Discussion Apple Silicon SoC thread

Doug S said:

Yes but Apple's SoCs only include things that can be economically justified. That is, they don't build M4 Pro chips and disable parts of the chip to operate it as an M4. Why? Because that would waste a lot of silicon. The non-analog portion of the modem may not occupy as much area as the difference between M4 and M4P but its enough that it doesn't make sense to add it for the low single digit potential market for cellular on a Mac. It's gonna remain a separate chip. Now if you could get me to believe that 50% of Mac buyers will want built in cellular you could make a case for building it in, but that's just not reality. People will use hotspots, and if they really want cellular you can buy a little USB stick barely the size of your thumbnail.

One may wish they could build an "everything" SoC but you can't mix digital and analog on leading edge processes, so there will always be separate chips for the analog world of PHYs and ADCs.

Click to expand...

mvprod123 said:

https://twitter.com/x/status/1894076780602151351

Click to expand...

Mopetar said:

I don't think they'd add a modem to anything but the lowest end M-series SoC if they did at all. There might be enough of a market for iPads and a MacBook with cellular connectivity, but there's no reason they couldn't use an A-series SoC for those products.

They will definitely make an A-series SoC with the modem integrated though. Every iPhone needs it anyway and the volume there is more than large enough to justify it. If they have the lowest power use, it makes sense for them to integrate it for their watches as well.

Click to expand...

SpudLobby said:

Well apparently they want to do an all-in-one Wifi/BT/Cellular chip long term but presumably in this case they'll still have a Wifi-BT standalone one for Macs and iPads unless the cost is just so low that it makes more sense to standardize at that point, but my suspicion is the RFE elements would make this a no go cost wise especially given most Mac users won't really care that much.

Click to expand...

Doug S said:

Why wouldn't they want to integrate that on the SoC? That gives it the best process so the best power efficiency. Every iPhone needs all three. (Just about) every Mac needs the last two. Obviously the RFE/analog piece has to be separate but those could be put onto one chip as well.

Click to expand...

name99 said:

Let's agree that by "the modem" we mean only the baseband logic.

Obviously there are various good technical reasons for probably eventually moving the baseband logic onto the SoC (along with adjacent ideas, like sharing the common pool of DRAM, rather than the C1 design of separate physical memory).
BUT on the economics side there is the issue of adding extra area that may (who knows...) mostly go unused in iPads, Apple TVs, MacBook Airs, and the various other A and M clients.

Click to expand...

Doug S said:

Yes there would be some wasted area in iPads and Apple TVs that use A series SoCs, but the former is offered with cellular as an option so they'd be able to make that option cheaper if it is built in anyway. That could pay for the "wasted" silicon area on models sold with only wifi by increasing the overall iPad ASP. For Apple TV it is truly wasted space but so is stuff like the ISP for the cameras, most of its I/O capability and heck probably half the memory controllers if we're being honest. Having a cellular modem block just gives them more crap they don't need they could theoretically bin on.

I think process cost is a bigger problem for Apple TV going forward than whether or not the SoC includes wasted area for a modem. They can follow the process train for a long time with high ASP iPhones, but Apple TV is at a lot lower price point (and honestly should target $99) so I'm not sure I see it ever being feasible to use chips made with A14 or whatever they'll be on by the time Apple starts integrating the modem.

I mean, what is the case for upgrading Apple TV beyond the A15 it uses today? You might want it to support newer wifi/BT standards but a new model could do that while still using A15. Apple Intelligence might push it towards A18 eventually. But it doesn't need support for anything more than 4K, and doesn't need better CPU/GPU performance unless they change course and start pushing it harder for gaming.

Click to expand...

name99 said:

What I see as the use case for a better chip in the aTV is "improving" the image on the fly
- upsampling and "sharpening" DVD/NTSC quality material
- stretching the contrast range to give HDR punch to SDR material
- temporal interpolating (correctly!) 25/50Hz material to 60Hz (and vice versa), and detecting the usual problem cases like backward wagon wheels and stuttering pans
- even colorizing B/W material

In THEORY amazing functionality is possible. Check out:

https://twitter.com/x/status/1887497602398073234

Can this be done in real time on aTV HW? Will it be possible in five years? Who knows?

I fear Apple TV will not get this because many Apple TV buyers are snobs who don't want anything to ever manipulate their precious dirty 1960s pixels. But hopefully other HW WILL get this, eventually forcing Apple to go along. Well, I have no control over any of that.

Click to expand...

okoroezenwa said:

Looks like the battery improvements are mainly from a meaningfully larger battery too compared to the 16.

Click to expand...

I’m actually surprised by this because I was hoping for battery life to be much longer given the efficiency of Apple’s new C1 modem. This plays a critical role in our testing because the phone’s display is set to 150 nits and runs a script that simulates normal web surfing over cellular until the battery’s depleted. Yet, the iPhone 16e's result comes up a little short.

Click to expand...

Nothingness said:

It indeed looks like the efficiency of C1 isn't what it's claimed. The linked Tom test says this:

Click to expand...

mvprod123 said:

A controversial statement

Click to expand...

