload absolutely requires it. In short, the more concurrent apps, the more control threads and the more „save guard memory“ is needed. Other important points to understand are that if a system has more RAM, MacOS is more generous with initial allocation of Ram for apps and processes. An app guaranteed more RAM at startup with a system with larger equipment in order to reduce the probability that later computing time must be used to manage it. Furthermore, the system does not automatically free up memory when an app is closed. The reason is that memory operations are "expensive." It "plans" the release of memory for a time when it "probably least bothers". For the majority of use cases there is not need for manual intervention e.g by the activity monitor. Actually, I wanted to bring a few direct examples from my experience in comparison with e.g. Windows, but the space here is far from enough for a sufficiently detailed look. The great scheme is that in real world you can get much more out of MacOS Unified memory, but not for all scenarios. Instead, a few more „rules of thumb“. If you work with VMs, Docker containers, emulations, etc. that require independent/additional memory allocation within the MacOS virtual memory, you have to add more RAM according to their characteristics. 8 GB Ram, as far as they can carry you under native 64bit apps on MacOS, are certainly not enough for demanding topics in these areas. I have a cluster environment with 5 VMs for customers here and under 64 GB Ram and 12 cores, testing makes no sense at all. My 16 GB M1 Pro MBP is not suitable for this in any way. The other important size is the number of large apps you want to use at the same time. Just think of this 1/10 rule of thump. If every app needs 4 GB for themselves and the files loaded to operate properly, how should this work on an 8 GB system without massive performance drops? As mentioned, of course, the size of files and data sets plays a huge role. With a 16 GB set for AI training, it's game over for my 16 GB MBP M1 Pro. I will have to spend money again when I want to process this sufficiently. (budget wise best currently a 64 GB used M1 Max). So if your files become very large in Photoshop, AE, etc., you have to invest in RAM or just wait accordingly until those are processed by swapping back and forth. If you only do this from time to time, those few minutes of waiting time don’t matter, bit if this is part of your daily workflow, think twice. Although somehow tricky a last rule of thump for 3D applications and games. The more detailed and complex the scenes, the higher not just the computational demand, but also the RAM requirement. 8 GB of RAM is quite sufficient at AAA under AS for optimized games at 720p, 900p and a maximum of 1080p (especially with MetalFX wich can cut Ram demand by upscaling). If you want to have a little more headroom and also want emulated games for 1080p, you should have 16 GB+ and for real 4K assets better 32 GB+. The assets and their storage requirements increase with the details and you not only need more computing power but also more RAM. AS and MacOS memory management take you very far and often make use cases possible that previously required much larger computers, but it does not override physical laws. 4K, high detailed assets, just need a lot of Ram. And as a last point, I want to mention that the speed of the hole system depends on how well the simultaneously running apps are optimized. One single app that consumes too much resources can change this massively. I therefore use specialized, optimized apps for most of my activities such as project management, notes, etc. and only use e.g. Chrome for testing code for the customers. I rarely have more than 5 Tabs open and honestly find tab management beyond 10 tabs annoying and time consuming in my daily workflow while multitasking. But that’s probably just me or the fact that my ancient 13’’ Intel MBP with 16 GB Ram was constantly beach balling when doing web development for customers on the road not being sparingly with tabs. So it depends on your workflow and the apps playing nicely with the virtual memory. It’s a bit sad that in some cases you can only find out after looking at the memory pressure whether you did the right choice. For me it’s not a problem that there is a 8GB base model, which most likely would serve me well for web dev on the go. My problem is currently that I’d need much more for other topics.

If you want to artificially push the memory pressure when using the Zig manual memory allocation, try filling it with random values (as opposed to just allocating the memory and letting it be zero filled). Use a Pseudo-Random Number Generator, and the result will not be easily compressible to fill the memory, and then MacOS won't be able to compress the memory to avoid the memory pressure.

You have to access each page of allocated memory to really use RAM.

Less swap also means potentially longer SSD life.

