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

How much RAM do I need ?

  • 16go

    Votes: 41 47.7%
  • 32go (+480€)

    Votes: 33 38.4%
  • 64go (+960€)

    Votes: 12 14.0%

  • Total voters
    86
Compiling is actually mostly limited by storage speed. Linking (a step in compilation) is most of the time single-threaded. Some parts of the compilation workflow will benefit from more cores, yes, but most of what we wait for when we compile won't benefit greatly. Plus you rarely compile all your components. That only happens if you pull in a significant set of git changes from others that fundamentally changes things. Normally your build chain is intelligent enough to only compile modified modules and relink them, and again, the linker is mostly single-threaded. So you wind up with storage speed being the biggest factor in compile times, by far. At least for day-to-day compiles. Compile the whole Linux kernel with all the modules and it's a different situation.

Not sure what you are compiling, but I can load 16 threads at 100% compiling c++ for 15 minutes. Each cpp file is a separate compile process.
 
Not sure what you are compiling, but I can load 16 threads at 100% compiling c++ for 15 minutes. Each cpp file is a separate compile process.

Compilation itself, yes. But as I said, linking is mostly single-threaded, and you'll in most circumstances not be compiling more than 1 or 2 cpp files, because your build tool will only recompile what has changed. Assuming you're under build management and don't just recompile everything each time.
 
Compilation itself, yes. But as I said, linking is mostly single-threaded, and you'll in most circumstances not be compiling more than 1 or 2 cpp files, because your build tool will only recompile what has changed. Assuming you're under build management and don't just recompile everything each time.

The LLVM linker uses multiple cores by default and poking around in a header file will recompile everything that somehow includes it. But most small projects won't see any benefint, compiling is usually not the time consumer. But if you're working on something the size of Firefox, you will se benefits.
 
The LLVM linker uses multiple cores by default and poking around in a header file will recompile everything that somehow includes it. But most small projects won't see any benefint, compiling is usually not the time consumer. But if you're working on something the size of Firefox, you will se benefits.

Alright, fair enough :). Most of what I’ve worked on is fairly small in scale. My information about linkers may well be out of date, or just not relevant to LLVM. And whilst I have written C++, I mostly write, Java, Swift, Scala, Python, JS and some Kotlin, so I don’t really fiddle with header files all that often. So thanks for including your insight into the debate :)
 
if you have to ask, the base version is for you

Pretty much this. I'm working with some enormous databases, some of them 50GB to 250GB, and I still went with the base model 16GB RAM 512GB SSD, excepting bumping the GPU to the middle ground for a little extra performance there. Maybe should have gone with the 1TB SSD, but I've got tons of good flash drives as well as multiple portable SSDs, feels like a waste not to use them.
 
What's a go?

Go noun (plural goes) informal
1 an attempt or trial at something: have a go atanswering the questions yourself.
2 chiefly British a person's turn to use or do something: I had a go on Nigel's racing bike | come on Tony, it's your go.
• chiefly British used in reference to a single item, action, or spell of activity: he drank a pint in one go | they now cost about fifty quid a go | Chris often covers 400–500 miles at a go.
3 [mass noun] British spirit, animation, or energy: there's no go in me at all these days. • vigorous activity: it's all go around here.
4 dated a state of affairs: this seems a rum sort of go. • an attack of illness: he's had this nasty go of dysentery. 5 North American an enterprise which has been approved: tell them the project is a go.
 
  • Like
Reactions: junkw
I went for 32GB just to give me some headroom for my current usage and future. You're spending thousands don't cheap out for a few hundred :)
This is terrible logic IHMO - and EXACTLY the way Apple wants you to think.

You're still talking about several hundred dollars extra. Cost is not a relative thing and $400 is still $400. Also, consider that in several years, a used Macbook with 32GB RAM is not worth $400 more than an equivalent used model with 16GB.

If your needs are ordinary, 16GB will last. I'm still using 8GB - macOS is much much better at managing RAM than a lot of people realize.
 
I'm assuming the base 16 is more than adequate for my needs since "I have to ask." However, I will be experimenting with Livestream Studio 6 and the RAM requirements are a bit ambiguous. Thoughts? Thanks!
 
I’d suggest 32 GB if you’re hanging on to your hardware for a while. My 16” MBP has 32 GB and I often hover between 16-23 when I have some Safari tabs and an empty Rhino file. With a big Rhino file, I’m swapping a lot.
So my MBP is going back, and I’ll order one with 64 GB.
 
This is terrible logic IHMO - and EXACTLY the way Apple wants you to think.

