I Returned the Mac Mini M2 Pro
jeremymorgan.com
jeremymorgan.com
I don't disagree that with the main conclusion, it is a lot of money and for most people it doesn't make sense. If I was spending $2600 I'd be looking at places that do certified refurb M1 Ultra Mac Studios with at least 64gb unified memory, which I think will be going a long way soon to helping future proof a machine for the next few years.
> 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.
It is true that Apple is pushing the ANE very hard in its marketing considering how little apps actually use it.
"You may like the Mac Mini if: You require the OSX environment for your work (I don’t)"
Author should have started the rant with that.
Then you have articles like these: https://techjourneyman.com/blog/great-macos-exclusive-apps/
> Despite having a version of it in Windows, there’s a good reason why people preferred using Photoshop / Illustrator on Macs. First and foremost, Photoshop was developed on the Macs first. Color is more accurate on the mac out of the box than on windows. macOS is more stable than Windows. These reasons make Macs the preferred platform for graphic illustrators around the world.
wut?
Color is more accurate on a Mac if you don't pay attention to the display you purchase with/attached to your device. Yes, Photoshop was first developed on the Mac... 33 years ago.
I get if you like macOS, but the marketing behind it has done a really good job to paint it as if it's the only option out there. It's insane.
As someone involved in both, most people don't work in commercial audio production or photography. And even for audio and photography, Windows has similar or even the exact same applications available, with perfectly fine performance.
Ableton, Kontakt, Lightroom, Premiere and Resolve all work fine on "not macos"
As for color accuracy, I have a plenty expensive monitor (bought on black friday, of course, never pay full price for expensive hardware) and a Spyder calibrator because factory calibration is a nicety that stops being true a month into using the thing. Although of course the mac mini doesn't even have a display, so it can't even get points there =D
It’s able to handle colour profiles correctly on a per app basis, rather than per display.
https://developer.apple.com/library/archive/technotes/tn2313...
I mean if they have that feature and it's used for something big I'm not privy to, cool, but otherwise it's something that most large OS's handle.
Different files, applications, media can be encoded in different colour spaces, as can different displays, printers, dyeing processes take them as inputs - and the conversion isn't always straight forward.
For most consumption activities this doesn't matter, but if you work in colour and your job is to make sure that that signiture 'Pepsi Blue' is consistent in print media, television advert, merchandise, and the product, some finer treatment of them in the OS becomes necessary.
To iterate - not a colour expert, but I just thought I'd politely enlighten you as why macOS is so highly regarded for this kind of work, it isn't just marketing. Though no doubt, there is plenty of that too.
He just says his workflow is not bound to macOS (like it would be if you would do iOS development in Xcode for example) so he's got other options too.
I'm in the same place. I also moved from Mac to a NUC with FreeBSD as my main driver. It was cheaper than a medium mac mini and that's with 64GB RAM in it and 1TB NVMe 980 Pro SSD. Whereas the mini that's more expensive wouldn't even have half that.
Additionally I have much much more customization of my user experience with KDE. So I can really tailor the interface to my workflow. This was a bigger reason to switch than the hardware price. The ability to max out the RAM for cheap was also not 100% necessary but it's nice not having to worry about this later. And it allows me to run VMs whenever I need.
Fascinating. I use Sublime daily on my M2 MacBook Air and haven't encountered this.
With all these GUI apps crashing and monitor problem, I wonder if there was an issue with his specific Mini Pro, like with the GPU.
Although in my case it might have been a specific Language Server Protocol (LSP) package I was using that was consuming 100%+ of my CPU (they fixed it).
Used to have an issue where my bluetooth headphones would annoyingly hiccup every minute or so while listening to music.
I've since kept the 3rd monitor on and haven't had any problems. So, if anyone has this problem, try fiddling with your DisplayLink connection: perhaps try disconnecting the adapter completely and/or try turning it off/on from the manager.
I'm not sure that you would be happy with the dual monitor performance - it was one of the few complaints people had about the first M1 chips. Does your MBP have an M1 or an M1 Pro?
I've got one and really enjoy it, but I can't get it to output to a 1440p monitor. Two different 4K monitors - no problem.
