but apple has no prior experience with modems, and qualcomm is not standing still either, they have more shares of essential 5g patents than intel.
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.
Apple Reportedly Wants to Have a Custom 5G Modem Ready for Use in Some Products by 2021
- Thread starter MacRumors
- Start date
- Sort by reaction score
It's a reasonable question BUT it assumes facts not in evidence. In particular it assumes that Intel's diffi
That's one small part of it, but there are multiple ways Intel probably screwed up.
- The division (like most of Intel these days) seems to be have been run atrociously badly with personal vendettas, micromanagement by unqualified managers, constant changes of plan, and similar stupidity. You get this impression by reading sites like TheLayoff where employees can vent about their companies anonymously.
So simply setting a single goal and sticking to it is one way management can be a lot more productive...
MY guess is that the division will be essentially destroyed and rebuilt, with totally different reporting and responsibility structure, and with management having to justify their retention. This sort of rebuilding is generally necessary to get rid of pre-existing toxic relationships. (The sort of thing where "algorithms" just hates "hardware's" guts, so both refuse to ever talk to each, and you never get both sides sitting down to say "you know, if we added this feature, you could remove that stage from your loop and everything would go 15% faster".)
- Apple is always willing to spend more hardware to do a better job. INTC appears obsessed with the bottom line, so was likely continually scrounging to try to make the modem smaller (less area) even though that's an idiotic constraint given that they have one customer who matters --- and that customer doesn't care!
But companies get stuck in a way of doing things --- we saw the same thing with Imagination which kept its GPUs too small and timid, to save area, even though same issue --- one customer who mattered, and they wanted more performance rather than small area...
- Intel was locked into doing things the Intel way.
Like they insisted on using an x86 core to run the baseband in their current Intel modem Apple uses, and were likely using an Intel DSP as well. Needless to say, these are NOT optimal choices along any dimension (performance, power, area).
Apple will probably kill this nonsense, with Johnny Srouji talking to their top engineers TODAY, describing the cores Apple has in-house (along with what ARM can sell them) and a SENSIBLE choice being made of core+DSP.
- Likewise of course, this was locked into Intel process. Presumably the mythical 10nm process that one day, real soon now, promise promise promise, will produce the greatestest fastestest densestest most awesomestest chips the world have ever seen.
Apple will, likewise, shut down that nonsense and retarget (most likely TSMC but, who knows, maybe SS, maybe some of the RF front-end on GF 12FDX?)
[doublepost=1564160159][/doublepost]
Apple has multiple years of experience with both WiFi (up to W3 chip by now) and BT (W3 and H1 chips), and has had at least some modem engineers employed for years. It's massively ignorant to claim they have no knowledge of either the RF or modem (ie baseband/algorithm) side of cellular tech. EM is the same, propagation is mostly the same (reflections are worse for cellular but rake receivers are old tech everyone understands).
Even the precise details (OFDM modulation and now, with 802.11ax, OFDMA access) are converging...
And ESSENTIAL 5G patents are meaningless. That very word, ESSENTIAL, means Apple gets to use them!
And whatever games QC might have wanted to play around patents this time last year, since they've been slapped down aggressively by Lucy Koh, presumably those ESSENTIAL patents will be available to Apple (and anyone else) at a reasonable fee and under no ridiculous restrictions.
These days many companies buy expertise rather than create it from scratch in house. Apple has deep pockets and there are plenty of engineers available for hire. And if you were an potential employee whose stock options would you want to own, Apple's, Qualcom's, or Intel's.
you mean the chips in my airpods that i have to forget and reconnect once a week? oh boy
[doublepost=1564160910][/doublepost]
ah right forgot about the pooching aspect.
Agreed. I've always thought Qualcomm made superior radios. Have always been happy having Qualcomm in my iPhones rather than Intel. Besides 2021 is far far away!
No, that's not going to happen. Apple has a licensing agreement with Qualcomm to use their technology.
This was an absolute steal for Apple. Great news if they can make it start to pay off. this quickly. Signature Tim Cook doing what he does best.
Yep, 5G requires more towers. We, in the US, already have too few towers for good consistent coverage. If the transition to 5G happens fast it can only mean that service will suffer even more as the new towers are built 5G only.
It's hardly a league, I mean, we don't even have matching t-shirts.
