81–90 of 135 posts
I think it'll be another month or two at least before doing scientific programming will be pleasant due to the lack of support from e.g. numpy and hacked versions of gfortran being used for R.
Are there any quality lapack/atlas libraries available?
I mostly work in VSCode, the latest insiders build work much faster compared to my 2019 MBP. Browsing is also significantly faster with Chrome. I have tried the Docker beta, it runs well, even with 8gb, but being limited to ARM images limits its use.
You can run intel/amd64 images too by passing '--platform linux/amd64'. I span up Hasura last week and it worked fine - there will be a speed impact but I couldn't notice it for that use case.
I got mine this week, I'm surprised it doesn't even blow hot air while building node from source. However, I am greatly disappointed in the fact that I cannot use both usb-c for displays, only one of them. I did not know of this limitation until I discovered things were not working as expected. Everything is super responsive, it's far better than the 2014 mac mini (4GB of ram) with 2020 SSD. Rosetta 2 seems to work w…
Beyond two monitors, definitely noticing less consistent external display connectivity. Often have to cycle between closing and opening the lid to finally get it to connect. Unsure if that's just a Big Sur regression, however.
Also M1 owners, I would also like to know how much 8/16 GB memory has affected your workflow/productivity assuming you previously had at least 32GB RAM? e.g. Those who never closed their apps before, kept several browsers with several dozen tabs, Note taking apps, IDE, VM etc. all the time. Is lighting fast opening of apps on M1 = Never having to close the apps in any machine with large RAM?
I have laptops with 64 and 32Gb of RAM, but I found my best work happens with 8Gb of RAM. I can't exactly explain how or why, but it seems to enforces discipline by reducing task (and context) switching - so much that I got myself another laptop with 4Gb.
One advantage of having it is that the company I work for has a very international and diverse user base, but gives us developers super modern machines. Some developers tend to overgineer and cause performance issues on slower computers. With an older laptop I can understand better how the app will be perceived by users.
Main thing: it just works, silently and quickly!
No OP but looking for people with any reactions on using Docker, Node or Python on the M1? Likely going to get an M1 mini at some point.
Native Node is fine, and generally runs better than the equivalent MBP. My 8GB M1 handles Node & JS about twice as fast as my 16GB 15" Intel MBP from 2019. Where you can run into issues is architecture support. I'm running Node natively, but this means that some packages that don't support ARM will fail to build properly. I've been able to work around/do without for now, but depending on your use case, it could be an…
Things that rely on disk speed like npm install are much faster, though.
I tried upgrading some projects so they could run arm native, but it some cases it would have forced my colleagues to update their OS since such-and-such updated library dropped support for their OS version.
I briefly went down the path of having a patchwork of arm and x86 stuff running together (like having webpack be arm native but the rest of the project not) and it was lots of fiddling. If I wasn't careful to segregate x86 and arm into completely separate shell environments, I'd end up with ldd trying to link to shared libraries with the wrong arch during [brew, npm, bundle] install. It was more trouble than it was worth.
I think it'll be another month or two at least before doing scientific programming will be pleasant due to the lack of support from e.g. numpy and hacked versions of gfortran being used for R.
Are there any quality lapack/atlas libraries available?
I've been unable to get native openblas yet through homebrew, though havent tried recently.
For deep learning inference I'm using Vulkan/moltenvk/ncnn got it to work and was able to hack around.
Also M1 owners, I would also like to know how much 8/16 GB memory has affected your workflow/productivity assuming you previously had at least 32GB RAM? e.g. Those who never closed their apps before, kept several browsers with several dozen tabs, Note taking apps, IDE, VM etc. all the time. Is lighting fast opening of apps on M1 = Never having to close the apps in any machine with large RAM?
TL;DR: 16GB is enough for most use cases on Intel Macs, so it should be as good or better on M1 Macs. In my experience with 16GB Intel Macs for years, I have never had memory problems. I'm sure there are some large data migration cases which could make it a problem, but for my typical development (1-3 instances of a JetBrains IDE open, PostgreSQL and MySQL servers running, Docker running, sometimes even VirtualBox ru…
If you don't need it, you don't need it but to say 16 gigs on an M1 is somehow "better" seems silly to me.
I got mine this week, I'm surprised it doesn't even blow hot air while building node from source. However, I am greatly disappointed in the fact that I cannot use both usb-c for displays, only one of them. I did not know of this limitation until I discovered things were not working as expected. Everything is super responsive, it's far better than the 2014 mac mini (4GB of ram) with 2020 SSD. Rosetta 2 seems to work w…
Dual monitor (1hdmi + 1usbc) and small performance improvement (from benchmarks on youtube) is why I went with Mac Mini 8GB (Because 16GB ram didn't make a difference)