Can the Mac mini be used as a development machine? Python, Airflow, SQL? x86 or Apple Silicon?

[Mity]

Is anyone here using their Mac mini as a development machine, either web, mobile or simple data analytics / engineering? What setup do you recommend? Is Python fully supported now Apple Silicon now or are there still workarounds for installation? If there is no native support for M1/M2/M4, is it better to buy a 2018 or older model and upgrade the RAM to 64GB and SSD?

I will mainly be using it for light work in SQL, Tableau, and some projects in Apache Airflow. I will eventually want to get into web dev and mobile dev too. For entry to mid-level development, what do you recommend?

 

[ShadowJamie]

Is anyone here using their Mac mini as a development machine, either web, mobile or simple data analytics / engineering? What setup do you recommend? Is Python fully supported now Apple Silicon now or are there still workarounds for installation? If there is no native support for M1/M2/M4, is it better to buy a 2018 or older model and upgrade the RAM to 64GB and SSD?

I will mainly be using it for light work in SQL, Tableau, and some projects in Apache Airflow. I will eventually want to get into web dev and mobile dev too. For entry to mid-level development, what do you recommend?

You may need to investigate further regarding tool compatibility, but generally speaking, Apple Silicon is the way to go. I would recommend choosing any M series Mac with as much RAM and storage as possible. Apple-designed chips outperform Intel in nearly every task, and this will also help future-proof your device.

 

[casperes1996]

Our build servers at work are Mac minis. We do app development including native iOS android and web apps. We recently dropped our last x86 dependency which was in-house. We now build everything as a universal binary for x86 simulator as well as arm sim and arm device. Even when we had an x86 dependency though Apple silicon was faster at running it through Rosetta than our intel build server was.

 

[bzgnyc2]

Is anyone here using their Mac mini as a development machine, either web, mobile or simple data analytics / engineering? What setup do you recommend? Is Python fully supported now Apple Silicon now or are there still workarounds for installation? If there is no native support for M1/M2/M4, is it better to buy a 2018 or older model and upgrade the RAM to 64GB and SSD?

I will mainly be using it for light work in SQL, Tableau, and some projects in Apache Airflow. I will eventually want to get into web dev and mobile dev too. For entry to mid-level development, what do you recommend?

It's kind of crazy but I've run SQL Server in a Docker + Azure Data Studio + Tableau Desktop + Excel on a Mac Mini 2018 i3 w/8GB of RAM. iTunes running in the background didn't skip a beat. That was with Mojave though and more recent OS have higher overhead. More recent SQL Server instances also weren't reliable until I gave them a little more RAM. Either way I would definitely go for more RAM to run a stack like that especially with larger datasets. Ideally 32 GB for any new system for that type of workload.

As far as processor architecture, an Intel Mac has some advantages as far as binary compatability and flexibility (e.g. dual boot/native Intel VMs as a fallback) while the upcoming Mac Mini M4 will get you a processor that still beats Intel's latest versus the 6 year processor in the Mac Mini 2018 (which weren't Intel's best years either). Plus a higher core count that can substantially improve build times among other things.

As an example of compatability, MS is only releasing docker images for SQL Server as x86_64. Since they only stopped their edge version of SQL Server for ARM recently, I doubt there will be a native ARM/Apple Silicon version in the forseeable future. So if you wnt to run MS SQL Server (as opposed to PostgrSQL, MySQL, DuckDB, etc), you should test that MS SQL Server for Docker runs well via Rosetta on Apple Silicon (I've heard that Docker takes advantage of Rosetta to run x86_64 images but have no direct experience with this).

On the other hand, Docker for Mac is finally native to Apple Silicon so that's nice.

Beyond that most open source builds for either Intel or Apple Silicon. Going forward the latter is likely to be better tested under macOS but building on Intel should work fine until Apple drops it from macOS. Apple's future support for Xcode and building and testing iPhone apps on Apple Silicon is likely to be better as well.

The issue you're going to run into on Apple Silicon is cost. To get to 32GB of RAM, you're either looking at Mac Studio, Mac Mini M2 Pro (w/32GB of RAM costs almost as much as the Studio but the Studio adds 2 more performance cores + almost double the GPU), or a Mac Mini 2018 (I wouldn't get anything older than a Mac Mini 2018) configured as you mentioned. That's going to cost a lot less than either of the first two. Compromising for 16 or 24 GB of RAM brings the price of the base Mac Mini M2 down quite a bit so could be a price/technology sweet spot for you depending on whether that's practical for your datasets, etc. The upcoming M4 Mac Mini may or may not change the equation as well.

 

[960design]

Is anyone here using their Mac mini as a development machine, either web, mobile or simple data analytics / engineering? What setup do you recommend? Is Python fully supported now Apple Silicon now or are there still workarounds for installation? If there is no native support for M1/M2/M4, is it better to buy a 2018 or older model and upgrade the RAM to 64GB and SSD?

I will mainly be using it for light work in SQL, Tableau, and some projects in Apache Airflow. I will eventually want to get into web dev and mobile dev too. For entry to mid-level development, what do you recommend?

