How much RAM do you ACTUALLY need in your M3 Macbook? [2024]

Which macbook to buy in 2024? How much RAM or unified memeory do you need to work with your macbook without brakes and limitations? In this video, Arthur Winer tries to figure out how many gigabytes of RAM your Apple computer actually needs for web browsing, macbook for games, or creative tasks. Watch until the end so you don't throw your money away.

-----------------------

GEAR THAT I USE ↓

As an Amazon Associate I earn from qualifying purchases.
-----------------------

-----------------------

[Thomas-poex]

As an engineer/software developer I got the 16 inch M3 max MacBook Pro with 36GB of memory. Honestly, I think 36GB is pushing my limits a bit. I am averaging around 30GB of RAM usage in a typical work day. Between software and hardware I have already spent over 4k on the laptop and since my budget was finite being a recent college graduate I am going to try and make it work. Honestly I can’t even imagine trying to make the 8GB version work. That would be a total nightmare and Apple should be ashamed they even sell that for a pro oriented computer 😂.

[gerald]

I read a post from someone who composes symphonies for a living. He maintains that he needs 128 GB of memory for his workflow. Folks involved in data science and machine learning fields also need massive amounts of memory. I read that photographers who process large amounts of raw photos or work with large panoramic photos with many layers could benefit from having 36 gig of memory.

[IOSALive]

Arthur Winer, This is fantastic! I subscribed because I love it!

[berndkiltz]

Been monitoring my Ram Usage on my M1 Ultra Studio for a Year. I am doing Video editing mostly, have a lot of tabs often, and i write music (with samplers, drum plugins etc that use a lot of ram). Guess what... I had 128 Gigs, and most of the time it did not go over 32. Sometimes it spiked, but it NEVER EVER got over 48. So...

[martinb]

Yes, I spent ~$3, 200 for my M3 MacBook with 64GB RAM, 1TB SSD, and a bulletproof Operating System with a useable CLI terminal that will last longer than 3-5 Windows devices used in succession. I don’t have to reboot once or more times a week for software updates.

The M3 is outstanding and the only issue I have had is finding ARM64 Linux ISO images to run under Parallels as a virtual machine. Also, not all software for Mac OS from my Intel MacBook worked on the M3 and I had to upgrade or reinstall the ARM64 version. Why 64GB RAM? Speed and the fact I fun multiple VM’s at the same time. I have hit 49GB utilization on occasion.

My attitude is buying RAM one up from what you need. Someone, somewhere, sometime, somehow will write some code or app (or bloated web page running an AI subroutine) that will require more RAM. In 2012 a MacBook Pro came with 8GB or 16GB RAM (Intel CPU) and that was sufficient. I remember when MS/DOS had a 512/640KB RAM limit. In 2008? Desktops and Notebooks had ~256MB RAM on a high-end Windows notebook. If you are shelling out the bucks for a MacBook (Pro or Air), Don’t stint on the RAM. Hard Drive? I use about 120GB of my 1TB SSD. MacOS is not a disk hog. You want VM’s? more space? Get a good SSD USB-c external drive. I recommend a Beetle 512GB or 1TB on Amazon with a 40Gbps USB-c cable. So far it has the best times for R/W using Backmagic disk testing application. My MacBook will last at least 7 years simply because I added the RAM.

Upside? I have had Windows 11, and ARM64 Linux VM’s working fine. Unlike the Intel 2019, I can run Windows 11 (Visio is the main reason) and 3-4 other Linux VM’s with no appreciable, noticeable performance issues. More storage? I added two Beetle 1TB SSD externals in a RAID1 configuration to keep the VM’s. Additionally, with SoftRaid, I added a 2TB RAID5 set using 4x1TB USB SSDs for archive and backup.
Excellent Video!

[wency]

Great video! You just made my decisions super easy. I really did not understand how memory works on an Apple computers. You broke it down perfectly. Thank you for a very informative video. You just have a new subscriber out of me.

[vancent]

I read that ML programmers may need the extremely high RAM for LLM's

[akin]

For 90% of working people, 16 to 32GB of RAM.

