Years back, I spent $19,500 for a brand new Toyota Prius-C. Today it still gets well over 50 MGP in the city. I don't pay any attention to the price of gas.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple Says Mac Studio and Mac Mini Will Be in Short Supply for Months
- Thread starter MacRumors
- Start date
- Sort by reaction score
Not only are the m-series minis good for local AI, but I have some late 2014 mins I use as a "mini PC after replacing MacOS with Proxmox. They are cheaper and more reliable than equivalent spec mini PCs.
Apple was so poor at marketing these computers for off-purpose use cases. These should be selling by the ruckload for industrial uses. Even for things like running security cameras, they are great.
Apparently, the word is out.
Oh yeah. Mac Minis were under marketed as “entry level” bring your own keyboard, monitor and mouse computer.
$8b in Mac revenue in one quarter is little when you have the most efficient hardware for local inference. In comparison Nvidia’s quarterly revenue for hardware is $68b. Clearly lots of room for Apple to find ways to ramp up production. Obviously they won’t hit Nvidia revenue but when high capacity Mac minis are sold out everywhere, demand was clearly poorly anticipated.
chars1ub0w
macrumors 6502a
Mac Mini goes up to 64GB unified memory.
There are also mini PCs with unified memory up to 128GB. They're based on the AMD Ryzen AI Max+ 395 chip. Very popular for running local AI models.
Correct. However setup is not as plug and play as it is for Macs. A big issue is also that on those PCs inference is faster on Linux than windows. However if you want to setup an autonomous agent, on Linux it has less access to popular applications.
Meanwhile with a Mac Mini, the ai agent has access to iMessage, mail and any application that requires mouse and keyboard input. Basically anything that you might have access to on your main machine, the agent can have access to own its on machine.
However, beggars can’t be choosers. I might have to buy an AMD Ryzen Mini pc later this year.
This news will not help those with a long term commitment to the Apple environment on the desktop. I suppose they must prioritize their iPhone line etc. but those of us who still are committed to them as a desktop system are left wondering if they really care. Maybe time to consider alternatives.
LFO8
macrumors 6502
Just common sense to not advertise to the outside world that you are worth robbing..
Apart from that, there’s very little one can do when it comes to the behavior of others. I assume ok enough intentions and a ‘no harm’ attitude in almost anyone until proven otherwise on a case by case basis. Because the alternative; assuming the world is out to get you, leaving in fear or paranoia is a real ****** way to live and most likely a very false assumption that won’t really make you any safer in the end.
And there is always randomness and sometimes insanity.. can’t prevent that even.
chars1ub0w
macrumors 6502a
Since the M5 is in plentiful supply on the laptop side, e.g. MacBook Air or Macbook Pro, there's no technical reason not to release the M5 Mac Mini. However, I'm pretty sure the laptops sell much better, are more expensive (and with a higher margin?) than a Mini, so Apple probably prioritizes those. How many Mini users will buy an Apple monitor?
Lol, so true. Also ironic and somewhat frightening. What are they going to do with themselves?
Laptops are better business because if consumers buy them, it increasingly ties them into the Apple eco-system and they end up buying more Apple devices and Apple services.
hoorayforhollywood
macrumors regular
I'm not a nervous guy but when I picked up my Mac mini M4 in London's Regent St store I walked straight out and hailed a cab for security simply due to paranoia.
Barely. Inference speed is dependent on the computer hardware and in particular the memory bandwidth. The operating system makes negligible difference. 1 or 2 extra tokens either way isn’t going to change the user experience.
Whether it's training or inference, the entire stack of hardware and software work in concert. In fact, Nvidia is preferred to AMD because of superior software(CUDA) and networking infrastructure. Raw compute and memory bandwidth aren't the end all be all.
Windows carry heavy resource overhead. It's highly bloated and inefficient regardless of task whether it's video editing, gaming or llm inference. 20-50% tokens per second difference in using linux instead of windows compounds to minutes and several hours for long horizon tasks just in one week.
Mac Mini is a bliss as it's well optimised for local inference out of the box whilst remaining consumer friendly.
woodbine
macrumors regular
here in the UK, some nefarious folk will steal a bag of spuds as you leave the store..../j
woodbine
macrumors regular
I'm not so sure Apple has dropped the 'AI ball' or is even 'late to the party'. When the big tech companies are throwing billions into AI which is looking very frothy, it's always been Apple's DNA to take the wait and see approach.
Firms who want to use AI, but have a business that involves client confidentiality, are realising they need to keep AI in house.
I am pretty much at that point. Latest hardware from Intel or AMD and NVIDIA still outperform Macs by a great deal. I just hate windows and I cannot use Linux. macOS is the best OS for me, but Apple's treatment on the desktop is overly frustrating.
tubular
macrumors 65816
Absolutely correct. I remember Apple trying to sell their 2014 Mac mini -- busted back to dual core -- as a starter kit for switchers, as if that were the only reason to consider the form factor.
I've been using minis as my main driver since maybe 2010 or so, and I've always thought there was a huge gap between their capability and their popularity.
Except 2014-2018, when Apple dumbed down the mini because it didn't know what the hell it was doing with headless desktops, but that's a different rant and fortunately they've seen the light since then.
The conversion was about Linux vs Windows. You shifted it to Nvidia vs AMD. Bizarre.
Yes, it is. The prices for compute hardware and the sanctions tell us everything we need to know, along with the benchmarks.
Windows does not carry a “heavy resource overhead”. You simply do not understand what you are talking about. You are confusing window decoration and a Start menu for the kind of bloat that impacts performance by more than a one or two percent either way. Sometimes Windows will be faster at a task or game, sometimes Linux will be. UI makes almost no difference and there are plenty of ugly bloated Linux UIs.
That’s even worse than the rest of the nonsense you posted. There is no 20-50% tps difference given Windows or Linux the exact same model, same hardware and same version inference app.
You simple tried to lie and manipulate the readers on this board because you thought they are ignorant enough to believe Linux is like some kind of magical OS created by fairies and unicorns. It’s so magical it can perform up to 20-50% faster than computer hardware allows.