Depends. Some carrier had theirs upgraded already and only requires Software update. ( This is not referring to the 5G mmWave, but 5G in Sub 6Ghz Frequency )
Depends, in terms of Sub 6Ghz and NR, those are easier and most of them are prepping for the 3.x Ghz. But If you are looking at mmWave 5G, I really don't know and as I have said before I have always doubt how it would work. And so far I haven't seen anything that is convincing.
On the subject of Modem.
It is interesting GUC are not really know for Mobile / Modem Design work. ( GUC and Mediatek are quite well known in Taiwan ) While I have said I don't see how Apple could ship an Apple Modem in 2021, I would love to be proven wrong on this one.
1. Infineon was very slow on the uptake of LTE (and the CDMA aspects. The protocol not the headset standard). It isn't just Intel but the ramp positioning they this Infineon group came in on. Intel bought Infineon just as the largest customer was lost ( apple going to Qualcomm because had one chip for one world solution and LTE. Infineon didn't).
2. TSMC or some Fab contractor. TSMC if go back far enough to where wasn't using Intel fabs there were other choices. The choices being the issue. ( if go back 6+ years TSMC wasn't necessarily the best option for cellular modems. Apple wasn't using them either. )
3. Did Infineon have much of an ARM chip in their modem anyway? They really didn't have much of a smart adaptive modem. If talking about the some app processor + modem combo they didn't have much traction there at all in smartphone space (at least in the post iPhone introduction notion of 'smartphone' ) .
x86 cores weaved into the modem itself or the modem die coupled to a x86 inside of package? In the latter case it is more so that is the only primary customer (consumer) of the model . That was a substantive problem from day one after the acquisition.
The discrete cellular modem market in smartphone space dried up pretty quickly (all the more so when trailing on LTE). Putting the modems in the same package at Atom x86 didn't work so well. Intel fabs only hit a brick wall until got to 10nm. Intel hadn't borked that lead then
What will help with Apple is that they will probably be shifted to making a modem that is only 'good' inside an iPhone ( and not much else. ). Or short term shifted to making a data only modem (no legacy voice) and only going into watches/iPad.
The bigger fail at Intel was that the rest of Intel could be the biggest "customer" for the Infineon modems. that is substantive easier at Apple because they only have 1-2 CPU products to couple to versus Intel's shotgun approach to products. Plus Apple also has the volume to support it at high run rate levels. Intel never really had that for the combos with modems.
you mean the chips in my airpods that i have to forget and reconnect once a week? oh boy
[doublepost=1564160910][/doublepost]
ah right forgot about the pooching aspect.
totally agree.
intel results of its work so far on its modems (that same work, patent base, and people) that apple is trying to leverage) + apple repeated failure at releasing advanced tech that is reliable in 1st gen products = problems in apple's in-house attempt to overcome Qualcomm with this 1st gen 5G tech apple modem
management can mean that have a different set of tools to choose from. Intel made a decision to use the modem as part of filling up their fabs with product. That was a dual edged sword in that if Intel's fabs fumbled then the modem would also. 5-6 years ago Intel was largely out in front of everyone else. That might have worked if they hadn't gotten very arrogant and come up with a change 5-6 major things at the same time and shrink plan ( the 10nm debacle).
If the fab process had gone sideways on Apple's management they'd have quit and found another option. If the modem needs a substantively different fab recipe than the CPU/GPU part more than likely Apple would follow separate tracks. Intel seemed to be more so in a mode where trying to make the modem 'eat' a fab process tweaked specifically for their CPUs. (even though Intel said they wanted to get into the "fab for 2nd parties" business that really rolled all that well at all. Even their higher margin Altera FPGA effort eventually turned in a 'buy-out' to make it work. )
One, I don't think Apple is looking to beat Qualcomm, only be more affordable and aligned with Apple's schedule. Because Qualcomm has to work for lots of different folks Apple may never be better than them because there are probably going to be contexts where Qualcomm looks and works on that Apple doesn't. [ As long as Qualcomm has the money to cover everything early and well they'll probably be better in more edge case situations. ]
Intel's modem division in a bit of a catch-22. If they don't have a big (high volume, wide distribution ) customer to get interoperation exposure to. But without very good interop they couldn't land a major volume discrete customer in smartphone space.
