Toggle sidebar Toggle sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Apple Stops Accepting Orders for Some Mac Mini and Mac Studio Models

masterin said:
Yes I know they are fused to the M4, but no they don't come out of the TSMC oven looking like that... it's a multi step process.
Click to expand...
TSMC don't make memory chips. LPDDR5X chips are shipped to TSMC from Samsung (vast majority of the time, some maybe also Micron) to be packaged. Almost all the memory go for iPhones, they're priority.

Korea news reported that Samsung committed on March 9th to 12GB LPDDR5X for the folding iPhone. All of the folding displays also come from Samsung. Good times for Samsung as they also got Apple to agree to double the price for the memory chips!
 
RashyIsBack2 said:
Heya, is there a ‚solid‘ middle-ground for something like roleplay or fanfiction/story chatbots? I agree that 24GB are pretty low, but going 128GB is costly. Are there solid models that would harmonize well with 48 or 64GB? I own a 32GB M1 Max and consider upgrading to a M5 Pro or Max soon. Thanks!
Click to expand...
Probably 32 to 64. You could do that on your current machine, but the M5 has better AI performance.
 
Wahlstrm said:
I have the 24gb M4 and while you can load some decent models with 24gb you basically have no memory left to actually use them all that much.

Especially if you have anything else running at the same time.

So you’re basically stuck with asking a good model short/simple questions or a not so good model complex questions.

I would recommend you look at minimum 32GB if running local LLM’s is a real priority.
Click to expand...
And the answers are mostly WRONG for those models. People who think they are getting a "ChatGPT like experience" using local LLMs with under 64 gb of RAM are in for a surprise. You can run them with RAM to spare, but the models are not great.
 
  • Like
Reactions: cavemandaveman
chars1ub0w said:
Good times for Samsung as they also got Apple to agree to double the price for the memory chips!
Click to expand...

Is there a news source for this? And we thought memory prices were already high. But, my complaint about Apple pricing has been the same for the last 10 years -- the cost of flash storage. That is where the Apple tax really bites. I've spent a significant amount of time over the years working around this in various ways. I was hoping that Apple would relent on flash, and, this AI bubble hit. The prices people are paying now-- ouch.
 
Tim Jobs the 2nd said:
More likely it just means new models are coming real soon
Click to expand...

And not just that new models are coming real soon, but which models were more popular than Apple anticipated. If Apple sold more 512GB M3-Ultras with than it anticipated, memory supplies are constrained, and new models are coming soon, then it makes sense for them to adjust 512GB M5-Ultra production upwards than make more 512GB M3-Ultras. It's better to sell off the 256GB M3-Ultras. If you go through the unavailable memory configurations, you'll find where Apple's sales projections were too low. I suspect the popularity of the 512GB M3-Ultra took them by surprise. They are going to be selling 512GB (or larger) M5-Ultras like hotcakes.
 
Last edited:
  • Like
Reactions: Ronarb
Very likely both are true, M5 is coming and ram shortages. Processor configs that are already built or in the pipeline are still for sale with extended lead times. Models that are rarely ordered and those chips are never stocked in large quantities are no longer available for purchase. Apple will not build new M4 chips with high ram configs as that ram is being reserved for M5 models and they dont want to deal with returns of M4 128GB models as soon as the M5 drops.
 
BSDnostalgia said:
Is there a news source for this? And we thought memory prices were already high. But, my complaint about Apple pricing has been the same for the last 10 years -- the cost of flash storage. That is where the Apple tax really bites.
Click to expand...
Yes, of course.
An official from the memory semiconductor industry said, "I know that it has led to a price increase that is almost double compared to the previous year," and "Nevertheless, Apple is supplied at a lower price than other customers."
https://www.thebell.co.kr/front/newsview.asp?click=F&key=202603091424197880103382
 
  • Like
Reactions: Ronarb and BSDnostalgia
vantelimus said:
And not just that new models are coming real soon, but which models were more popular than Apple anticipated. If Apple sold more 512GB M3-Ultras with than it anticipated, memory supplies are constrained, and new models are coming soon, then it makes sense for them to adjust 512GB M5-Ultra production upwards than make more 512GB M3-Ultras.
Click to expand...
The M3 generation used LPDRR5-6400 RAM chips. With respect to Apple Silicon, these are obsolete now.
M5 CPUs announced so far use LPDDR5X-9600 RAM chips.
 
