Tim Bray

My laptops since about 2018 have all had 32G of memory. No combination of Lightroom, video streaming, Jetbrains IDEs, Slack, dozens of browser tabs, and recently generative graphics have ever caused detectible memory pressure.

I’m beginning to think that for the vast majority of people, 32G is enough memory, for the foreseeable future.

Apr 21, 2023 at 12:43amWeb
Show threadThrow AwayApr 21, 2023
@timbray I’m on 24 and it’s _just_ a little bit tight. Not so much that I complain, but …. juuuuuuuust a little tighter than I’d like. Next box will be 48 min.
Show threadJeff AtwoodApr 21, 2023
@timbray I think we'll settle on 64gb over time, so I mostly agree
Show threadTill KrüssApr 21, 2023
@codinghorror @timbray I think the modern SSD speeds also play a factor why even 16GB feels snappy, at least on M1/M2.
Show threadJeff AtwoodApr 21, 2023
@tillkruss @timbray plausible argument, but even a fast SSD is so very much slower than RAM: https://blog.codinghorror.com/the-infinite-space-between-words/
The Infinite Space Between Words

Computer performance is a bit of a shell game. You’re always waiting for one of four things: * Disk * CPU * Memory * Network But which one? How long will you wait? And what will you do while you’re waiting? Did you see the movie “Her”? If not, you should. It’

Coding Horror
Show threadJeff AtwoodApr 21, 2023
@tillkruss @timbray especially since Apple is now embedding the RAM on the same SoC, so it's even closer/faster
Show threadbagigioApr 21, 2023
@tillkruss @codinghorror @timbray i have an m1 with 8 GB and I'm surprisingly ok with it's performances.
Show threadmempkoApr 21, 2023

@codinghorror @timbray

Anyone consider Alan Kay's argument that if you want to build for the future, you should pay to be 10 years ahead? 32G is probably fine now but in 10yr where will we be at?

I'm at 64GB now on my main machine and have a server with 256GB. I run a lot of jobs and 256GB sometimes doesn't seem enough...

I think to be 10yr ahead today, you probably want 1 terabyte of RAM.

Show threadDr. Juande Santander-VelaApr 21, 2023
@timbray in other words, 32GiB are the new 640kiB, got it 😉
Show threadMans RApr 21, 2023
@juandesant @timbray For most people, that's likely true. Personally, I'm making good use of 128G.
Show threadTim BrayApr 21, 2023

@mansr @juandesant

Wow, that's interesting. What are you doing?

Show threadDr. Juande Santander-VelaApr 21, 2023
@timbray @mansr also curious… model training, inference? Or massive simulations?
Show threadMans RApr 21, 2023

@juandesant @timbray Nothing quite so exciting, I'm afraid. Mostly, I'm building system images using OpenEmbedded. Keeping most of the temporary files in RAM really speeds things up.

Then there was the time I accidentally created a 100k square matrix in Matlab. That can hardly be considered a good use of RAM, though.

Show threadDon StewartApr 21, 2023
@timbray wait till there's a local LLM doing ranking, search and generation in each tool. Gonna feel tight
Show threadbrennenApr 21, 2023
@timbray don't give them any ideas
Show threadAlan MartelloApr 21, 2023
@timbray I think it depends Windows or Mac. I run both and the Windows machines seem more memory greedy than the Macs. I’m typing this on a 10 year old MacBook Pro with 16GB Ram running Ventura (by using OCLP) and it’s mostly fast and amazing. Honestly this configuration is as much as most folks would ever need (except code devs, video/audio/image editors, Adobe CC users, etc.) https://dortania.github.io/OpenCore-Legacy-Patcher/
OpenCore Legacy Patcher

Experience macOS just like before

Show threadSlaíneApr 21, 2023
@amart interesting. I also had oclp on my 2013 mbp but it randomly stopped working. Back on macOS 11 for now but thinking of giving it another chance.
Show threadmattApr 22, 2023
@amart @timbray I have to help a relative from time to time use their MacBook Pro with 8 GB ram. It's painfully slow, like launch Photos and wait 30 seconds slow. I've always thought that's wrong but don't know enough Mac to dig in. ... So it's good to read some validation from 'out there' that it doesn't have to be this way.
Show threadCorey QuinnApr 21, 2023
@timbray I think you’re right. I’ve hit weird edge cases but only when I’m doing something horrible.
Show threadSpaceLifeFormApr 21, 2023

@timbray

640K should be enough.

Show threadSuchandraApr 21, 2023
@timbray I've had clion run out of memory with 32g. Had to go to 64 g. It wasn't even too large of an app although I was doing compilation and debugging in a podman container.
Show threadSlaíneApr 21, 2023
@timbray feels like we’re at the top end of the S curve for today’s needs. Once we got SSD and 16GB my use has had plenty of headroom with some beefier workstations for special use scenarios. Latest work machine is 64GB and I suspect will last me a long while
Show threadPaul Murray 🇮🇪 🏳️‍🌈Apr 21, 2023
@timbray I recall someone once saying something similar about 640K
Show threadD PawsonApr 21, 2023
@timbray Many of us thought that of 8... then 16Gb. And so it goes.
Show threadChris MallinsonApr 21, 2023
@timbray 32 is what I get too, but I also have a M2 Air with 16 and I push it hard and can’t tell the difference. It’s so crazy how computing power is just something I never think about any more.
Show threadAlastair CampbellApr 21, 2023
@timbray I wonder if it's a relative thing? I.e. most people have 8 or 16GB, so software gets tested to that. (To the point of mostly working, not perfectly.) If you have 32GB, you're going to hit issues a lot less.
If 32GB becomes typical, time to move to 64...
Show threadsankaraApr 21, 2023
I have 16G. Yet to push that limit.
Show threadChris NicolaApr 21, 2023
@timbray I don’t see why 640k isn’t good enough for everyone?
Show thread joApr 22, 2023

@timbray I’m still on 16G on my MacBook and I rarely push it. Editing Minecraft maps or working with large disk images are the only two examples I can think of though.

That said with the M1 and a 2TB ssd memory seems to be the obvious bottleneck on this laptop.