5 ms·
This is a sincere question: what do people need with 32 GB of RAM on a dev laptop? I work on an MBA 11" with 8GB and I'm using Parallels to run Visual Studio 2
by rufius 10y ago
This is a sincere question: what do people need with 32 GB of RAM on a dev laptop?
I work on an MBA 11" with 8GB and I'm using Parallels to run Visual Studio 2015 for a work related project (100k+ C#/C++ codebase). Other than that work I do some Haskell and C++ work locally. It chugs along just fine and I don't feel like I'm hurting with it compared to compiling on my dev box at work.
Are people just spinning up a lot of VMs? I've got a beefy box at work for when I need to do heavy lifting because I value battery life on the laptop over doing it locally. Additionally, I like the static context - work on code at work ok beefy workstation then go home and remote in if I still have more to do.
TL;DR: what are people's use cases for 32gb on a laptop if you're a developer? Why not remote into big box?
- sigzero 10y agoEvery time I've seen someone ask for more RAM it has been because they are spinning up multiple VMs.
- monkmartinez 10y agoOr like 12 chrome tabs....
- rufius 10y agoYa - that was my assumption though I can't envision spinning up more than one a laptop at a time. That said, I don't do client work so my context is relatively static.
- strictnein 10y agoYes, VMs, used in a wide variety of ways. And being reliant on connectivity to a "big box" to simply get your work done is suboptimal. Also, how big of box do you need when you've got hundreds if not thousands of devs? Much easier (and cheaper, since there's no upkeep) to throw another 16GB of memory inside every developer's system.
- rufius 10y agoWell when I say big box, I mean big box per dev but I get your point. The message I'm getting from folks is that if you do contract/client work, you want everything local for many reasons. In my case, I work for a Large Software Company so my scenario is more constrained.
- lanaius 10y agoTo be fair, being a Professional at a computer covers a lot more than just C#/C++ code. Back when I was doing neuroscience research, I could eat up 32 GB of RAM in a MATLAB simulation easy. Using microscopy data people could eat up 32 GB with images (or videos) in no time flat. There's tons of algorithmic work that needs large enough in-memory data to be useful.
- rufius 10y agoThat's fair and that's why I'm asking the question. I'm curious what others are doing that they feel they need more than 16gb. I'm not contesting people's needs, I'm curious what they are.
- e40 10y agoWhy not remote into big box? <sigh> Because when you demo, being tied to a box on the other side of the planet is asking for trouble. I really can't believe the "why do you need that much RAM?" question still gets asked.
- wattt 10y agoIt suggests you work on one client at a time. I'll often have a few VMs running as I switch between 2-3 clients in a given day. For me 8GB is a minimum.
- Spartan-S63 10y agoWhat type of work do you do exactly? I'm just curious because in my experience, I've never run virtual machines with front ends. As of late, I haven't been running VMs at all and have been using Docker containers on my Mac to emulate the environment I need something to run in.
- lj3 10y agoVideo game developers, video editors and graphic artists are the traditional professions where more ram is always better. But, in the days where nobody pays attention to memory usage, casual use can peg all of your ram and page the living daylights out of your hard drive[0]. I've had chrome eat up 32gb all on its own. No VM required. [0]: https://www.macintouch.com/forums/showthread.php?tid=1046 https://www.macintouch.com/forums/showthread.php?tid=1046 (Spotify is killing SSDs)
- calvinbhai 10y agoMost people need 32gb ram for bragging rights. You don't seem to understand why that's a big deal :)
- TurboHaskal 10y agoNo. They buy insane amounts of RAM they don't need, open up the resource monitor, see the greedy apps taking over 19GB of RAM and scream "see guys!! i told you i needed 32gb!1"
- xemdetia 10y agoI am in spinning up a ton of VM's category as well as the artistic category. With networking options available I can have a complete, closed environment that only disturbs my PC. From the artistic side once you start slinging 4k+ images around on multiple layers RAM vs. no RAM is death. What's great though? On my 32GB+ systems I can just do both on the same time and then come back to it later. If you have the ability to leave things spooled up you really miss when you start having to do a hard context switch. It's more realizing the promise of hibernate/suspend options- if you don't have to reboot or stop what you are working on to do something else, why would you? And when it's in a portable laptop you can just take it anywhere, where a big box may be unavailable/down. Not to mention the TCO of a big box versus a laptop in electricity and everyday costs is actually much lower and comes with free battery backup, embedded keyboard and screen! I repurpose many of my old laptops to be servers/development support for that reason and I want my systems to be expandable, because RAM is what I really want to keep things speedy.
