The iPhone 14 is barely used and we are already talking about the iPhone 15's unmentionables. I have felt so polarized lately about whether I will upgrade or not. Macrumors on iPhone 14 doesn't look any different from Macrumors on an iPhone X. I am happy with the quality of the photos except for cases where I might be capturing images in a dimly lit setting.
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.
iPhone 15 Rumored to Still Feature Qualcomm 5G Modem as Apple Continues Work on Custom Chip
- Thread starter MacRumors
- Start date
- Sort by reaction score
given geekbench from a10-a14, it was very clear apple silicon was going to beat intel both on battery and performance, by what margin was the unknown factor.
I would want an Apple modem if it was as good or better than what Qualcomm would produce. To me cellular reception is more important than many of the other bells and whistles offered with the newer phones. I currently have an iPhone 11 with the intel modem and I see inferior performance reception wise than phones with the Qualcomm modem
At they rate they are going they should just scrap it and work on a 6G modem. 😂
I thought iPhones were suppose to have Apple's secret modems with the iPhone 12 or 13, if I recall from articles 3 or 4 years ago. Guess Apple is finding out, making them is much harder than they thought. I really wouldnt want to own a first gen model.
Apple spent a ton of money on Intel's modem division a few years back along with their patents in hope to build their own modems.
The gotcha is that Qualcomm has a ton of patents/intellectual property too.
Hence to build a modem that does not infringe on Qualcomm's patents and yet delivers as good results would obviously be the secret sauce.
The gotcha is that Qualcomm has a ton of patents/intellectual property too.
Hence to build a modem that does not infringe on Qualcomm's patents and yet delivers as good results would obviously be the secret sauce.
You can't build a modem that doesn't infringe Qualcomm's patents (or Nokia's/Ericsson's/Huawei's/Samsung's etc). They all have patented some aspects of 2g/3g/4g/5g standards and each of them will have to pay licensing fees to each other. When Intel was building modem chips for Apple, they also had to license stuff from those companies (and vice versa). Apple building their own modems by using Intel's tech doesn't really change anything in that regard; they'll still have to pay the licensing fees for all of those companies in order to comply with the standards.
However, building their own modems that are baked in to their SoC would be pretty nice since it would most likely lower the power usage significantly and thus provide a better battery life. Currently all iPhones are using a separate chip (Qualcomm) that is communicating with the SoC while other SoC's like Qualcomm's Snapdragon, Samsung's Exynos and MediaTek's Dimensity can just have the modem stuff built in.
I don't see Apple being able to to catch up to Qualcomm for 5G. Maybe for 6G when they both have a similar head start? On the other side Apple doesn't need to go for 5G, what if they first integrate 2G/3G/4G in their SoC and leave 5G for Qualcomm modem? That will still be good for iPhone users as you are not always on 5G, specially outside of cities. Having tighter control and better efficiency over power usage on 2G/3G/4G could save a lot of battery.
Why are you here if that's how you feel? The butterfly keyboard debacle is now history. Apple learned the lessons and evolved. Can you recognize that? 🙃
Providers are starting to shut down 2G and 3G networks entirely. Yes, some users will be affected, but Apple (usually) aims to skate to where the puck will be. Investing into older technology is not how you change the world.
Machine learning is a better term.
The basic point is that any modern spec (for WiFi, for cellular, even for codecs) has many many options. When I connect by cellular which frequency should I use? How many bits per symbol? How strong a FEC code? Which of the many many multi-antenna modes should I use? Etc etc.
One possibility in making these decisions is a set of heuristics that looks like an if/then tree (if signal strength >x and if bit error rate> y and ... then ...). But it's difficult to optimize that sort of decision process, especially when you start taking into account cross-connections (if the bit error rate is high AND the cell tower is using a high frequency AND the connection is more crowded than usual AND ...)
Machine learning has the possibility to help with this. Like other use cases, you give the model a large number of outcomes where the connection parameters were what they were, and the packet had a satisfactory (or not) particular outcome, and you hope that the machine learning builds a model for an optimal (or at least better than the alternative) choice of cellular connection type under all circumstances.
This is the same way (mostly) that machine learning (ie "AI") is used in the implementation of modern video encoders (including by Apple).
In the codec case, IN THEORY machine learning could also be used in other ways, for example to pre-segment the image before subsequent coding steps. It's unclear to me if anyone, so far, is actually using segmentation data in their encoding, but it's just a matter of time.
In Manhattan there is lots of tall buildings and subways and sub-basements (2 levels down in parking garages, etc) and upgrading every year on iPhones (currently iPhone 14 Pro Max) to get the latest modem is important. Every year places that I could not get signal are getting signal.
