You could be correct, but previous iterations of their CPUS used standard components, or at least standard compatible components. So I’d be surprised if Apple wasn’t using something they couldn’t source from at least two different vendors here.
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.
M1 Ultra Mac Studio is 2 Pounds Heavier Than M1 Max Version Due to Thermal Differences
- Thread starter MacRumors
- Start date
- Sort by reaction score
You could be correct, but previous iterations of their CPUS used standard components, or at least standard compatible components. So I’d be surprised if Apple wasn’t using something they couldn’t source from at least two different vendors here.
Explain why extra conductive material in the path helps latency?
In a portable device keeping things secure is more critical than a desktop, and especially in phones and tablets where they are more often used while in motion. Solder and glue are the two easiest ways to accomplish this, however there are trade offs. Serviceability and upgradability being the main ones.
Now if you guys want a better argument, maybe Apple has found all other methods impossible or highly problematic? But I would counter that’s likely in portables, however desktops have way more flexibility since they are largely stationary. Apple has created a series of clever component clamps inside several Mac Pros models alone, and have a well earned reputation for high precision mechanical engineering in their electronics products and cases.
There are likely overriding business reasons, like maybe wanting to help avoid the situation with the pre-2013 Mac Pros, which institutional users have been able to keep in service long after Apple’s ideal product life cycle.
Allowing that to happen further down the chain in the “prosumer” end, could create a similar situation but for a much wider swath of the customer base. Products with Apple silicon already have relatively long reliability and lifespans. A Mac Studio likely has a similar long lifespan, and with upgradable components even longer.
The bean counters have assuredly factored this in, and it’s no secret Apple has been increasingly stingy with the extras, especially since the last pre Mini class desktops (no keyboards / mice). For portables there are better arguments for permanent components. But for desktops, it is a bit disappointing on the permanent hard drives alone.
Nothing about ARM CPUs is particularly standard. And, again, since the A6 (so for almost a decade now), the CPU design is custom.
The RAM is standard LPDDR stuff, if that's what you mean. As has been pointed out, putting that in the same package as the CPU comes with latency benefits, so while you might regard it as a cynical way of Apple to make more profit, it does have performance benefits as well.
Yes, others can theoretically manufacturer an Apple SoC as well, and Samsung has in the past, but it has nothing to do with the M1 being "an ARM CPU; many people make these".
You can’t just use pressure to hold the memory in. You need continuous metal between each pin and each solder bump. It just doesn’t work the way you think it does.
Continuous metal helps latency because it conducts and has a low impedance. A slot, clamp, or other type of connection has higher impedance which increases latency. This is simple electrical engineering. Not to mention work function effects. That’s advanced electrical engineering. You keep ignoring the laws of physics.
What are you talking about?
It’s a full custom cpu designed by apple and dabbed by tsmc. No other fab could make the die. Other fans have other design rules, so apple would have to redo the physical design for each fab.
What are you talking about “extra” conductive material? Compared to what? ZERO conductive material? With no conductive material a charge can’t be carried.
Apple could sell their own memory modules, they clearly make different types that are interchangeable on the M1 chip “socket”. See previous post links.
Also it’s likely had they made these serviceable, any of the major brands would have made something compatible. Right now there’s no incentive. Are you familiar, like at all, with the history of the very active Apple 3rd party market?
If there’s a serviceable slot, somebody is always ready to plug it. Bank on it. It’s one of the biggest companies on the planet, and this was true when Apple was tiny.
They could, but again, they're soldered, and they're soldered in part due to latency reasons, so, no.
Barking up the wrong tree here. My first Mac was in 1992.
And yes, I miss the ability to buy a new Mac every x years and then upgrade the RAM mid-cycle. I do. But it's gone. And it's gone in part in favor of other benefits: the newer devices are smaller, lighter, less prone to reliability issues, and have lower latency.
What was also true when Apple was tiny is that Steve Jobs strongly believed the Mac shouldn't need a lot of internal expansion. Look at the original Macintosh 128k. And the 1998 iMac. So you're bringing up an argument from 38 years ago, and, like, come on, time to move on.
I said components, not fabrication. Apple and TSMC have more than one source for those.
You are really hung up on this all in one CPU notion (marketing), the memory modules are clearly fabbed in separate processes and later soldered to the CPU assembly. Check the previous links and photos.
You weren't (just) talking about memory. You said: "The M1 is an ARM CPU, many people make these and the associated memory component types."
