Today, I’d like to talk about something that is a bit off for this place: the endless discussions and uncertainty connected to how much computer memory one needs. Whenever someone wants to buy a new computer, one of the first questions they typically ask online is: “How much RAM do I need?” And, especially on Reddit (which, typically, is being Reddit), the answer is almost always “You need more.” Since I’m mostly browsing Mac forums, this question pops up a bit more since Apple tends to price its RAM closer to the price of gold than to other electronic supplies. But, thanks to Sam Altman and his co-conspirators, this question has become a frequent occurrence in the Windows space, too.
Now, this article is not going to be ground-breaking. If you want or need the fastest and latest, go ahead and get a 192 GB set of DDR5 memory for … let me check … ah, the price of an entire M5 Pro MacBook. This article is about guiding you towards a good estimate for how much memory you actually need, not for saying you should “just get more.”
So, today I want to give a layman’s guide to “How much memory do I actually need?” I’ll walk through what memory is, why it’s hard to judge how much you will need, and how you can learn how to estimate how much memory you effectively use day-to-day.
The Basics of RAM
First, some basics. Memory, or RAM (short for “Random Access Memory”), is a piece in every computer that holds data just like your regular computer storage, but with four distinct differences: it is faster, closer to your CPU, has much higher bandwidth, and can look up data quicker.
Let’s compare it a bit with your regular storage that you have in your computer; typically an SSD or an NVMe-drive. Those have gotten huge increases in their capacity over the past decade. When I bought my first SSD, I believe it had 64 GB of capacity. Now 2-4 TB of capacity are quite common. On them, we store all our data – primarily files and applications. And the operating system, of course.
Storage is usually connected to your computer using an SATA cable or, in the case of NVMe drives, dedicated PCIe-lanes. However, both SATA- and NVMe-connections are relatively slow. Even though modern computers read data quite fast, the speed limitations of your storage mean that, if you open a very large application, it can take a few seconds until you see something on screen. The smaller an app is, the faster it typically starts. That’s the core reason why we have RAM: it uses a completely separate set of connections to your CPU and is also physically colocated with your CPU. Therefore, it can read and write data much faster. It works similarly to your storage, but because of its physical position and dedicated connection, it is much faster. You can imagine the difference of connection speed between regular storage and memory like the difference between a school-zone street and a twenty-lane highway.
When you start an application – especially one you rarely use (this is important for later) – it needs to be read from your computer’s storage, and placed into the computer’s memory. That’s why it sometimes takes a while to open a program: It quite literally needs to be moved from your storage to your RAM. Once it’s in there, it actually starts. The same happens with most regular files when you open them: The Word document you just clicked also has to be moved into your system memory before it will be displayed. (Some files can be streamed, but that is a different discussion.)
Once an application or file is in memory, your computer can productively work with it. Don’t get me wrong: theoretically it is possible to work with files directly from a storage device, but in almost all cases (i.e., what you do in your own personal home) that won’t work nearly as well as we would like. That is the fourth difference between storage and memory: Storage is meant to hold large amounts of data for a long period of time. Memory is meant to hold a bit of data momentarily, and make it accessible in an instant. Storage is optimized to read or write large chunks of data (such as files or an entire program) at once. But as your program does its thing, it will frequently jump between many different instructions. That requires — you guessed it — random access. And computer memory is optimized for that random access.
That’s why programs need to be kept in memory: If we didn’t, it would take an eternity for your program to do anything. For example, when I press a key in my editor that I write this blog post in, quite a lot has to happen. The key press needs to be registered, then handed off to the program. That program will then handle the key by, say, adding it to a file buffer, recording a “change” event in my editor that I can undo, and more. All of this means that the program needs to access several parts of its own instructions to do so. If all of that were to happen on your storage, you would feel it, because storage is magnitudes slower and less-optimized for this kind of task than your memory.