I use Apple Silicon Mac Mini's (several stacked), an iMac 24", and a MacBook Pro for development with no issues. Currently using satyrn (https://satyrn.app/) for quick python testing, Anaconda, Miniconda and PyCharm for Python work. I've used them to run LLMs, Computer Vision, AI bots, EW command/control, etc. Most of my work today is nodejs / react stacks for prototyping PWAs.

Definitely NOT better to purchase 6+ year old tech. If you are just getting into web dev start easy with a 15" MacBook Air, M3 (~$1200 - $1500US) or a Mac Mini M2(~$500). Good luck!

 

[Mity]

...Apple-designed chips outperform Intel in nearly every task, and this will also help future-proof your device.

I have an M1 mini but the issue for me has been RAM. I'm used to 64GB.

...Even when we had an x86 dependency though Apple silicon was faster at running it through Rosetta than our intel build server was.

That is pretty cool, actually.

...Ideally 32 GB for any new system for that type of workload.

...

As an example of compatability, MS is only releasing docker images for SQL Server as x86_64. Since they only stopped their edge version of SQL Server for ARM recently, I doubt there will be a native ARM/Apple Silicon version in the forseeable future. So if you wnt to run MS SQL Server (as opposed to PostgrSQL, MySQL, DuckDB, etc), you should test that MS SQL Server for Docker runs well via Rosetta on Apple Silicon (I've heard that Docker takes advantage of Rosetta to run x86_64 images but have no direct experience with this).

On the other hand, Docker for Mac is finally native to Apple Silicon so that's nice.

...

The issue you're going to run into on Apple Silicon is cost. To get to 32GB of RAM, you're either looking at Mac Studio, Mac Mini M2 Pro (w/32GB of RAM costs almost as much as the Studio but the Studio adds 2 more performance cores + almost double the GPU), or a Mac Mini 2018 (I wouldn't get anything older than a Mac Mini 2018) configured as you mentioned. That's going to cost a lot less than either of the first two. Compromising for 16 or 24 GB of RAM brings the price of the base Mac Mini M2 down quite a bit so could be a price/technology sweet spot for you depending on whether that's practical for your datasets, etc. The upcoming M4 Mac Mini may or may not change the equation as well.

MS SQL Server is a good point. I can refrain from using it but it would be nice to not use my NUC just for this purpose. I'm not in a production environment; just projects on Github so Postgres and MySQL will be fine. I haven't used DockerDocker but compatibility here too is a good point.

The issue with an Apple Silicon Mini with 32GB of RAM is that an M1/M2 Mac Studio enters the equation as the price difference is negligible, exactly as you stated. I already have 64GB of RAM and 1TB M.2 SSD lying around from my NUC. If I'm just coding casually, I wonder if the 2018 will be sufficient...

The largest dataset that I've worked with had 5 million records (NYC crime stats) and I will do my best to avoid project like that in the future. But working in Pandas, RAM becomes important.

I use Apple Silicon Mac Mini's (several stacked), an iMac 24", and a MacBook Pro for development with no issues. Currently using satyrn (https://satyrn.app/) for quick python testing, Anaconda, Miniconda and PyCharm for Python work. I've used them to run LLMs, Computer Vision, AI bots, EW command/control, etc. Most of my work today is nodejs / react stacks for prototyping PWAs.

Definitely NOT better to purchase 6+ year old tech. If you are just getting into web dev start easy with a 15" MacBook Air, M3 (~$1200 - $1500US) or a Mac Mini M2(~$500). Good luck!

It sounds like it doesn't make sense to buy the Intel mini. Maybe I can find a 32GB M1/M2 Studio or M2 (Pro) mini on sale once the new M4 comes out. Thanks.

 

[bzgnyc2]

MS SQL Server is a good point. I can refrain from using it but it would be nice to not use my NUC just for this purpose. I'm not in a production environment; just projects on Github so Postgres and MySQL will be fine.

Another alternative is to have an Azure instance that you turn on and off as needed. Not quite as convenient as having a DB you can just leave running on your own computer but the price/minute for one of their large servers is hard to beat (as long as you don't let it sit idle while the meters runs) for what will likely be spikes of batch needs.

However if you have no hard requirement for MS SQL Server, I would try to work within DuckDB and PostgrSQL. Beyond some nice syntax and standards comformance advantages, a particularly nice bonus with PostgrSQL is the PostGIS extension. I am guessing that could be valuable based on your other notes.

Then DuckDB is competitive with MS SQL Server's clustered columnstore for in-SQL analytics work (and better in many ways) with the additional option to be used as a datastore for Python.

I haven't used DockerDocker but compatibility here too is a good point.

The issue with an Apple Silicon Mini with 32GB of RAM is that an M1/M2 Mac Studio enters the equation as the price difference is negligible, exactly as you stated. I already have 64GB of RAM and 1TB M.2 SSD lying around from my NUC. If I'm just coding casually, I wonder if the 2018 will be sufficient...