You're still talking about several hundred dollars extra. Cost is not a relative thing and $400 is still $400. Also, consider that in several years, a used Macbook with 32GB RAM is not worth $400 more than an equivalent used model with 16GB.

If your needs are ordinary, 16GB will last. I'm still using 8GB - macOS is much much better at managing RAM than a lot of people realize.
You're assuming everyone here is willing to go through the hassle of selling their current MacBook Pro, and buying a used one with more RAM, in a few years? This is not a very practical assumption.

Yes, Apple is overcharging for the RAM upgrade. It's arguably not worth $400, with a sizable percentage of that cost being nothing more than Apple greed. That being said, with a shiny new MacBook Pro it's nice to never have to worry about the amount of RAM. I upgraded my iMac to 32 GB three years ago for that reason, and since then I've never have to worry about closing apps or browser tabs, running occasional VMs, etc. I was surprised to see swap regularly being used even on 32 GB, though the memory pressure typically stays low.
 
One other thing to consider about RAM is that while the two base models get discounted, BTO models typically do not. For me, the difference between 16 and 32GB of RAM was not $400 but more like $650, since Adorama was only discounting the base models. I definitely don't need 32GB now, and I'm not spending $650 more just for a vague feeling that my computer is more future-proofed. (This a concept that I question anyway, since by the time the people who don't need 16GB now will need it in the future, the other parts of the computer like the CPU and GPU will be quite out of date.)
 
  • Like
Reactions: LuciusWrong
This is terrible logic IHMO - and EXACTLY the way Apple wants you to think.

You're still talking about several hundred dollars extra. Cost is not a relative thing and $400 is still $400. Also, consider that in several years, a used Macbook with 32GB RAM is not worth $400 more than an equivalent used model with 16GB.

If your needs are ordinary, 16GB will last. I'm still using 8GB - macOS is much much better at managing RAM than a lot of people realize.


Just one caveat- people who buy Mac are looking for something a little better than normal. Buying more than you need comes with the territory. It is the idea and fear of being limited that concerns people.

My 2013 had 16gb, and while the ram usage often says 12+gb, the graph is always green and sits near what must be 5%. So the performance data is a bit misleading.

I thought i would upgrade to 32gb, because every new machine must have more than the next. But after this thread, im realizing i should just do the base 8 core machine. I knew i wanted atleast 1 tb (i have 512), because a larger SSD will theoretically last twice as long with the same usage. But my SSD is going stro g after 6 years, so that is, again, a theoretical issue.

And i will use some of the money i saved on ram for a NAS.... i almost thought i could do a 512gb + NAS, but i think the 1TB is the upgrade that really makes sense.
 
This is terrible logic IHMO - and EXACTLY the way Apple wants you to think.

You're still talking about several hundred dollars extra. Cost is not a relative thing and $400 is still $400. Also, consider that in several years, a used Macbook with 32GB RAM is not worth $400 more than an equivalent used model with 16GB.

If your needs are ordinary, 16GB will last. I'm still using 8GB - macOS is much much better at managing RAM than a lot of people realize.
Right, but you spent $400 more on it. So if you are getting $200 more than a device with 16GB are you really making more money?

It’s the way Apple has—for some time now (under the guise of thinness)—forced us to think: to try to predict our future use and needs, leading to these purchasing dilemmas. Selling and setting up a new computer is a hassle and many people want to keep their machines for 5 years or more. I see posts from people that still have a 17-incher!

macOS may be really good at RAM management, but not all software is; and a free-RAM bottleneck can be quite a limitation if you encounter it. I’m on the fence myself. There’s no question that a year ago I would’ve stuck with 16: I hated the Butterfly keyboard so I wasn’t invested in keeping those models long-term. This 16” has allowed me to reconsider.

I have encountered RAM thresholds a couple of times previously with 16 GB; not often, but I have. It was hard to tell the impact because I couldn’t exactly stick another 16 GB in to find out: my research work varies so individual tasks are not consistent. Considering my 2013 and 2015 MacBook Pros came with 16 GB as standard, I have a hard time not considering the jump to 32 GB at this time on the cusp of 2020.

The way I see it, even if you don’t end up using more than 16 GB very often, a $200 loss after two years mightn’t be so bad compared to the alternative: encountering frequent limitations and regretting it. I’d rather get 32 GB RAM than the i9 personally, given the thermal constraints in the design. The hex-core would still be a big step up from my 2015’s quad. Then again, the idea of an OLED or microLED screen in future is really rather tempting as well. Decisions!
 
Last edited:
You should obviously get the most you can afford. If you've been fine with 16, you only NEED 16.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.