My workflow isn't nearly as heavy as OPs (mostly DevOps stuff; some Kubernetes clusters here, some toy Go apps there), but it does a great job at these tasks while being super quiet and fast about it.
Our other Mac Mini sits in our living room and is used for Zoom game nights. We play Jackbox on it.
The Intel Mac mini that it replaced would really struggle with Jackbox despite having "fast" chips. On the M1 mini, the fans don't even turn on, even while sharing it over Zoom with other people's video on.
Although it is true that an Intel machine with a dedicated GPU is just “easier” and can be assembled for far less (I did just that a month ago: https://taoofmac.com/space/blog/2023/02/18/1845), none of the other criticisms are valid and seem to stem from a lack of understanding of how to actually use the machine and what it excels at (I intend to get one myself, despite having built a CUDA box, because it excels at audio and video editing).
The only thing that is worth taking away is that Apple kneecaps the entry-level M2 Pro configuration with only 512GB of storage (it should start at 1TB).
Classic case of “holding it wrong” and doing it for the clicks/viewers/controversy.
I did not manage to boot from USB drive at all.
"Why advertise some super AI chip if it doesn’t do AI stuff? Who knows." => Neural engine is fp16/int8 limited so it wouldnt be good enough for training anyway but its pretty good for inference with CoreML in a lot of cases. Also training anything outside Nvidia ecosystem is very bad cause software was written with CUDA first and everything else second (AMD for example have similar story to Metal backend).
As someone who works on Deep learning for past 6 years I can easily say CUDA wont be replaced anytime soon at least for training, almost every software out there was written for either CUDA or CPU.
And for local model development you don't need your rig to be powerful as training wont be done on your local machine anyway and MacBooks are good enough for that (a lot better than windows for sure).
To make a meaningful cost comparison we need to include an equivalent computer in which to put Nvidia's GPU.
Then we get into what workloads you need the GPU for.
Now he mentions tensor flow which probably has very different requirements than running an LLM. But the point about the toolchain is a good one IMO.
Also, running an LLM is pretty niche for now since there are no open models yet (meta's is only an unofficial leak)
Also - it was unclear - did the software have issues with the M1? Or only the M2 Pro? I have no idea what OBS is.
The bugs mentioned, sporadic crashes across a range of apps, seem odd. Maybe a faulty unit or a bad OS install? I’ve occasionally had Macs come “fresh out of the box” with buggy OS installs.
As mentioned, unless you absolutely need OSX (which tends to not be true for the vast majority of users), the "why don't I just put a 4080 in a PC" route is a pretty reasonable path to go down once you realize that doing the work you need to do is more important than the machine you do it on.
One thing he told me sticks in my mind though.
The airline would periodically have problems - cancelled flights, bumped passengers, etc. He said this was disservice, and it was handled ok by most normal people. But one situation was different. He said that first class customers, who would routinely pay 10x the price for a ticket - they did not handle bad service well. Disservice a first-class customer and they would be MAD.
Or even performance-per-watt?
Is it so hard to make a list? Ideally, a sortable one?
Sorry, but your reply is really funny, even if it's a jab at me.
I outlined the reasons for the 4080, full knowing it's drastically less performance than the 4090. I balanced power usage, heat and noise vs how much performance I need.
Doesn't make your comment any less hilarious though.
I strongly considered the 4090, at $500 more you get a lot more performance. (if you can find one). However it mostly came down to this:
1. I'd need a new power supply as well (need 4 PCIe Ports) 2. The power usage is drastically more (I care about this) 3. More heat and noise 4. I don't need the extra performance that much.
So yeah from a value perspective 4090 is a better bet for most people, but after careful consideration I went for the 4080. It's been fantastic so far.
It replaced a macbook pro (with touch bar, no less) that costed me $4500. It swelled on me about two year after purchase, shortly after the expiration of the standard warranty.
I live in vscode, a plethora of browser addons and tabs (with auto-suspend extensions after 30min of idle), some Teams and Google Meet sessions (video piped through OBS) and some terraform every once in a while; no discernible performance difference between the two machines 99% of the time, although the remaining 1% is a swapping-party on my Mac Mini because 8GB of RAM is certainly not enough.