Also important in crowed places including conference, having latest cell modems.
Also, in NYC at least, use the Verizon high end plans. I do tethering so I get the 5G Get More which allows for 50GB on 5Guw (which uses band n77 == c-band).
Verizon is still expanding the usable amount of c-band by buying out satellite providers still using the channel early.
The X70 follow-on to be used in the iPhone 15 promises to be a big improvement in terms of signal reach and power consumption compared with the X65 which itself was a huge improvement over the X60 in the iPhone 13.
Also important in crowed places including conference, having latest cell modems.
Also, in NYC at least, use the Verizon high end plans. I do tethering so I get the 5G Get More which allows for 50GB on 5Guw (which uses band n77 == c-band).
Verizon is still expanding the usable amount of c-band by buying out satellite providers still using the channel early.
The X70 follow-on to be used in the iPhone 15 promises to be a big improvement in terms of signal reach and power consumption compared with the X65 which itself was a huge improvement over the X60 in the iPhone 13.
As I have said before, this is a foolish attitude.
Apple is certainly generating a large number of recent cellular patents, many of which are about ways of saving energy. (1% here, .5% there; each seem small, but they add up...) This makes sense insofar as radio energy is a substantial factor in the battery life of iPhones (and presumably the same tech and ideas will be used by Apple in subsequent BT and WiFi chips).
So one issue is that they're doing more than just "take Intel modem tech and move it onto an A-series chip".
A second issue is: how exactly do you want to implement the modem chip(s)?
One option is to place it on the A-series chip. That's "easy" but sub-optimal, because analog tech is not shrinking as rapidly as logic; you are paying a lot for N3 square millimeters that don't need to be in N3.
Alternatively you can have a separate chip placed on the PCB, like Apple does today. Problem is that uses depressingly large amounts of energy to communicate with the main A-series chip via PCB.
Optimal (probably) is to have the modem on a separate chiplet that's packaged with the A-series chip. But this involves a whole new set of design ideas and options. Do we place the chiplet on top of the A-series chip (like is done with A-series DRAM) or on the side (like is done with M-series DRAM)? How do we define the communications protocol between the two – use an existing (but sub-optimal) protocol or invent something new that can be used for future Apple designs (there will be more chiplets...) How do we split functionality between the modem chiplet and the main chip (for example should we have cache and a small CPU on the chiplet, or should most of the logic live on the A-series chip, and just analog circuits on the modem chip?)
Point is: Apple is defining here not just "let's move the Intel modem onto A-series". They are defining their RF-architecture for everything they make for the next five to ten years. Decisions made at this point will be what governs how everything from AirPods to Apple Watch to Mac Studio interact with their RF chips...
Better to get it right (and slip a year) than get it wrong!
These same claims of impregnable patent walls were made about CPUs and then about GPUs. In both cases they were nonsense. There are LOTS of ways to do things differently, especially if you're not locked into a design that you've been using for fifteen years.
As someone whose actually looked at a lot more of these patents than most people, I find it unlikely that "can't work around a patent, and can't license it" is at all relevant to their timetable.
Yeah - stop ruining our echo chamber with alternative points of view!
They make incredible silicon. I just compared an M2 MBA to a Galaxy Tab 8 Ultra for someone today. The silicon comparison was insane. Apple knows how to make chips.
I would imagine they are using a neural net for signal processing to better handle interference and for improved error correction to enable better throughout. But yeah, I've tired of the ai this ai that. The vast majority of it is not intelligent, and therefore should not be called ai, but I guess that boat already sailed in the marketing departments of companies.
Qualcomm chips are very good. But expecting Apple modems to consume less power.
It still puzzles me that people still choose to doubt Apple on these things. Considering that 2 of the years they have had to work on this were durning COVID shutdowns and remote work. Not an ideal environment for Apple’s model of development.
If/when Apple comes up with something it will be after a few years of chasing a fairly rapidly moving target so it had better be good. This report quotes Qualcom’s claim for the X70 of “up to 60% improved power efficiency”. There are those dreaded “up to” weasel words in there so no real idea of how it will perform in real life but the fact that Qualcomm calls attention to it does make me think that some quite impressive improvements in power efficiency might be coming down the line with the X70.
The 13PM was bad because it would never drop to LTE on Auto even when 5G was super slow or even nonexistent. I don't know if 14 is any better. The almost worse thing about the Z Flip 4 is it was actually great at auto toggling to LTE until the Sep-Oct update then it is just are bad as the 13PM. And you have to jump through hoops to manually toggle. Can't just do settings networks and change it. You have to get the band switcher app, and then tap more network settings, and then it is just picking the carrier and changing the default