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

MacRumors

macrumors bot
Original poster
Apr 12, 2001
67,493
37,779


Update 6:44 p.m.: Samsung has informed MacRumors that the original report from The Elec is "completely incorrect" and that the "details are false." While The Elec has pulled its report entirely, we are leaving our original article intact below for reference to ensure clarity around the situation.



Apple is reportedly planning a significant shift in its iPhone hardware design by transitioning to discrete memory packaging to enhance on-device AI performance.

apple-silicon-feature-joeblue.jpg

Samsung, a key supplier of Apple's memory components, has begun research to accommodate the change at Apple's request, according to a new report from Korea's The Elec. The shift will mark a departure from the current package-on-package (PoP) method, where the low-power double data rate (LPDDR) DRAM is stacked directly on the System-on-Chip (SoC). Starting in 2026, the DRAM will instead be packaged separately from the SoC, which should significantly improve memory bandwidth and enhance the iPhone's AI capabilities.

The current PoP configuration was first introduced in the iPhone 4 in 2010 and favored since then for its compact design. Stacking the memory directly atop the SoC minimizes the physical footprint, which is especially important for mobile devices where space is at a premium. However, PoP packaging imposes constraints that limit its suitability for AI applications, which require faster data transfer rates and more memory bandwidth.

With PoP, the size of the memory package is constrained by the size of the SoC, capping the number of I/O pins and therefore limiting performance. Moving to discrete packaging will allow the memory to be physically separated from the SoC, enabling the addition of more I/O pins. This design change should increase the data transfer rate and the number of parallel data channels. Separating the memory from the SoC also provides better heat dissipation.

Apple has previously used discrete memory packaging across the Mac and iPad product lines but later shifted to memory-on-package (MOP) with the introduction of the M1 chip. MOP shortens the distance between the memory and the SoC, reducing latency and improving power efficiency. For the iPhone, adopting discrete packaging may necessitate other design changes, such as shrinking the SoC or battery to create additional space for the memory component. It may also use more power and increase latency.

In addition, Samsung is reportedly working on next-generation LPDDR6 memory technology for Apple, which is expected to offer two to three times the data transfer speed and bandwidth of the current LPDDR5X. One variant under development, LPDDR6-PIM (Processor-in-Memory), integrates processing capabilities directly into the memory. Samsung is said to be collaborating with SK Hynix to standardize this technology.

The changes could appear beginning with 2026's "iPhone 18" devices, providing Apple can overcome engineering challenges related to SoC miniaturization and internal layout optimization.

Article Link: Apple Looking to Fundamentally Change iPhone Memory Design to Enhance AI Performance [Update: False]
 
Last edited:
Basically, Apple is re-inventing PCs. This could be the start of user swappable RAM, SSD and potentially discrete GPUs.

Imagine a Macbook with the base GPU and the option for Apple discrete GPU

For ex [M5 Pro 18-core GPU w/M5 Max 40-core discrete GPU]
 
Basically, Apple is re-inventing PCs. This could be the start of user swappable RAM, SSD and potentially discrete GPUs.

Imagine a Macbook with the base GPU and the option for Apple discrete GPU

For ex [M5 Pro 18-core GPU w/M5 Max 40-core discrete GPU]
We shall see about that, but I’m interested to see what comes from this. I don’t have high hopes for user upgrades though.

Meanwhile, i’m very much looking forward to swapping my RTX 4090 for a 5090 in January ;)
 
Basically, Apple is re-inventing PCs. This could be the start of user swappable RAM, SSD and potentially discrete GPUs.

Imagine a Macbook with the base GPU and the option for Apple discrete GPU

For ex [M5 Pro 18-core GPU w/M5 Max 40-core discrete GPU]
Do you really think that is what will happen? :)
Sounds good to me (modular computers where components can be upgraded should be able to last longer), but I don't know it this is something Apple aims for. But I'll be happy to be wrong.
 
Doesn't packaged memory offer significantly reduced latency though?
Yes but the gain from reduced latency may no longer worth the reduced capacity and bandwidth for AI purposes. We all know Apple raises the minimum ram to 8GB on many of their devices just for effectively running the first gen AI. Would be a struggle for more higher bandwidth RAM on iPhone especially later down the line.
 
Last edited:
This seems like pretty bad move for consumers (good for apple as this would be a lot cheaper to produce)
Umm it may not be a bad move depending on how things pan out. Memory latency is just one of many aspects determining the performance target for memory. It needs to be combined with other factors for optimal result.
 
Doesn't packaged memory offer significantly reduced latency though?
Yes, if true this is almost certainly intended to cut costs due to the amount of memory being used now or projected for the future, not for performance reasons.

It may also be a test run for future M series because they’re having trouble getting high enough memory density within the SoC package and are exploring engineering options that somewhat mitigate the tradeoffs.

edit: re-reading this it might be the other way around and they’re folding some of the Memory on Package stuff back into the A series silicon. This article is convoluted.
 
Last edited:
  • Like
Reactions: navaira and mevans7
Basically, Apple is re-inventing PCs. This could be the start of user swappable RAM, SSD and potentially discrete GPUs.

Imagine a Macbook with the base GPU and the option for Apple discrete GPU

For ex [M5 Pro 18-core GPU w/M5 Max 40-core discrete GPU]
This is nuts because I’ve been saying I want Apple to do this on Macs so we can upgrade our RAM again and Apple’s line has been that RAM in the SOC is way more performant

Now I’m hearing it’s not!!?? I feel like I’m taking crazy pills
 
Basically, Apple is re-inventing PCs. This could be the start of user swappable RAM, SSD and potentially discrete GPUs.

Imagine a Macbook with the base GPU and the option for Apple discrete GPU

For ex [M5 Pro 18-core GPU w/M5 Max 40-core discrete GPU]
I don't think we'll ever be able to swap components in iPhone.
Even for a Mac, right now, you'd have to swap the entire SoC.
If I could swap the entire SoC in a Mac, it would be pretty expensive, but I'd be happy to. But I don't think they'll do that. Even if the design stays the same, sometimes they'll improve cooling methods (i.e. the fan).

And as per the article, I understood before that an SoC gave the ability to have great memory bandwidth, because the memory is closer to the CPU/GPU. I'm not sure how another solution can improve memory bandwidth even more ?
 
Last edited:
Yes, if true this is almost certainly intended to cut costs due to the amount of memory being used now or projected for the future, not for performance reasons.

It may also be a test run for future M series because they’re having trouble getting high enough memory density within the SoC package and are exploring engineering options that somewhat mitigate the tradeoffs.
That's what I initially thought - I mean memory is packaged even on the highest-end Apple chips at this time with some pretty serious memory bandwidth and frequency. So I think this is a camouflaged reason to cut costs.
 
  • Like
Reactions: ksec
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.