Performance-wise for what you are doing I believe so. Of course who doesn't want faster? The issue will be slower processors + more RAM versus more faster processors but less RAM (or more everything for a lot more money). Right now I already have a Mini 2018 and I like RAM but I get a little more tempted with each new generation of Apple Silicon.

The largest dataset that I've worked with had 5 million records (NYC crime stats) and I will do my best to avoid project like that in the future. But working in Pandas, RAM becomes important.

Within reason, I wouldn't choose my projects based on my hardware but try to get the hardware that let's me do what I need to do (or learn, etc). That was probably a cool project and would be unfortunate if you had to shy away from cool projects because Apple's RAM prices are high...

It sounds like it doesn't make sense to buy the Intel mini. Maybe I can find a 32GB M1/M2 Studio or M2 (Pro) mini on sale once the new M4 comes out. Thanks.

I think we can share our experiences with you and highlight the key trade-offs but as a sophisticated user I think you'll have to decide what's best for you based on what you do and plan to do.

We also can't judge your budget. You should be able to pick up a Mac Mini 2018 for $2-300 and take it to 64 GB of RAM, etc for another $2-300 (~$500 total). I suspect any Apple Silicon with 32GB+ RAM will run >$1K. For a business this would all be a drop in the bucket and a no-brainer if it meant more produtivity/faster client response time. For other people I know even that first $500 may be hard to come by.

 

[Mity]

I forgot to ask you guys but would your opinion be any different if I told you I needed the mini for 1 year only (2 years at max) before I buy something else?

 

[bzgnyc2]

I forgot to ask you guys but would your opinion be any different if I told you I needed the mini for 1 year only (2 years at max) before I buy something else?

If it were me and money is tight today, I would tend towards upgrading a used Mac Mini 2018 now and then getting a loaded Mac Studio M4 after those become available late next year.

If not urgent, I would probably wait a few more weeks to see if the Mac Mini M4 will be offered with more reasonable RAM prices. If not, I would debate whether I could be productive / unhindered within 16GB or 24GB of RAM and consider a base Mac Mini M2 (though still $800-$1K -- subtract 15% if you can find a refurb model). If not, I would go the used Mac Mini 2018 route knowing it's a bridge purchase.

Another upside of a Mac Mini 2018 is that it will reinforce writing more efficient code as it is less forgiving than faster processors...

 

[Mity]

...I would debate whether I could be productive / unhindered within 16GB or 24GB of RAM and consider a base Mac Mini M2 (though still $800-$1K -- subtract 15% if you can find a refurb model). If not, I would go the used Mac Mini 2018 route knowing it's a bridge purchase.

It's going to come down to RAM and ease of transferring my projects from my 2019 Intel MBP, which I'm giving away to a family member. That machine has 64GB of RAM and a 5600M but I've never used it. The x86 machine's dynamic allocation was quite predictable and I haven't found that to be true for Apple Silicon (I don't want to have this discussion as I know there are benefits to not "wasting RAM"). I also don't want a Studio because I know I will never use the GPU.

Another upside of a Mac Mini 2018 is that it will reinforce writing more efficient code as it is less forgiving than faster processors...

🤣

 

[960design]

It's going to come down to RAM...

I work on billion+ rows of data in Python with <1 second response times with 16 GB of RAM. You may need to 'upgrade' the code to handle large data sets; instead of just getting more RAM. There comes a point where RAM just isn't enough.

...and ease of transferring my projects from my 2019 Intel MBP...

Git and docker. Two cli commands and the software can be moved to the new machine and started.

 

[bzgnyc2]

It's going to come down to RAM and ease of transferring my projects from my 2019 Intel MBP, which I'm giving away to a family member. That machine has 64GB of RAM and a 5600M but I've never used it.

One advantage you have right now is that you have a machine with the RAM you need already running at least your current projects. How much RAM are each of the items in your stack actually using? Are you sure you are using all that 64 GB?

I've noticed a lot of apps allocate far more RAM than they need and then it sits there like old junk in a storage unit doing nothing until the process passes away. On the flip side this can get Compressed or paged out to disk with little performance penalty. That's not the way I like to write code but not much we can do about it when it comes to commercial or even giant open source packages.

Given all the dynamic virtual memory allocation and management behinds the scenes, you might play with memory_pressure while running your programs to see what kind of slow down you experience under your current system as less memory becomes available for your programs. I would generally want my CPU-bound programs to run at 99% of peak but memory beyond that isn't doing much for you.

If Apple's memory prices were like $5/GB, a few levels of rounding up or buying just in case memory wouldn't matter so much but unfortunately it isn't for any recent or forseeable models.

I also don't want a Studio because I know I will never use the GPU.

Don't think of it as paying extra for more GPU. Think of it as getting free GPU with the purchase of extra RAM and cooling capacity...

Also as analytics libraries are increasingly tuned to automatically use GPU and NPU, more GPU might become more interesting than you expect even if your future endeavors don't include 3D graphics and visualization.