Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

SteveJobsOG

macrumors newbie
Original poster
Feb 11, 2025
20
7
Hey guys , i'm a software engineering student and I need a new macbook pro
I have kept an eye on the m4 pro , but i dont know if i should go with 24gb or 48gb of ram ?

Im doing mainly web dev , i will install some dev apps , productivty apps , many browser tabs

and maybe in the future i will use llm's locally and develop ML
i want this laptop to last for at least 5-6 years for me

thanks guys :)
 
Hey guys , i'm a software engineering student and I need a new macbook pro
I have kept an eye on the m4 pro , but i dont know if i should go with 24gb or 48gb of ram ?

Im doing mainly web dev , i will install some dev apps , productivty apps , many browser tabs

and maybe in the future i will use llm's locally and develop ML
i want this laptop to last for at least 5-6 years for me

thanks guys :)
You might tackle a particular project where 24gb would be more than enough but... do you only have one app open at a time? I am old school with tried and true, go with more RAM if possible to compliment the power of your processor. I would not (myself) hesitate to get 48.

I still smile as I recall, just before Apple upped the min spec of RAM to 16gigs for some devices, telling others that 8 gigs is plenty. It never was plenty but sometimes enough to squeak by.
 
When working with my usual number of Safari tabs open (often more than 20), I'm very glad I've got the 48GB. On my MacBook Air, which has 24GB RAM, I have to be careful which tabs I keep open as you can easily bump into that limit and things start slowing down when it resorts to using the SSD for caching.

IMHO As it's no longer upgradable, I'd get the largest amount of RAM you can as future OS's are unlikely to need less, but more RAM.
 
Last edited:
When working with my usual number of Safari tabs open (often more than 20), I'm very glad I've got the 48GB. On my MacBook Air, which has 24GB RAM, I have to be careful which tabs I keep open as you can easily bump into that limit and things start slowing down when it resorts to using the SSD for caching.

IMHO As it's no longer upgradable, I'd get the largest amount of RAM you can as future OS's are unlikely to need less, but more RAM.
I seriously doubt that a single browser tab will use up 2 GB of memory. As an experiment, I just opened and refreshed 59 browser tabs in memory-hungry Chrome, and it pushed my total "Memory Used" to 14 GB. But in actual use, this isn't what happens; the tabs you haven't recently used become inactivated after a while and memory use plummets.

So with 24 GB of physical RAM... in the first place, you don't have to "be careful" to avoid paging to SSD. It's not a big deal if the system does that. Paging speed is incredibly fast and you probably don't even notice it.
 
I seriously doubt that a single browser tab will use up 2 GB of memory. As an experiment, I just opened and refreshed 59 browser tabs in memory-hungry Chrome, and it pushed my total "Memory Used" to 14 GB. But in actual use, this isn't what happens; the tabs you haven't recently used become inactivated after a while and memory use plummets.

So with 24 GB of physical RAM... in the first place, you don't have to "be careful" to avoid paging to SSD. It's not a big deal if the system does that. Paging speed is incredibly fast and you probably don't even notice it.

Safari doesn't have to use 2GB per tab to force the system to start paging with 24GB RAM, partly because Safari isn't the only application that uses memory (the OS needs memory too, along with apps like MS Outlook and Powerpoint).

Looking at my 48GB machine right now, the top 10 processes are using 17GB RAM, and this only covers one tab I have open in Safari - along with two more in Firefox.

While other tabs don't always use as much memory (typically around 0.4GB to 1GB each), to put this into context, on my 48GB machine, the amount of memory currently in use is 38GB RAM + 11GB Cache.

While you might not notice the delays caused by paging, I certainly do. This isn't hugely surprising when an M4 chip can reach a theoretical 120GBps to 500GBps in raw memory bandwidth and an M.2 drive can hit a max of 0.15% of these speeds - depending on the spec of the machine.