- rufius 10y agoGot it. Thanks!
- amjaeger 10y agoRobotics Engineer(ing student). Also annoyed that only the most expensive pro has a real graphics card. Things that need a lot of RAM: Solidworks, Fusion360 (I can't even run it on 8bg with and an integrated gpu) And using Solidworks remotely is a bad joke. Robot Path planning, motion planning AI Assignments, ROS + RVIZ + Gazebo + actual robot software Anything written in python [EDIT almost forgot] Matlab, VMs Playing music or having a web browser open while any of the above things are running... 16bg would probably be fine for 2 years. But >$2000 is a lot to pay for only 2 years of use. People seem to think that just because its no longer current_year-1 that their desktop app can use previous_version's_RAM+0.5gb There are enough people with enough reasons for having only a laptop. The dell precision m5510 is around same price as mbp (depending on setup) and yet it comes with a real graphics card, and the option for 32 gb of RAM.
- mamcx 10y ago> Why not remote into big box? Internet is not reliable. Pls, we are talking about PRO work. > What do people need with 32 GB of RAM on a dev laptop? Databases. I run Sql Server, PostgreSQL, MySql, Firebird. All at the same time? Yes. Also, the thing about "Big Data" is need as much RAM as possible. --- I could live with a iMac (what I have) or a MacPro (that I can't buy because the dolar in my country double in price!) but the main issue here is the signaling. Apple is like telling "We don't care anymore about you". How will be trust the (maybe next?) upgrades to the rest of the line will be almost good now? And what about the next after this ones?
- rufius 10y agoRegarding internet - are you referring to cafe wifi or your own home internet connection? What exactly is insufficient about internet? I work from home 2-3 days a week so good internet is key - especially since my main machines need to be domain joined so RDP is ideal. Just trying to understand what your scenario is.
- mamcx 10y agoI work from home too. The internet is ok here. But is reliable? Not. Also, the internet somethings get slow just because, I need to reset the modem, or call to get it fixed (so the provider replace it with another (?)) and things work for weeks, until not again. Also the latency make some things slower.
- rufius 10y agoFull transparency - I work at a "Big Company" and I've got gigabit. So when I say internet is good, it's quite good. No latency issues and no hardware issues in the 1.5 years I've had it. Either way - fair enough. You don't like the reliability of your internet so the power of the machine locally is important.
- flurdy 10y ago16Gb is ok for now I am already feeling the constraint and know my next laptop which should last 3+ years will have to have 32 gb to be comfortable with my future usage pattern and apps memory feature creep. My usage is obviously not like everyones: lots of docker container and vagrant vms both with dedicated memory, unhealthy amount of browser tabs, memory hogging intellij (especially with scalaz implicits and having lots of projects open at the same time). And the biggest memory killer: working for clients with very large ecosphere stack of JVM based microservices that you spin most of up locally including dbs. Whenever I pair with devs using 8gb MBAs on these projects they continually apologises for the MBA's slowness and their need to continually shutdown part of microservices stack as they run into 8gb constraints. 16gb is a sweet spot today, but not in 2 or less years. Remoting into private/cloud vms or integrating with staging environments is possible sometimes, but I work on the train for 2 hours a day as well as from home a lot, and in demos, meetings, coffee shops. So relying on dodgy VPNs over fragile connections is painful. So spinning up the whole stack locally has always been a must for me. Again, this is not most people's pattern. ( Though spinning up microservices not just locally is a frustration why I in anger started https://github.com/flurdy/philharmonic https://github.com/flurdy/philharmonic )