Why Your Memory is Always Full
Now with a basic understanding of memory, let’s tackle a common comment we can see on the internet: our memory is always full. When you open the Task Manager (Windows) or Activity monitor (macOS), you’ll likely see that almost all of your available memory is taken. For me, for example, it currently shows that 14 GB of my 16 GB are taken.
Oh no! I need more RAM! But do I, really? I myself personally (!) actually do need more memory, yes. But not because I am running out of it. You see, when you go and buy some storage, let’s say an external SSD, that is meant to hold data persistently. You move some photo backup from your phone onto it and then it stays there. The only thing you can do to make it go away is by actively deleting that data. Also, because of physics™, your storage has a limited lifetime. It’s made up of small memory cells that simply wear out over time. After a few million read- and/or write-accesses, it will just straight up break. Now, the rated lifetime of your storage is not a hard number, and there are ways to destroy it faster or slower. But it is something to consider.
This is why we oftentimes tend to keep some space on our storage free. This both gives us the ability to store more data if needed, and the safety that, if some sectors on our storage die, there is leeway for our computer to use the unused space to avoid the damaged memory cells.
Computer memory does not have this limitation. It does not care how many times you read or write data to or from it. And as such, any byte of unused memory is a waste of energy. You could literally write all the data from your storage into memory many, many times over, and your storage would fail much earlier than your memory.
That is why your computer will try to keep as much of the data from your storage in your memory as possible. Any application that is already in memory does not have to be read from storage first. This increases load times (you have to wait less until the app opens), and reduces wear and tear on your storage. A win-win situation. However, if your laptop’s battery runs out, all the data in your memory will be gone. That’s its Achilles Heel.
How your computer decides what data to keep available in memory differs, but it will likely decide based on how often you use an application. Your browser is likely always in memory, even if you quit the app. And it is probably the first thing your computer loads into memory even before your wallpaper even appears after a reboot.
The more applications your computer can keep in memory even if you don’t use them, the better. These apps will start blazingly fast and allow you to do more in less time. But of course, it will only keep applications in memory if there is space. Any application that you actively use must be in memory because otherwise it would be unusably slow, as mentioned earlier. So when you start apps you rarely use, your computer must first load it into memory. And if there is not enough space, it will start dropping apps that it has thought you might use, but didn’t.
That’s why your memory will in many cases always be (almost) filled to the brim: It is inconsequential, but has many benefits for you. (Nota bene: you may see, especially in new computers or on Linux servers that memory isn’t actually filled. This is usually because you rarely open and close apps on a server. There, the programs the server runs are in memory, as is their data, but that doesn’t change often. But this is different on your personal computer, where you constantly open and close apps.)
Swap, or, When A Lack of Memory Actually Becomes an Issue
Now, with this knowledge at hand, it’s easy to identify a situation where all this neat “keep memory used all the time” can break down. And that is if you keep open so many applications at the same time, and work on so much data, that the amount of data your computer considers to be “important” is so large that it exceeds your available memory. If you then open that one large Excel spreadsheet or that one additional application, you will suddenly start noticing your computer become sluggish.
This is because now you have presented your computer with a challenge: You told it that you need that one big spreadsheet, but there is no more memory marked as “optional.” But because it needs to load it into memory, it has to make a decision. And so it will take a look at all your applications and documents, and identify one that you haven’t touched in a while. It then takes that block of memory, and moves it out of memory and onto your storage. That is known was “swapping,” because it swaps a part of your current memory with a new app or file so that you can continue your work. As soon as you pull that application that the computer just swapped to disk into the foreground, your computer then has to quickly move that data back into memory, and swap it with another block of data. You may be able to notice that when it takes a perceptible longer time to bring an app to the foreground. That’s typically a sign of your computer having to swap memory.
So that’s bad and a clear sign that we need to upgrade our memory, right? Here’s the thing: It might not be. Because how often does this really happen to you? All the time? Probably not. Typically, you can make it easier for your computer to actively close apps you don’t use. That will then mark that block of memory as “optional” and your computer can remove that if necessary. If you immediately restart the same app, it will be fast because it’s still in memory, but if you have opened another one in between, it can always re-load it from your storage. If you want to avoid upgrading your memory, the first step is to simply reduce the amount of memory that your computer deems important.