For the small 10% of working adults, 48GB of RAM or more! This includes Machine Learning, daily routine 8k video editors, or extreme engineers.

[PeterRince]

I use a lot of RAM for ram-disk. If you care about SSD and longevity, you use ram-disk for cache, temp data etc. If you have your memory usage under control, you can even turn the whole swapping mechanism off, and then there's no unnecessary swap files created in your system at all.

[rafaeldomenikos]

Just bought my M3 Max 64GB. It’s arriving tomorrow, so excited!!!
I hope the RAM is enough, but I couldn’t afford the 128 for $1000 more.
I do AI so need all the RAM I can get 😂

[NatePeriat]

As a music producer, more RAM equals more virtual instruments available in a session without having to utilizing a server, more audio tracks and more plugins. However, someone who is just recording audio doesn't need to max out the RAM 18gb is workable, 36gb is the sweet spot.. Virtual instruments take up the most RAM in a session and even with 128gb it's totally possible to utilize all of it. Most composers that reach that limit however, do run servers to host their virtual instruments. It all depends on if you want to have all that horsepower on the go.

[mikesmith]

I have the base m1 8GB and primarily use it for music production. 8 gig is enough to handle smaller projects, it the moment I fire up Superior Drummer 3, the swap kicks in and immediately jumps to 4 to 8 gigs

[Ronniezim]

For what I do + comfort that I would be good for a number of years, I went with the 36gb m3max. I initially bought the m3pro with 18gb, but in the week I had it, I was in the yellow on memory pressure a ton..and on my previous machine with 16gb of ram (2019 mbp), I would somewhat regularly get out of memory messages..

[RickWilson-kdzv]

My daily workflow, with various applications open, word, adobe, excel, powerpoint, maps, Pixelmator, VS Code, Teams, Apple Email, safari, Edge, and Parallels running Windows, is typical; I will easily occupy 64 GB. Indeed, I could close each when done and start them up again, but it would slow me down. So I have 128 GB, and I am never slowed down. I tried 16 GB Air, but there was no way. I am a consultant charging $250 per hour, easy math. Can't believe my workflow is that unusual.

[AyeItsJT]

As someone who has been working in music production with only 8 gigs of RAM I rage purchased an M3 Max with 128 gigs of ram and I dont regret a thing

[JayzBeerz]

I got 128GB RAM for my PC for $180.00. I know unified memory is different but the cost is ridiculous.

[WildlifeZambezi]

I’m a software engineering consultant, and I often have to create separate MacBook profiles for for each client, which means multiple docker containers, up to 3 or 4 VSCode windows per profile, multiple terminals, multiple browser windows with multiple tabs per profile. Then I also do some Astrophotography as a hobbie which is very resource intensive. I’m going to be getting the 98GB RAM MacBook for this reason. Just in case you were wondering why/how someone needs that much, there’s a real world example!

[gyozolohonyai]

There is no such thing as too much memory for teaching an LLM model. This type of memory is a huge competitive advantage over all other manufacturers.

[kpetsas]

My M1 Max MacBook Pro regularly uses 36-40Gb of RAM while rendering in Adobe Premiere, and with only a few Chrome tabs open.
Thankfully I got the 64 GB version, without listening to the hype that you need less unified memory than RAM in windows... I guess 32GB could work in a pinch, but it would eventually turn out to be the bottleneck.
With that said, for people using Final Cut or Davinci Resolve it could be slightly better as these applications are a bit less RAM-hungry.
I would suggest that for people doing professional work 48 or 64GB is the sweet spot leaving some headroom for the future, and 32GB is the minimum to get.
It is a pity to spend so much money to get a laptop and then be bottlenecked by RAM...

[Cordovan]

i specced my MBP with 32gb and on any average day i sit between 20 to 22gb of RAM in use

but if i want to run an LLM locally that ALONE will already occupy between 20 and 28gb all by itself (depending on model and quantization), slowing everything else down to an absolute CRAWL

even opening a simple finder window can take up to a minute to load thumbnails or expand complex folder structures on an unbinned MAX chip

my next MBP is going to have 96gb at the LEAST

but yeah, everybody's use case is always unique