Doug S said:

The fact that Apple's 1.0 modem is trading blows with Qualcomm's is the news here. Even if the power savings turn out to have been overhyped a bit, that result where the 16e is pretty much keeping up with the 16 in cellular speed on a train ride (which was a brilliant idea for a test - tests not only varying signal strengths but constant handover between cells, and possibly some no signal dropoffs in tunnels) is a major accomplishment. And if it is faster in areas of low 5G strength because it is quicker to switch to LTE then that shows it is already BETTER than Qualcomm's modem in that regard. You want to switch to whatever network is the fastest, not hold onto a weak 5G signal just because it is 5G.

I'm really impressed, if you read back a few posts I was pretty skeptical about Apple's first effort and was just expecting it to be "good enough not to look foolish" so to speak but not really competitive in terms of performance. I was wrong, it falls short in some ways but is superior in others so even if you had a choice between the two when you bought an iPhone it would be hard to make a compelling case to choose Qualcomm.

Now I'm sure someone will find a corner case or two with performance/handoff issues or some small carrier where things go haywire, and that will be overhyped because there are always people invested in generating clicks by telling negative stories about Apple. But their job in finding that bad news is going to be a lot harder than I would have guessed.

Click to expand...

Raqia said:

Would wait for more detailed reviews as it is as relatively difficult for review sites to test cellular subsystem performance compared to CPU/GPU performance as it is for companies implement (even a client-side accelerator for, never mind design) the 5G cellular networking standards. Nevermind the lack of mmWave (which is necessary as lower spectrum frequencies are simply running out of capacity):

1) This is not 1.0 (more like 2.5 or 3.0)...
2) It is unclear how carriers are configuring their cellular bands for this early release, lower performance model modem: it very likely not apples-to-apples...
3) The C1 early defaults to 4G whose towers have something like 5x the range as 5G and have worse network recovery characteristics:

4) Furthermore, Apple is intentionally making comparisons to just the specific modems which they use in other Apple models: notably NOT the Qualcomm X80 at 3nm but the X71 fabbed at Samsung 7nm which they intentionally picked for the iPhone 16 series instead of a faster model. They also have had a habit of throttling modems (which could have been run faster) in select product lines in the past to homogenize their supply chain diverse product line at the expense of product quality for an important cellphone feature.

It's a bit disappointing as an Apple customer and shareholder to see the fawning press tout edge-case advantages of an essentially inferior cellular subsystem which I think is diverting corporate resources and attention from its bread and butter client facing interface strengths such as rapidly aging features like Siri. Even with all their money, this seems like an obsessive holy war and a short-sighted, bean-counter initiated distraction for Apple in the end rather than an actual innovation or even a refinement.

Click to expand...

Raqia said:

This is not 1.0 (more like 2.5 or 3.0)...

Click to expand...

Raqia said:

It's a bit disappointing as an Apple customer and shareholder to see the fawning press tout edge-case advantages of an essentially inferior cellular subsystem which I think is diverting corporate resources and attention from its bread and butter client facing interface strengths such as rapidly aging features like Siri. Even with all their money, this seems like an obsessive holy war and a short-sighted, bean-counter initiated distraction for Apple in the end rather than an actual innovation or even a refinement.

Click to expand...

poke01 said:

This sounds like exactly like how I would expect Qualcolmm to respond and looks to me you are projecting your personal feelings instead of looking at facts. This is Apple's first modem and it will be Apple's worst modem and certainly is better than Intel's old modems.

Apple shouldn’t have to pay Qualcomm forever for it’s grossly overpriced modems.There is a reason why Google stopped using Qualcomm modems. Sure Apple won't compare to X80 because Apple doesn't use X80 modems, Geekerwan compared them though and Apple's modem is 20% more efficient,

Click to expand...

Raqia said:

Apple added value w/ their custom CPUs and feature oriented SoC design / tight OS integration on top of which they created a massively profitable store. The C1 project by Srouji's own admission was Apple's most complicated silicon effort ever, being years late, inferior to existing designs and invisible to users unless it breaks. It simply isn't what Apple does well or should do.

My point isn't at all that they should have to stick with Qualcomm, but that they should opt to add Mediatek (I believe the watch uses a Mediatek modem...) or Samsung to the mix rather than dump vast resources into a lagging solution which still uses an external RF solution. Unless they get in bed with infrastructure manufacturers and become a leading cellular IP house, they will still have to pay the system level IP licensing so they should continue to shop around for the best implementation; it's still very unclear that they will be differentiated in some way due to their own efforts.

Click to expand...

jdubs03 said:

They literally just spent the past five years working on the solution and it comes out as a pretty solid implementation for a first version.

Each subsequent version is going to an increased feature set and faster speeds. And since it’ll be fabbed on denser processes, the power draw will be even lower.

Seems like you’re being pretty unreasonable. This was a good thing for Apple and their own ecosystem.

Plus, it’s not like they don’t have the money.

Click to expand...

Eug said:

I don't follow this at all. It makes perfect sense to me that Apple would want to take this in-house. If I were Apple, I personally wouldn't want to rely on Mediatek either for the rest of my days.

And it's not just about product differentiation, but also cost and control.


Apple Vision Pro is the ultimate high risk and high cost but low volume product. Uber low volume in fact, and I'd be happy if it didn't exist at all, as it seems pointless to me. But it's good that they're willing to try new things.

Click to expand...