I'm not saying you need 48GB (and the machine will still work with 24GB or even 16GB RAM), but one shouldn't rule it out. I'm glad I have the 48GB, because I notice the difference every day.
 
Last edited:
Safari doesn't have to use 2GB per tab to force the system to start paging with 24GB RAM, partly because Safari isn't the only application that uses memory (the OS needs memory too, along with apps like MS Outlook and Powerpoint).

Looking at my 48GB machine right now, the top 10 processes are using 17GB RAM, and this only covers one tab I have open in Safari - along with two more in Firefox.

While other tabs don't always use as much memory (typically around 0.4GB to 1GB each), to put this into context, on my 48GB machine, the amount of memory currently in use is 38GB RAM + 11GB Cache.

While you might not notice the delays caused by paging, I certainly do. This isn't hugely surprising when an M4 chip can reach a theoretical 120GBps to 500GBps in raw memory bandwidth and an M.2 drive can hit a max of 0.15% of these speeds - depending on the spec of the machine.

I'm not saying you need 48GB (and the machine will still work with 24GB or even 16GB RAM), but one shouldn't rule it out. I'm glad I have the 48GB, because I notice the difference every day.
How exactly do you notice it everyday? Do you have a second, identical machine with 24gb? Just curious.
 
so guys 24gb wont be enough for my needs ?
well, for your dev, 24GB's fine.

But you also said local LLMs.

Take a look at a list of available models.
Here is one: https://ollama.com/library?sort=popular,
just a subset of all available models,
Then see if there are any you want/need to use but cannot on 24GB machine.
For example, gpt-oss 20b is 14GB but gpt-oss 120b is 65GB.
Another, deepseek-r1 14b is 9GB, 32b is 20GB, 70b is 43GB.
Llama3.1 8b fp16 is 16GB, 70b q4km is 43GB, 70b q8 is 75GB.
and so on. This will count in addition to system resources.

I would see if you can ask around and try a friend's 24GB machine for a couple of days to get the feel for what these models provide and at what speed. You cannot do just off the size of the models, as some you can squeeze into working, but slowly, and others you think will fit just fizzle.

fyi I just traded in my M4 studio 36GB for one with more RAM in order to run more local llms. At some point you have to settle for what you really want/need.
 
How exactly do you notice it everyday? Do you have a second, identical machine with 24gb? Just curious.
I'm using a 24GB MacBook Air and a 48GB MacBook Pro.

I got the MacBook Air a month or so before buying the Pro, and I'm glad I have both. They are both great machines, however, one can be better than the other, depending on what I'm doing at that moment in time.

Given the choice again, I'd increase the RAM on the Air, because the stuttering could be irritating when it was my only machine. Maybe I'm more sensitive to delays than others, but I'll stick with 48GB because that works for my use-case.
 
so guys 24gb wont be enough for my needs ?

Only you can answer that. All I can say is I never regret buying more than I think I need, but I have regretted it when buying too little.

Another aspect to this is that, unless you're planning to keep the machine for less than a year, it's quite possible your needs will increase over the time you own the machine. It's just a shame that memory can no longer be upgraded and it's subject to 'Apple Tax'.
 
maybe in the future i will use llm's locally and develop ML
i want this laptop to last for at least 5-6 years for me

For your (initial) criteria, I would say "Yes!"

To the quoted, you're going to eventually find that you are basically chasing your tail...

...the basic consumer technology available now will inevitably be eclipsed by that which we will enjoy in 2030.

Think "Two Years" before you invest in six.
 
  • Like
Reactions: chown33
It's nice to have. 16 GB is too low for me. I plan on getting way more memory when I upgrade.
 
well, for your dev, 24GB's fine.

But you also said local LLMs.