That is, coincidentally, what Apple refers to as “memory pressure.” If you check the memory tab of the Activity Monitor, you can see it in the bottom of the window. When memory pressure is low, that means that your computer thinks that there should be plenty of memory available for any unexpected move you might make. Then the memory graph in the Activity Monitor will be green. As you fill up your memory with applications and data, it will turn yellow, indicating that your computer still thinks it can fulfill all your requests, but might have to think a bit harder about how to do so. But once your memory pressure reaches red, that tells you that you have so many applications open that your computer simply cannot cope with what you’re doing. In that case, switching applications will almost certainly involve swapping. That is, while the one app you are currently using, is fast and in memory, no other app is, and so switching apps always means you’ll have to wait a moment.
And that’s when we can actually start to think how much memory you may actually need.
A Note on Browsers
Before we do so, however, I have to say a word about your browser. Until now, I have just talked about applications and data. But browsers are something special. The reason is that, nowadays, most websites are less a website and more full-blown software packages. And that means that your browser works more like its own mini-operating system inside your actual operating system.
Right now, I have two tabs open that take up more memory than the actual, main, Firefox process. That is because these are tabs that run an entire web application. Have you ever wondered why you can edit spreadsheets both with Excel on your computer and in Google Drive at the same time? Well, you certainly can, but it’s important to realize that Google Drive will consume just as much memory and processing power as Excel. The only reason is that this application will only be downloaded on demand when you open the website, rather than when you install the app.
That’s why browsers nowadays typically come with their own little task manager. Effectively, they have to reproduce the memory management of your operating system. Sometimes, you may see that clicking on a tab takes a noticeable amount of time until it actually opens. That is because your browser, likewise, takes a look at all your tabs and starts to unload unused tabs while you’re not looking. Then, when you focus the tab again, the browser has to load it once more. That’s the reason web browsers are typically the biggest memory consumers. Try to remember: Most tabs you have open are less boring websites, and more like all the apps you additionally keep running in the background.
How Much Memory do You Really Need?
Now we can finally talk about ways to determine how much memory you need. Of course, the simple answer is that “more is always better.” But we’re not here to discuss the maximum. Instead, we’re searching for the minimum. To figure out how much memory you actually require, there are two important numbers to consider. First, how many apps do you typically use at the same time? Their size is a good starting point for figuring out the amount of “mandatory” memory your computer needs. This includes figuring out the amount of browser tabs you typically need. When you have those identified, you’ll need to understand how much data you are working with. Every spreadsheet you open in Excel will count towards Excel’s memory usage. This is not a clear 1:1-relationship because every document has some overhead in terms of memory usage. So while the file sizes will give you an estimate, it’s not going to be exact.
Armed with that information, the minimum amount of required memory should be the memory consumption of your largest app with the largest amount of data that you can find. Because that needs to fit entirely into your memory. If it doesn’t, parts of it will be swapped back and forth, and that will make your experience worse.
Then, you can start adding more apps and data, because that number only tells you how much memory you need to just run that one app. But you oftentimes run more than that. So start adding the memory footprint of other apps you use very often. For me, this would be Firefox, Zettlr, and my mail program. Most other apps I typically keep closed until I need them. And I tend to keep the amount of tabs in Firefox small, so I don’t differentiate between open tabs. But I also don’t regularly use big web-apps. I try to keep actual work out of the browser. Your use-case may differ, so take a second look at how much your browser actually consumes.
Once you have a set of your “most common apps” and “most-used data,” you’ll have a good understanding of how much memory will be sufficient for you. And I’d argue that for most people, 8 GB of memory will be plenty. If all you actually have to use for work are two programs at any single time, but you always run out of system memory, try not keeping every program installed on your computer open at all times ;). Quitting programs from time to time is a good habit to foster.