  • Like
Reactions: vantelimus
glockenSquish said:
This makes me so angry. I was literally just about to buy a Mac studio. I have an enormous year-long project starting right now (music production) and desperately needed the upgrade (still running an Intel Mac). I wanted the Mac Studio for fear of the M5 MacBook Pro making fan noise when working on a large session. Also wanted higher ram in the Mac Studio. If they really are releasing a new line, announced at WWDC, that is pushing it for me and my project. Plus, it’s still a gamble.
Click to expand...
I would not wait. Buy one M studio now. You can always resale it later for a M5.
 
kalsta said:
Right. We’re really talking about two different things here aren’t we.

Is AI an ‘overhyped’ fad as many people seem to think? No. It’s a major tech revolution that, for good or for bad, is going to change life on earth forever. But it’s not surprising that many people can’t see that. I’m quite sure when the motor car first arrived, many people laughed at all its flaws and claimed that it would never replace the horse. I even remember trying to convince my dad that the World Wide Web was not just a fad.

Are some early investors in AI going to fail? It certainly looks that way. But some will succeed, and as always, a few filthy rich billionaires will get even filthier from it.
Click to expand...
100%. There are different categories of pet when it comes to tech acceptance.
Laggers are the last one I believe. They finally make the move when they think it’s safe and 75% of the population is using the tech !!
I’m 200% into AI. Taking classes about AI strategies. Implications are huge.
I just hope Apple pays attention to MacMini clawbot frenzy and continue the momentum.
Steve created a new era with music, video ; digital revolution while MS was still fighting for browser dominance. They never caught up.
Apple needs to realize that video, sound, music creation is gonna be done With AI as assistant. So AI assistant need to be part of Mac hardware to run local LLM.
Nvidia is pushing hard with DGx, Dell, HP too, but at $4000 no thanks.
Need something between M studio and M Mini Pro. Buying a M studio to run LLm is a waste.
 
  • Like
Reactions: Ronarb
Padre_Pio said:
And the answers are mostly WRONG for those models. People who think they are getting a "ChatGPT like experience" using local LLMs with under 64 gb of RAM are in for a surprise. You can run them with RAM to spare, but the models are not great.
Click to expand...
Untrue. It depends on what you need to run. Gemma apparently is pretty good. Mistral and Arwen 3.5 too.
Some have reasoning. You can do the grunt work on Local LLm and send to Claude for more advanced reflections
 
  • Like
Reactions: Howard2k
theluggage said:
There's more to a "bubble" than some early investors failing. That happens with virtually every development and it's often the second mouse that gets the cheese.
…
When the bubble bursts (e.g. the dotcom bubble) it doesn't just remove the also-rans from that specific market - it causes a wider stock market crash and "colateral damage" that costs everybody.
…
You are trying to dismiss the existence of the AI bubble and the chaos it is already causing as just another case of there always being winners and losers in business. There is a lot more to the current AI bubble than that.
Click to expand...

I’m not going to argue with you for two reasons:
  1. Firstly, I don’t completely disagree with you. What we’re seeing here is a major disruption. It’s already having a negative impact in some ways, and I believe we’re only just scratching the surface.
  2. As with the comment I replied to here, there’s a fair bit of straw man in your remarks to me. I enjoy engaging with people on topics like this, but not so much when they make assumptions about what I think or put words in my mouth. Again, I’m not wanting to start an argument here—just being open with you.
 
