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.
interview Apple VP platform architecture
- Thread starter themichiel
- Start date
- Sort by reaction score
At around 24 minutes or so they talk about integration of non ARM applications, at 25 minutes or so they rifle off a bunch of application titles saying they work great. World of Warcraft, which has a couple of active threads running questioning compatibility, was one of the mentioned titles. Watching this basically sold me on ordering one (13” MBP). It’s a good watch.
A definite improvement... but I still shake my head at why it’s 720p to begin with.. and I’ve been shaking my head about that for years now.
I presume it's because they just want to use all the same old parts until they update the cameras to accommodate Face ID.
I reckon, it’s just funny how they can be blazing new paths here, and straggling so far behind there. Personally I could care less, it looks fine and if you REALLY need HD quality there are plenty of 3rd party options. I just get a chuckle from it.
Very interesting video for multiple reasons 
Those guys went through some impressive training in talking to journalists.
Those guys went through some impressive training in talking to journalists.
No, I think this is the first mention of it. It will be interesting to see how the next batch of Apple Silicon Macs will handle larger memory. There's only so much room within the Apple SoC package.
Anandtech writers mentioned it on Twitter before. As to next gen Macs, they will use more RAM chips, possibly stacking them, while using a wider interface.
DDR4 is not higher-performing, it has lower transfer rate. Assuming Apple uses LPDDR4X 4267 with a 128-bit interface, it has about 60% more bandwidth than the DDR4 2666 used in the 16" MBP
I inferred higher performance for DDR4 than LPDDR4X from the fact that DDR4 consumes more power, and yet Apple uses the latter in both the 16" MBP and the iMac. Logically, if DDR4 offers reduced performance while consuming more power, why would Apple use it, rather than LPDDR4X, in those devices?
DDR4 has higher densities (this is what allows 64GB on the 16"), not to mention that if I remember correctly 14-nm Intel CPUs don't support LPDDR4X. And of course, cost and availability.
With Apple moving to on-package high-bandwidth memory configurations, I am curios to know where are they going to get all those premium RAM chips...
They may move to off chip RAM on the mid-range and high-end SoCs, and make up with the performance loss with even wider data busses. They obviously can't stack 256GB and higher RAM capacities for the higher end machines. And while they will probably end up using denser and denser RAM chips when they are avaiallable, those are on longer development cycles than the AS SoCs appear to be on.
720p streaming video uses relatively little bandwidth. This is particularly important on slower Internet connections. That's why 720p has been the standard for years. While you might have a fat 5G pipe today, it wasn't the case for 99% of users a few years ago.
Note that for canned video conferences, while the actual interaction might have taken place over a 720p service like Zoom, it's possible that all parties recorded high quality 4K video on more robust hardware and uploaded those files to be edited by the video publisher.
Apple executives have participated in other such interviews. That's why Apple's recent media events look great. They're not live Zoom conferences. They are likely using some of the best 4K video and audio hardware. Tim Cook is not standing in front of a MacBook Air's FaceTime camera.
The RAM in the M1 SoC package is already on a separate die:
Apple Silicon M1 Chip in MacBook Air Outperforms High-End 16-Inch MacBook Pro
No, you are wrong. In all of Apple’s A-series processors, and in the die photo that has been circulating of the M1, there is no RAM on the SoC die. The RAM is on separate die, in the same SoC package. There are multiple reasons for this: 1) so you don’t have to make a separate chip for each...
forums.macrumors.com
Last edited:
While that is likely the case, upgrading RAM in a stacked SoC package isn't realistically something that the end user can do and may likely be beyond what an Apple authorized service provider can accomplish.
A socketed SoC might allow for an upgrade path but my guess is that the SoC on the new Apple Silicon Macs are soldered to the motherboard effectively making this discussion moot.
Apple has pushed an increasing importance on system integrity and security. With the on-die Secure Enclave, it is unrealistic to think that Apple would use a chip socket for the M-series SoC.
A socketed SoC might allow for an upgrade path but my guess is that the SoC on the new Apple Silicon Macs are soldered to the motherboard effectively making this discussion moot.
Apple has pushed an increasing importance on system integrity and security. With the on-die Secure Enclave, it is unrealistic to think that Apple would use a chip socket for the M-series SoC.
Yeah, you're correct on the compatibility. I looked it up—the chips in the 16" MBP and iMac aren't compatible with LPDDR4X
Don't know about the cost, but Apple could have addressed the memory density (at least to the point of doubling it)—and also gotten both higher performance (5,500Mb/s) and more energy efficiency—by going with Samsung's 16 GB LPDDR5 chips (https://www.androidcentral.com/samsung-first-begin-mass-production-16gb-lpddr5-dram).
Apple certainly had enough lead time to incoporate these into the first AS Macs: Samsung publicly announced them in 2018 (so Apple probably knew privately before this), and they went into mass production in February 2020.
Also, according to https://www.econotimes.com/iPad-Air...00-starting-price-difference-worth-it-1592484, the recently-released A14-based iPad Air has LPDDR5. Though I don't know if this is accurate; I haven't been able to find confirmation on other sites.
Anyways, given all the seeming advantages, why didn't Apple use LPDDR5 RAM in its first-gen AS Macs? My best guess is cost and product positioning; based on both product pricing and the number of I/O ports, both the M1 Mini and M1 MBP were positioned as lower-end models (and the MBA is inherently entry-level, at least for Apple). So I suppose not using cutting-edge mobile memory is consistent with that. Alternately, perhaps there is some technological reason—e.g, maybe it would have been difficult for Apple to design the M1 to take advantage of the increased bandwidth. But the latter is just pure speculation. I would guess that LPDDR5 will be in the next generation of mobile AS Macs.
Last edited:
This is a simple question that I can answer with some confidence. There is simply not enough LPDDR5 being produced out there, that’s all. Samsung (and maybe others, don’t know) make a limited amount of those chips for them to be put in some a droid phones, but Apple ships their devices out in such a large quantity that there is no chance to satisfy their demand with current production capabilities. This was explicitly mentioned by various people with sources in the industry, including Anandtech
P.S. I think it’s important to remember that these supply constraints can exist. We are focusing a lot on tech that is available and not enough on how much that tech is actually available. By the way, similar stuff can happen to M1. There are already reports that TSMC won’t be able to supply enough chips if sales go well.
Last edited:
More chips. If you look at the M1 is has two chips on the right next to the SoC. They probably did this for small size and low power. They could theoretically have 4x as much memory by just putting it on all 4 sides. DDR5 is on the horizon which will increase capacity per chip too.
Can you stack DDR RAM? I know you can stack HBM but I wasn’t sure about other types of memory. If you look at a GDDR GPU it’s surrounded by a lot of chips compared to an HBM GPU which has two or two small stacks of memory. If so that will free up a lot of space.
I really doubt this is true. I’m not sure why they would waste silicon space supporting two memory standards. You often see this on PC chips during memory transitions but that makes sense because buyers might still want to use older memory because it’s cheaper. It doesn’t make sense in a small highly integrated SoC for a phone as Apple controls the whole thing and knows exactly what they need.
Last edited: