Last Update: Nov 12, 2023
I was super excited when my M2 Mac Mini arrived. So excited I wanted to take the rest of the day off work just to set it up. That excitement didn’t last very long.
I returned my Mac Mini M2 Pro, and this is why. I’m not trying to bash here. This isn’t a hit piece. I generated a lot of talk and tons of instant messages on Twitter and other platforms. I thought it would be easier to explain here and send people the link.
If you want to skip to the bad stuff, click here
Also I’ve posted an update regarding my video card swap.
What I Ordered
When I decided to get the new Mini I knew I had to get the Pro. If I’m spending the money I won’t wonder what more performance is like. I’m fortunate to have a yearly computer budget and can swing “the big one” again.
So I got the beefiest SoC combo:
Apple M2 Pro with
- 12‑core CPU
- 19-core GPU
- 16‑core Neural Engine
- 32GB unified memory
I selected 2TB of storage and the 10 Gigabit Ethernet. Cool.
Total Price: $2699. Yikes, more than twice what I paid for my M1 Mini. But, if it’s fast and problem free for the next few years, It’s worth it. It arrived in a nice little box, and I set it up.
Why I Ordered It
I purchased the M1 Mac Mini way back in 2021. It is less than half the price, but half the hardware as well:
Apple M1 chip with
- 8‑core CPU
- 8‑core GPU
- 16‑core Neural Engine
- 16GB unified memory
- 1TB SSD storage and Gigabit Ethernet
And it was (and still is) a fantastic machine. I was so impressed with it over the last two years. I expected the M2 to exceed my expectations, be screaming fast and super reliable, and make me a happy little coder.
I expected it to keep up with or even beat my Digital Storm workstation I’ve used for “big jobs” over the years. I even thought I may leave the Digital Storm box powered off the rest of its days.
What I liked about the M2 Pro
Performance is super important to me. Naturally, the first thing I did after setting it up was run some benchmarks.
Definitely faster than the M1. CPU speed can be difficult to notice these days, though. Nearly every machine made in the last 10 years is “pretty fast” for browsing around and doing regular stuff.
What about the heavy stuff?
- Software development
- Video Production / Rendering
- Occasional AI/ML work
I push my machines with this type of work nearly every day. The tasks above require a GPU+CPU combo, so what’s the GeekBench 6 compute score?
Wow, impressive! More than twice the compute power. This doesn’t surprise me much. More cores, better SoC, etc. Unified memory helps a lot too. So the machine is undoubtedly faster than the M1.
I liked the performance. The form factor, of course, is great. The Mac Mini doesn’t take up much room or use a lot of power. Ideally, I could power down my “big computer” and work for days with the Mini, as I’ve done with the M1.
Once I was satisfied with the benchmarks, it was time to put it in service. BTW, There are more detailed benchmarks here.
What I Didn’t like
I used this machine for about five days. During that time I put it to work. I had some issues I’ll describe in the next section, but let’s talk about performance. Specifically, applied performance. The real world stuff.
Rendering - I filmed and produced some videos and noticed that it wasn’t much faster than what I’m used to. It shaved off some seconds here and there and a minute off the time with 4k. Less than I expected, but still an improvement.
I expected it to exceed the performance of my Digital Storm workstation (see chart), but it was slower in rendering by a significant amount.
Interesting right? Maybe my expectations need to be adjusted because this is a tiny workstation compared to a giant box with a dedicated GPU. Here’s what Geekbench had to say:
According to this, it should be marginally faster than the Digital Storm machine, yet less fast than my laptop (with an NVidia RTX 3060 Card in it). It turns out it’s slower than both when rendering.
Not good, considering that’s a big thing I do pretty frequently.
Machine Learning - OK, what about Tensorflow/Machine Learning stuff? After all, it’s “built for AI” with the Neural Engine at your fingertips. I usually do this on my Linux machines, but I figured I’d give it a shot.
Running Tensorflow on the M1 has always been a hack at best, and is even worse with the M2.
Despite going thru the “official Apple” docs, several Tutorials, and YouTube videos describing different methods (Using Anaconda, MiniConda, Forge, or blah blah), I ended up empty handed.
Every tutorial said to do something different and something has changed since each of them were published, so I never did get Tensorflow to recognize my GPU. CPU only training was meh.
Now, I could have spent more time on it, but many people got the GPU to work with Tensorflow, and apparently, it’s super slow.
Why advertise some super AI chip if it doesn’t do AI stuff? Who knows.
But all of this is acceptable at this point. I have a machine for rendering and ML stuff I can use. However, I did start to think to myself:
Maybe I should have put my yearly “new computer” budget towards a better GPU for my Digital Storm machine instead.”
Which is exactly what I did. Read on to find out why.
So, all things considered, the performance above is acceptable. I expected more, but it’s alright. If it were my only machine, I might be able to live with it all at this price point, especially if it were as reliable as my M1 has been.
This machine was riddled with software errors the entire time I used it. All small errors, but they add up. Here are a few:
Cannot close windows in OBS - Minor annoyance, but any windows other than the main OBS window would simply not close. Weird.
OBS does not communicate with Stream Deck well - I kept losing connection with the Stream Deck, even in the middle of the stream. Just quit.
OBS shows weird display errors when streaming - On a live stream, I was talking away, and someone mentioned only seeing the GitHub screen. OBS only showed a browser window underneath the windows I saw on my monitor. Super weird.
Chrome lags then crashes - Yeah this happened a lot. Chrome would randomly die an agonizing death. Really annoying.
Typing lags horribly in Sublime Text - I couldn’t figure this one out. I couldn’t use Sublime at all because of a weird lag with typing and navigating through text. It never went away. (Yes, I still use Sublime every day. Sorry not sorry.)
Hugo crashes sporadically - I run Hugo locally to build content for websites. It would error out and crash.
GitHub desktop crashes sporadically - The window just dissappears, with no logs.
VS code lags and crashes - This was more intermittent than the Sublime text issue, but still a problem.
Finder window does not open to the right window - You click on a folder, which will take you to the root or another folder you visited. This one frustrated me to no end.
Premiere rendering bleh - I expected more performance. Many renders crashed all together.
Windows don’t close - Many windows refused to close when asked. If I logged out and logged back in, it was fine. (Imagine doing that several times a day!)
Second Monitor lost signal - This happened a few times daily. My Second monitor would lose signal and power off. I tried swapping the HDMI cables (since I just purchased new ones), and that didn’t help. It’s never done this with any other machine. The monitor appears to be functioning fine.
These could all be problems with
OSX Ventura and not the M2 or the machine itself. But I have nothing else to compare it to other than a Macbook, the mini with the M1, and Monterey that do exactly none of these things.
So with all of these problems, I decided:
The M2 Mac Mini is far overpriced for what it is.
I didn’t feel good about this purchase anymore. I felt duped. The hype machine got another victim.
Why the Mac Mini M2 is Over Priced
For my use, this machine is pointless. It is a $1200 machine at best.
You could say, “well, it’s not designed to be a Machine Learning trainer” to which I say then stop marketing the super duper AI chip.
You could say, “It’s not meant to be a super video production machine” to which I say $2600 is too much for a web browsing box.
You could say, “In six months, they’ll figure out these software bugs” to which I say I’d rather have six months of productivity.
I loved the other Mac Minis. I’ve owned three of them since the late 2000s. But now, I struggle to see the point of the Mac Mini. Not at this price. For nearly three thousand dollars this thing should run like a top and knock your socks off. I’m sorry, but it doesn’t. If performance is a concern you can do better for less money.
The form factor and OSX are now the only selling points.
And you can buy three Intel NUCs of the same size for the price of this Mac Mini.
I paid half the price for my M1 Mini, which was great. But I expected more at this price point. Since it was unusable, I returned it.
Note: The return process was painless. I walked it into the Apple Store and handed it off. The person I dealt with was super friendly, and five minutes later, I was done. Great customer service. They made my sad process easy.
Is the Mac Mini M2 Pro Right For You?
Well, maybe. I know my expectations for such a small computer may be a bit off. But all things considered, my laptop isn’t much bigger and runs laps around it performance wise. So there’s that. The errors were the biggest issue for me.
You may like the Mac Mini if:
- You require the OSX environment for your work (I don’t)
- You can put up with weird bugs till they figure things out (I won’t)
- You don’t need to do any machine learning/AI stuff
- You don’t render video as much as I do
- You need a small form factor machine that’s out of sight but has good performance
Many folks who do web development, graphic design, writing, etc., will probably love this machine. I don’t require OSX, but I do enjoy it. I like airdropping things and how my M1 “just works,” as the Mac folks like to say. I like OSX’s look and feel; it has been stable for many years. I can’t justify the price just to use OSX instead of Windows or Linux. Also, maybe Ventura is a lemon of an operating system right now.
You might like it. I Didn’t. I will continue to use my M1 Mac Mini for casual OSX stuff, and I redirected my yearly computer budget towards an NVidia RTX 4080 for my Digital Storm machine (way less than $2600). I’ll talk about that on this blog soon.
I purchased the RTX 4080, which was less than half the price of the M2 mac, and it’s amazing. Ripping fast.
Personally, this is what I should have done first. This is the Geekbench Compute test. It also renders screaming fast (4K even!)
I’m still bullish on the M series chips, and ARM in general. I’d like to see where they go. But today it’s not the best bang for your buck.
For about $1200 my machine beats a maxed out $5,799 Mac Studio. If I chose the 4090 the gap would be even wider.
Yeah, not for me.
Questions, comments? Want more details? Yell at me!!