Yes, many make LPDDR memory. No, not "many people" make the Apple M1. In fact, nobody other than TSMC makes them.
Explain how solder helps latency?
Steve changed his mind a lot over the years, and made expandable desktops (and laptops) over his entire career at Apple.
There are contacts on both the CPU assembly and memory module.
Ever use a AA battery? Do you solder those in?
Check the previous photos. The memory is on seperate modules.
NOBODY knows how to make a LPDDR memory module like Apple’s? They can’t literally pry it apart and reverse engineer? Goofballs on social media are swapping these things. Whose IP is Apple licensing? Come on.
Wow, what a straw man argument. He never said nobody knows how to make LPDDR like apple’s. He said M1 is a CPU, and nobody knows how to make the M1 other than TSMC. He knows the memory is on separate modules. He pointed out that you stated that “many people make these,” referring to the *CPU*
Apple uses off-the-shelf LPDDR modules sourced from Hynix. There is nothing unique nor proprietary about those modules and they are soldered in a lot of things from a lot of OEMs. Same with their SSDs and other components.
The only item that is probably truly "Apple proprietary" is the M series SoC and as cmaier noted, they are designed to be fabricated by TSMC using TSMC's proprietary fabrication process. Hence why only TSMC manufactures M series SoCs and why Apple can't just call up Samsung or Intel or some other fabricator to make them because their fabrication processes are incompatible with the design.
That is a profoundly foolish statement. So nobody else makes 'puter stuff?
Nobody else in the entire electronics industry can look and copy what Apple did, ditto for the memory modules they stuck on top of it (which are less difficult to design/re-design than the CPU part). Nobody at Samsung can reverse engineer the memory and figure out Apple's "pin" layout? That's nuts.
Is this just crazy fanboy blindness? Need some pets? Yes, the M1 is great, Apple was a very good boy. Apple will continue to make great stuff and be a very good boy. Happy?
No, just you putting up more straw man arguments and changing the goalposts. Nobody said only one supplier makes LPDDR.
Why?
I think others in this thread have explained the latency thing plenty.
Yes, when he considered it a market necessity. It was never his preference. The computers he truly loved weren't expandable. The Mac, the NeXT Cube, the iMac, the ill-fated Power Mac G4 Cube, the iPod, the iPhone: none of those offer much in the way of internal expansion, and all of those are designs he adored.
He would've approached the 2019 Mac Pro with a deep "do we have to? alright, fine" sigh, not with enthusiasm.
Nobody is arguing against that.
Is there a latency component related to AA batteries that has heretofore gone unknown?
Otherwise, a AA battery invariably requires a connection that is almost, but not quite, entirely unlike a high speed data connection.
Do you know the difference between a static DC voltage - a voltage which is *required* to vary very slowly (so that, for example, parasitic capacitance is *good*), and a high frequency data signal, which required to vary very quickly (many millions of times a second)?
I guess you don’t.
As I said, if Apple hadn't "glued these in" another manufacturer could fabricate their own memory to fit into Apples design, which as you stated contain off the shelf components.
The memory modules, using off the shelf parts can't be put onto another module sized for the mechanical dimensions and to the electrical spec? What is stopping others besides Apple licensing? This SoC contains separate fabricated components and you're sure the memory part can only be fabbed by TSMC? How about next year?
They've said the word latency.
Uh, all those computers were expandable. And yes, its called the difference between wishing and actually doing.
I wish the actual live London Philharmonic Orchestra would be my alarm sound in the morning, but out of market necessity, I have to settle for the Los Angeles Philharmonic every morning. See the differnce?
Steve still decided to make upgradable computers, because he realized Apple couldn’t anticipate every need, and that’s especially true for desktops. Remember what he said about computers like cars and trucks? Tim Cook Apple is making trucks with the back bed cut off, doesn’t matter how compelling the truck engine is, if it can’t haul anything, its a car. And people still want and need trucks.
Well yes, they were. And as I said the SoC labeling is a bit disingenuous, separate fabbed pieces, later stuck together. I believe the argument is now whether or not Apple could make the memory components work as well if they weren't "permanent". People have already swapped and upgraded memory on the M1, but they soldered it back in.
Yeah. Guess which one he was more interested in making? It wasn't the trucks.
If you want to argue with a literal chip engineer who used to work at AMD (not talking about myself) that they're using "SoC" incorrectly, enjoy.