My guess is that Apple is going to find the flaws in the current Intel modems and then as step 1 just simply move them over to another process ( TSMC as likely can get bigger bulk buyer discounts and max out on tricks have learned along the way with that specific fab process family. ). Then they'll just iteration. It isn't something Intel couldn't do. It is probably more so just more money than Intel wanted to spend (given Apple could flake out on them at any next cycle of design bake offs while demanding low margins on the modems. Apple scrooge mc Duck they're own process just puts more money in Apple's pocket. At least until a major recall and warranty extension. ).
I think it is telling that Intel isn't dumping the modem business outside of the relatively small battery , ultra low power consumption cell phone modem business. It is more so smartphone modems they are having most problems with rather than data modems that might even be plugged in most of the time ( on other side of the link from the phone in the field).
A healthy smartphone business. Apple doesn't have to build volume for a consumer for these chips. What Apple is going to do is switch suppliers.
If iOS cellular modem devices sales start to crater substantially in 5-6 years Apple will then be in a very similar boat that Intel is current in now. ( only there probably won't be any "buyers" left for the modem business if want to try to sell it off. Just a couple $100M hole in the ground. ).
Remember though as Qualcomm complained about at one of the trails they had with Apple was that Apple will be getting all the newer Qualcomm modems to test with their headset designs and indirectly passing info along to Intel to fold into those trailing modems. All the feedback , errors, workarounds, and new testing plans that come out in getting the iPhone/Qualcomm phones to market can be directly applied to another variant with an Intel/Apple modem in it. Apple can just keep Intel on track now with their 'late' 5G' solution and just continue to improve it until acquistion is complete. That would be gen 1 . Doesn't ship in volume but all the 'mistakes made'.
They take that and port it over to another fab process and get gen 1.5 ( limited changes and just new fab. ) and done. That to is only on subset of iOS modem devices so get another round of gen 3 Qualcomm stuff to do blackbox version A versus version B. testing with. If one of the major hold ups is Intel modem power consumption then that could be uncorked largely with the fab change.
Even with proper "firewalls" between projects some info will leak across. Where Apple's modem fails to perform will be a fail on the "fix it' list even if there is no "exact' fix specified in the bug. Just working with Qualcomm's gen 3 will allow Apple to come up with better defined requirements for Apple's gen 2 .
this.
really like your thought process.
you know the tech side, the engineering side, as well as the development side.
thanks.
A good point. What would you suggest?
Apple Defense Tea Party?
Apple Defense Mob?
Rabble?
Apple yesterday announced that it has agreed to acquire the majority of Intel's smartphone modem business. The $1 billion transaction is expected to close in the fourth quarter of 2019, subject to regulatory approvals.
Understandably, the acquisition may enable Apple to accelerate development of its own 5G modem, with Reuters citing a source who claims the iPhone maker wants to have an in-house chip ready for use in some of its products by 2021, compared to previously reported timeframes of between 2022 and 2025.
Apple's transition to custom 5G modems will likely happen in phases, starting with lower-end and older models of devices, according to the report. Apple has a multiyear chipset supply agreement with Qualcomm, and a six-year patent license agreement, so it certainly does not have to rush the process.
The report does not explicitly mention the iPhone, so the first product with an Apple-designed modem could very well end up being an iPad. In any case, the transition away from Qualcomm will likely take years, as its modems lead the industry in performance and worldwide compatibility.
In the interim, Intel is expected to supply LTE modems for 2019 iPhones, with Apple returning to Qualcomm for the first 5G-enabled iPhones in 2020.
Article Link: Apple Reportedly Wants to Have a Custom 5G Modem Ready for Use in Some Products by 2021
No signal
Apple also invested a ton of money in equipment for sapphire that the CEO of that company lied about progress
Sounds like the argument against Apple making processors, or even phones in the first place. Remove the complexities of having to be compatible with thousands of devices made by hundreds of companies and the results can be astounding.
Last edited:
I am guessing is that:
1-Apple hopes it reaches or surpass qualcomm levels. Maybe future chips with integrated GPU-Modem that saves battery life.
2-Apple expand business by being modem supplier themselves
3-Apple has leverage in negotiation modem prices in the future, they can always rely back on their own
4-If you are the only customer, I figure if they bought the business and do the numbers, they will have higher profits per device since they don't pay higher prices for modems.
If they just shave off $3 off each modem price, and they sell 50M devices... thats $150M dollars right there. Now imagine if they shaved off $10.