Patroclus said:
Untrue. It depends on what you need to run. Gemma apparently is pretty good. Mistral and Arwen 3.5 too.
Some have reasoning. You can do the grunt work on Local LLm and send to Claude for more advanced reflections
Click to expand...


Exactly.

And architectural changes have made models more efficient over the last 6 months.

To some extent and in some cases, what you needed 24GB to do in the past can be done on 12GB/16GB now.
 
Hoping the M5 Ultra is released very soon as I'm verging on buying an M5 Max and I don't need the screen or portability. I need 10GB Ethernet too.
 
AbeLincoln1984 said:
You might be surprised how few people share your opinions.

Sometimes great things are unpopular. But this ai stuff is being forced on us all everywhere all the time, and we cannot get away from it fast enough (or at all, unless we abandoned computers and the internet entirely, which we can also not do if we want to live and do business).

I can't remember a piece of technology shoved down my throat quite as hard as these slop generators.

I'm not sorry for wanting it out of my face. I never asked for this. I don't want this. Please, leave me behind as fast as possible.

I'd still like to use technology the way I did three short years ago. But we can't even hide this stuff. It's everywhere.

I recommend you sit and think about the wisdom of forcing technology on people who don't want it. Because it seems morally wrong to me.

If a sit-down restaurant forced me to order and pay with a tablet, I'd never go there again. If I sit down at a table and there's a tablet on it, the tablet goes facedown on on the seats.

Technology exists to serve us. At its best, technology is simply swapping one set of problems for a preferable set of problems. If technology doesn't serve us, I'm eager to let go of it.

You might notice the Amish still reproduce. Might want to think about that too.
Click to expand...
So much this. I'm so so so sick of AI being shoved into every damn thing. There are some good valid use cases for it, yes and we should focus on improving it for those things, but I want it mostly out of my life. All I have seen is the reduction in quality in damn near EVERYTHING since it has been shoved down our throats. As far as I'm concerned, AI can **** off for the most part.
 
  • Love
  • Like
Reactions: krell100 and AbeLincoln1984
Padre_Pio said:
That's what I did. Let me tell you--the local AI models when you use 24 gb of RAM are terrible. They get most answers wrong. It will look like you are running ChatGPT or whatever locally, but its knowledge is trash. Wasn't worth it for me, and I'm told you need at least 100 gb to get an experience approaching the paid, cloud models.
Click to expand...
Yeah I told my buddy he needs 64GB's minimum. I do hope they bring back the 512GB's models on the Studio just because that has got to be insanely useful for at least a handful of people lol. That or at least give us a 384GB option.
 
HVDynamo said:
So much this. I'm so so so sick of AI being shoved into every damn thing. There are some good valid use cases for it, yes and we should focus on improving it for those things, but I want it mostly out of my life. All I have seen is the reduction in quality in damn near EVERYTHING since it has been shoved down our throats. As far as I'm concerned, AI can **** off for the most part.
Click to expand...
I use AI quite a lot but agree here. If someonme wants to use it, have them install it. Don't install all the crap by default.
 
  • Like
Reactions: krell100 and HVDynamo
Wow! I guess I ordered my Mac Mini with 32 GB of RAM just in time. My ship date slipped to August at first, but it has changed to the end of April. I needed a second system for a cluster I am building.
 
atomwork said:
A company that makes 40 billion on rev each quarter cannot build their on ChipFab in USA? Yet Tesla/SpaceX can?
Click to expand...
I'm sure if they call ASML and ask nicely they'll send one right over

Musk's announcement that they were going to fab processors was just that--an announcement. He has to pony up like everyone else if he wants a fab plant.
 
  • Like
Reactions: uczcret
hail_ants said:
I'm sure if they call ASML and ask nicely they'll send one right over

Musk's announcement that they were going to fab processors was just that--an announcement. He has to pony up like everyone else if he wants a fab plant.
Click to expand...
The most technically complex and precise machines ever created by man. Only manufactured by a single company in a foreign country.

Super easy. /s
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back