No regrets after almost a year of using the 16gb one.

can you compare 16gb vs 24gb ram models?

You did Zig memory test wrongly. When you only allocate memory every modern OS will provide you a virtual memory which cost nothing. No real RAM used in such a case. So you need at least write something to this allocated memory. And only after that memory compression/swapping will take a place and try to compress/swap memory pages which not used frequently. So to prevent extremely efficient memory compression you might write random numbers to memory.

how about running a python script that stores data into an array, ex: a=[] for i in range (0,10**12): for j in range(0,10**6): a.append(i**6+j) print(i)

Do you know how you can clear swap without turning off your MacBook? On Linux it was swapoff -a ; swapon -a

which is best for video editing

I got some heavy algorithms that we can use to test the RAM in parallel processing.

I have a M1 MacBook Air for three years now. I thought it over to upgrade and I decided to order a 1TB SSD option. 8GB RAM is fine for me.

We missed you!!!

I enjoyed this video and made my decision based on it. Thanks

Great video!

This is called "Apple's Upgrade Hell Dillema"

Coding with base m1 air modal here. No different than the devs with M2 16gb Pros in the office.

The problem is i use VM often. And it can be cached

Im a random guy on the internet and my advice is to always prioritize true RAM, i cant believe im saying this but 16gb isn't enough for some dev environments, for example my work's Angular/SpringBoot dev Environment. we have a lot of dependencies like Mysql, Kibana, ElasticSearch, Redis ..etc, running on docker containers locally. i had to switch from an M1 Macbook Air 16gb to a thinkpad with 40gb of RAM running linux and the project still takes 25gb of RAM on a lightweight linux desktop. Also Virtualization is getting really good on Apple Silicon, the latest UTM virtualization allows you to virtualize Macos and Linux operating systems with GPU virtualization thanks to the Apple virtualization framework for Macos, Asahi GPU drivers for Linux VMs. and for that 16gb isn't enough asigning 14+ GB of ram to the vm crashes the program. i love Apple silicon but im probably switching back to x86 laptops due to "Affordable" RAM upgrades and Nvidia gpu support for LLMs. Love your content Alex, Glad to see you back

Hey there, really interested in newer (apparently slower) SSDs where I understand the bigger SSDs are faster. So maybe 8GB RAM models with the slower SSDs vs the faster ones

Awesome comparison video! Crazy to see how MacOS is so intelligent.

liked the video before it even hit the 2 second mark. There are hundreds and hundreds of M series Air 'reviews' out there all with the same script or spec reads. all these reviewers just don't want to put in the work. this is actually a good and important topic for any one looking to buy the macbook air.

memory compression is a pretty nice thing

can we get a 16 vs 32 gb ? please , I just love your POV

It was expected to me. But now... Do I really need to choose 24 RAM on MBA M3?

thx for your videos!my M1 Mac mini base model I spend my time klling apps! google even safari take all the memory! sometimes it slows the comp so to keep it working you need to stop everything this is really annoying! damn we never had issues with browers before

Fill the SSD 80% on all of them and the RAM size difference effect will be noticeable

hi when u by mackbook 2023 haw mach cycle if battry?! 0,1,2.....?

The Unified memory feature on MacBooks never fails to amaze. Thanks Alex.

Would you be open to doing a similar video with Windows running instead?

Hi, i am considering the M2 base model, i plan to use it only for google docs, opening 15-20 tabs, and microsoft office, and discord. Is 8 going to be enough for at least the next 5-10 years? Ofc in your video the 8gbs seems to be doing fine, but i was just wondering if opening all those things i mentioned at once would be fine.

How would this be if you used a Linux distro?

Thank you sooo much

I use macOS every day, and Linux on the server every day. macOS is too aggressive in swapping memory pages out to disk, and it's less configurable for swap options. I jumped through lots of hoops to get macOS to use as much RAM available. If I have 64 GiB of RAM why does the OS start swapping at 16 GiB of mem usage? Also, Rust > Zig...