Hacker News Reader

Comment by mh8h | Hacker News Reader

I wonder what it takes to have more than 16GB of RAM in these "for developers" laptops. I really like the form factor and the fact that it works well with Linux out of the box, but I am sure I will regret buying a laptop with 16GB of soldered memory in 2019.

View on HN

I'd like developer laptops to max out at 2 GB of RAM and have really slow single-core processors. Then I wouldn't be subjected to software written for machines with 64 GB of RAM and the latest high-end Intel processor.

I'm with you, but I'd rather they have 64 or 128 super, super slow cores. That way they'd make better use of threads, and hopefully uncover and fix more race-conditions and other concurrency bugs during development.

This is a silly attitude - it can take 8 GB of RAM to compile an application, which then runs happily in 2 GB.

clusters are cheap and apple charges $200 for 8gb of RAM

By using a remote machine, you gain in upfront cost but you lose dev time due to poor tooling. It's possible to use X11, SSH, SSHFS, or Kubernetes but it never truly replaces the flexibility of developing locally with an IDE. Kubernetes tooling _might_ get there in the future though.

If you have a team at work dedicated to the tooling problem, remote dev can become really nice though.

Company code should be shipped to source control, and all test, buil/compilation handled by automated process that can be reviewed by all.

IMO it doesn’t matter if you’re developing local desktop apps or cloud apps, anything that’s intended for broad distribution.

What’s the point of one extra test run and compilation in a local fiefdom of personal customization?

It’s even sillier for containerized apps where desktop metrics are way different than cloud providers.

You think we should push our code every single time we want to compile our code, and then download the binary?

Never run any tests locally?

Going to be tricky to develop on a flight then, isn't it?

I was about to reply the same. It's 2019 and many applications still take forever to start and are unresponsive as hell. How did we take orders of magnitude improvement in every metric and make it just as bad? The magic of Software Engineering I guess.

These machines are literally supercomputers but they still struggle to perform common basic operations without delay even though much slower computers could manage a tolerable UX 20 years ago.

How? By focusing on developer productivity and not the consumer. We can't get any more anti-consumer in the the tech industry if we tried at this point.

Guessing you've never coded with a robust IDE before...

What exactly does “robust” mean in this context? It’s not a given that a full-featured IDE needs to eat all your system RAM.

Robust as in its set of features is on the level of, e.g., PyCharm and not RStudio.

You can get 32GB of RAM in a 15" form factor, just not in 13".

This is due to the well known engineering limitation that links LCD size to the density of RAM modules that can be supported by a motherboard.

can you explain that engineering limitation? I never heard about that.

Semi-

Autonomous

Remote

Cache

Assignment

Synchronization

Monitoring

It's absurd to me that this "Developer Edition" laptop can be purchased with, at most, half of the 32 GB of RAM that my 6.5 year old ThinkPad has had since day one.

Why do you even need that much? Seems a niche demand when 8GB should just do fine for most.

I've been developing on 8GB Mac for years with no problem. Doubt Linux needs more to do the same.

8GB isn't really enough when you're running a few docker containers like Elastic Search and the like.

I have 8GB on my Mac and I'm constantly slowed down by lack of RAM. Built a Hackintosh with 32GB and it's a world of difference. I don't know how people survive with such little RAM.

I do all my development on a Mac with 8 GB of RAM. It’s fine for my needs; I’m just very picky about keeping Chromium from running in the background on my Mac.

As a developer sometimes I end up running an IDE, sometimes not even of my choosing. IDEs like eclipse are famously resource intensive.

Sometimes for testing a VM or container (or three) are needed for testing.

Sometimes something heavy needs to run local for testing, maybe even one that specifically requires an OS you aren't currently running.

Given how dirt cheap ram is, I think it's really silly for all laptops/desktops to not support at least 32GB ram.

Seems like a selfish short term perspective to limit ram to 16GB, although I do expect that it significantly shortens the useful life of a laptop.

My 16GB of ram starts running out after a few instances of IntelliJ + docker.

Also I can't compile servo and run IntelliJ at the same time which really slows down turn around.

That's not even accounting for Firefox + other programs.

You're just not over-engineering your software hard enough.

Spoken like someone that just uses a text editor to write JS. I used to spend a long time in massive Java codebases where the IDE would need some serious horsepower to index the code. Then I'd be running an application server for testing (inside a VM) and running another VM to for the SOLR index and another VM for the very heavy SAP ERP system. Then I've got Chrome and other apps asking for memory and processor power.

This setup is pretty common for people working on large integration projects.

Yeah, I use IntelliJ, Chrome, Photoshop and a few other smaller apps and sometimes even run Win10 on Vmware but I never see any noticeable slowdowns. I do tend to quit apps that I don't use as I hate cluttering my alt-tab list.

Call me old, but these days everything including people are so ram hungry I cannot believe the days when things were being done with less than a GB ram.

People talk like having more ram feels like a champ but as developers you might as well want to think about how things can be achieved using less ram.

I don't care about using less RAM, why should I? My applications run in our private cloud and sometimes in public clouds. We optimise the parts that need optimising but it's often cheaper to scale the environment, add more SOLR servers etc than it is to jump in to the opcode and figure out the best way to get the JVM to JIT something.

Which thinkpad is that?

W530

Came here to say the same thing. I have an "old" T480 with 32GB of RAM. There is no way I'm buying a new laptop with half the RAM. I just don't get why the memory needle is not moving with manufacturers.

Swappable RAM and storage is a feature. I don't care if it adds 2 oz and makes the laptop thicker (the X220 was both thin an light enough for me). Bring them back, at least on "developer" machines.

I agree, which is why I went with a System76 laptop. So far it's been great and I've been enjoying the 32GB of RAM.

It's because of missing LPDDR4 support. In laptop CPUs it's only available on Ice Lake, which just started shipping. AMD doesn't support it at all.

there are several ice lake 10th gen laptops, but none of the linux ones i found have 32gig. why?

Wouldn't additional RAM chips consume appreciably more power? That is definitely relevant for the laptop form factor.

Also, even among software engineers, the need for more than 16 GB RAM is presumably rare.

Would you care to mention what do you need it for?

VM with SAP, VM with SOLR, VM with application server. Massive Java codebase indexing in an IDE, Chrome and multiple other applications.

This is a standard setup that I've seen hundreds and hundreds of developers use on large enterprise / consulting and systems implementation projects.

It's still very common for many many developers to live in world that doesn't solely consist of a text editor with JS.

this is a non-starter for me. i wont buy a machine with less than 32g of memory, period.

Not to mention a decent-size screen.

Reply on news.ycombinator.com