Take a look at a list of available models.
Here is one: https://ollama.com/library?sort=popular,
just a subset of all available models,
Then see if there are any you want/need to use but cannot on 24GB machine.
For example, gpt-oss 20b is 14GB but gpt-oss 120b is 65GB.
Another, deepseek-r1 14b is 9GB, 32b is 20GB, 70b is 43GB.
Llama3.1 8b fp16 is 16GB, 70b q4km is 43GB, 70b q8 is 75GB.
and so on. This will count in addition to system resources.

I would see if you can ask around and try a friend's 24GB machine for a couple of days to get the feel for what these models provide and at what speed. You cannot do just off the size of the models, as some you can squeeze into working, but slowly, and others you think will fit just fizzle.

fyi I just traded in my M4 studio 36GB for one with more RAM in order to run more local llms. At some point you have to settle for what you really want/need.
I have an M3 Max with 36 GB Ram. Swaps all the time running Ollama. LMStudio is a larger user of RAM as it can load more then 1 model at a time. I typically use 8B parameter models like Llama or Qwen.

I’m waiting for the M5 Max in the spring, and will go with at least 64 GB RAM.
 
It comes down to what you’re doing.

If you don’t know if you’ll need more than 16-24 GB, you most likely don’t.

If you working with specific workflows that do need large amounts of memory, look at those specific workflows to see how much memory they really need.

People don’t need a Range Rover to drive from their suburb to their nearest supermarket, but they might buy in anyway, which is fine, but they’re really not using what they bought - A farmer using a Range Rover will use both the road and off road aspects of it. Equally they might use a normal road car and a tractor, which would not be suitable as a road car.

You question doesn’t have an answer, it’s the specific use cases that will require certain levels of memory, and, even more importantly, how much you are willing to pay to have excess memory as overhead you may never need.

I’d argue that, if you are paying personally, don’t get more memory than you need, and live with some swop Agee years down the road, when upgrading is already to start to appear on the horizon. If you are buying as a business, and the purchase is going into the accounts as capital expenditure or as an expense, then buy more memory as an overhead, as a certain amount of the cost will be tax deductible.
 
so guys 24gb wont be enough for my needs ?
The only RAM-intensive use case you have is developing and running LLM’s, which can use massive amounts of memory. I think you’d know better than us how large your LLM’s might be.

If you’re really on the fence about it, spend the extra money because these machines are non-upgradeable.
 
Hey guys , i'm a software engineering student and I need a new macbook pro
I have kept an eye on the m4 pro , but i dont know if i should go with 24gb or 48gb of ram ?

Im doing mainly web dev , i will install some dev apps , productivty apps , many browser tabs

and maybe in the future i will use llm's locally and develop ML
i want this laptop to last for at least 5-6 years for me

thanks guys :)
5-6 years, max out today's RAM for sure. Anyone who suggests otherwise is flat wrong. The usual wrong-headed advice relates to something you do today, while you only use a new box tomorrow - - and 5 years from now.
 
Last edited:
so guys 24gb wont be enough for my needs ?
Are you intentionally trolling? "enough for my needs" is something only you can determine. The Mac OS will make 24 GB work. However at some point in your described life cycle it will be sub-optimal. How much sub-optimal will depend on what you do, especially how you multi-task, and on how smooth you expect operation to be. You mentioned LLMs, which might quickly make even 48 GB inadequate.

Historically RAM demands have always increased. Always. That trend is more likely to exacerbate than it is to reduce.

Personally I buy loaded MBPs and always max out RAM (currently an M2 MBP with 96 GB RAM). I then typically feel the effects of needing more RAM in ~5-6 years, requiring me to buy a new box. YMMV.
 
Last edited:
two answers (1) having 48 gb is neither good nor bad nor is extended lifetime much of a consideration if (2) the target audience has a lower memory requirement cpu/gpu dependency - always remember the customer base. be prepared to embrace a new development system quickly without remorse (don't get overly driven by depreciation schedules - an occasional lose due to resell price being lower than depreciated value is to be expected) good luck with your cost benefit analysis. my best guess is 48 gb is generally a good idea, but have a general target machine for wet testing.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.