If you are planning to use vibe coded programs then you better plan for a lot of RAM. Well I say planning when its really being thrust upon you in a lot of cases.
But be extremely mindful when looking at anything touched by Anthropic or OpenAI.
These coding agents have terrible feedback cycles for lag and memory leaks. A particularly egregious one is GPU usage which is absurdly high in most of these apps due to terrible design and react render loop issues. Someone needs to make RL enviroments than punish this.
I will caveat this with saying being wasteful with resources is not new at all. But at this point it seems like no one is doing any amount of basic due diligence.
It depends on who’s promoting as well as the stack and the app and the complexity.
To be honest (no pun intended here), I’ve actually benefited from finding leaks and optimizing code paths by such tools. As long as something is measurable and testable you can benefit from AI.
Worth mentioning, I mostly do C++ code at my work.
I run Windows 10 on a laptop with 16GB RAM and on fresh boot it uses ~2.5GB, which leaves a lot of memory for other stuff, but I guess when even the new Notepad uses 20x more RAM that the old one it makes sense to require 32GB RAM.. One can hope that due to the high price of RAM more effort will be put into reducing memory usage in general, because right now it's ridiculous
Had a laptop purchased in December 2024 with 16 GB of RAM. Running a single VS Code instance and Edge with maybe 10 tabs consumed almost 100% of the RAM. Doing any actual development work required me to upgrade to 64 gigs of RAM. Thankfully it required DDR4 SODIMMs and I had two lying around from a recently deceased Mini PC.
Meanwhile the MacBook Neo is amazing with just 8 GB of RAM. I just got it for using as a "thin client" for AI but it can handle at least Godot gamedev without hiccups
If you are planning to use vibe coded programs then you better plan for a lot of RAM. Well I say planning when its really being thrust upon you in a lot of cases.
But be extremely mindful when looking at anything touched by Anthropic or OpenAI.
These coding agents have terrible feedback cycles for lag and memory leaks. A particularly egregious one is GPU usage which is absurdly high in most of these apps due to terrible design and react render loop issues. Someone needs to make RL enviroments than punish this.
I will caveat this with saying being wasteful with resources is not new at all. But at this point it seems like no one is doing any amount of basic due diligence.
I think that is a bold statement.
It depends on who’s promoting as well as the stack and the app and the complexity.
To be honest (no pun intended here), I’ve actually benefited from finding leaks and optimizing code paths by such tools. As long as something is measurable and testable you can benefit from AI. Worth mentioning, I mostly do C++ code at my work.
This recommendation was targeted to gaming, not office work. And even for gaming, MSFT said 16GiB was good enough for most games.
Sensationalist article which doesn’t belong on HN.
Gaming is such a broad category. That said not even modern triple A games recommend 32 gigabyte yet.
I run Windows 10 on a laptop with 16GB RAM and on fresh boot it uses ~2.5GB, which leaves a lot of memory for other stuff, but I guess when even the new Notepad uses 20x more RAM that the old one it makes sense to require 32GB RAM.. One can hope that due to the high price of RAM more effort will be put into reducing memory usage in general, because right now it's ridiculous
I have win 10 iot running on devices with 4gb of ram (they mostly have to run one dedicated app) and it works fine!
Had a laptop purchased in December 2024 with 16 GB of RAM. Running a single VS Code instance and Edge with maybe 10 tabs consumed almost 100% of the RAM. Doing any actual development work required me to upgrade to 64 gigs of RAM. Thankfully it required DDR4 SODIMMs and I had two lying around from a recently deceased Mini PC.
Q: How much of that memory requirement is due to Co-Pilot?
Meanwhile the MacBook Neo is amazing with just 8 GB of RAM. I just got it for using as a "thin client" for AI but it can handle at least Godot gamedev without hiccups
Spyware and bloat ain't cheap these days.
Linux users don't rejoice too much, the Fedoras and Ubuntus were eating so much of my RAM I had to take refuge in Alpine.
"Linux users" is a pretty big spectrum. My XFCE install uses barely any ram.
You might consider MX, it has a much lower footprint for eating ram.
Debian XFCE is still strong with a low amount RAM, too.
CachyOS was using 3.25gb of ram while streaming video at 4k on my system.
My Windows work PC is currently using 28.2 gigs to have outlook, teams and 6 browser tabs open.
Kind of odd an apple to oranges comparison you're making there, streaming video uses hardly any RAM.
How much RAM does CachyOS uses if you're running outlook, teams, and 6 browser tabs open?
Who can afford that much RAM?
I'm going back to the days when 640k was all we could ever use.
Given all the other phenomena going on around computing, I may soon go back offline.
I just got a decent HP laptop for my mother (browsing/emails/light videos) for <$1k, it has 32GB. Good deals are out there.