When you Actually Need More Memory
Now, that gives a good way to estimate the amount of memory for the average person. But there are indeed groups that need more memory, and here I want to shed some light on why I believe to belong to one of these groups, and how to identify if you belong to one such group, too.
Because while checking which apps you typically use is a good first indicator, it will be a false flag for many people. Let me use myself as the Guinea pig. Sometimes I’m very promiscuous with keeping apps open, but for most of my day, I would actually be quite fine with just 8 GB of system memory. Right now, looking at my memory consumption, the only obscene number is Java, which currently has 5 GB of memory. This primarily means that I have to complain to LanguageTool (my spell checker) that their app is quite hungry, but aside of that the actual “important” memory on my machine right now sits at about ~4-5 GB. Which means, 8 GB gives me plenty of overhead to open a few more apps.
However, the guide I provided above is a red herring in my case. The reason is that multiple times a week I work with large amounts of data. “Big data” as we’ve called it 20 years ago. My datasets are typically about 60–200 GB in size. Does this mean I need to buy the MBP worth of DDR5 RAM? I mean, technically it would make everything faster. But no, typically it is possible to stream data, which means that, of these 200 GB of my data, I keep only ~10 GB in memory at all times, closing files again if I don’t need them anymore. This requires a bit of dabbling with the memory management, but it’s easy once I write code that actually keeps the memory requirement low. It’s just another constraint to keep in mind.
But in any case, this does add to the minimum RAM requirement. Because there’s a balance to be had between memory efficiency and speed. I could write code that requires 100 MB of memory at all times, but that would be awfully slow, because my computer would have to read and write data from and to storage much more often. So it’s beneficial to keep larger amounts of my data in memory at all times. (Again, remember that your storage is good at reading large chunks of data, but often struggles if you load many tiny portions.)
That’s why the minimum requirements for my memory are at about 16 GB which is exactly the amount of memory I have.
But I mentioned that I wanted more, and the next computer I will get will have at least 64 GB of memory. Why the quadrupling? Well, there is a cheesy reason, and a pragmatic reason. The cheesy reason is that I really want to run bigger LLMs on my computer. An LLM, just like an app, really should be loaded completely into memory at once. So if you want to run a model of 16 GB size, you will really need to have at the very least 24 GB of memory. But, more pragmatically, while I can make every data analysis work with 16 GB of memory, and I have done so for the past six years, it is slowly starting to become a bottleneck. As my analysis skills improve and the data I handle gets more complex, the benefit of having more fast memory available to do more becomes more appealing. Because, at the end of the day, time is money, and being able to more quickly churn through my data will have a huge impact on my ability to deliver research results.
To be absolutely clear: For more than half of my average work week, I won’t need more than 10–14 GB of system memory. But for those days when I actually have to run my entire pipeline to fix a problem in some paper? I absolutely will be thankful for having those 64 GB.
Final Thoughts
This article deliberately gave no fixed numbers, because there are many variables at play. Rather, what I wanted to do is give a how-to guide to determining how much memory you actually need without either falling for the “you always need more!” trap nor accidentally kneecapping oneself. I strongly believe that if you don’t do much fancy stuff on your computer and 8 GB of memory is insufficient, this is something you can fix yourself, and I’m sure you have more apps running than you realistically need. But I also believe that, if you feel like your computer is slow because you have only 16 GB of memory but deal with 200 GB of data, you won’t need 200 GB of memory.
One can say what one will about the memory pricing of Apple, but the 8 GB of the MacBook Neo are more than sufficient for their use-case. Nobody who needs to run data analysis should buy one, but data analysts also aren’t the target audience for the Neo. I feel like a lot of people tend to get anxious with choosing memory because it’s so weird to think about it. It’s not like some storage where the formula is literally “the size of my current data + 20 % overhead.” It is a dynamic number that will always fluctuate. And I do see why that makes it difficult to make a confident decision as to what you might actually go for.
But please, folks, especially on Reddit, stop suggesting some high school students computers with 64 GB of memory. That is a waste of money. Especially in this economy.