In my stint as an analyst, as well as in my long career as an engineer and a marketer in large corporations, I have developed a believability yardstick: believe less than 50% of what you are told, about 60% of what you are shown, and almost 100% of what you can hold in your hand and experience yourself. This hasn’t failed me so far, and has been a valuable tool when analyzing all the recent 5G news, including Intel’s latest announcement about XMM 8160 5G-NR modem.
Until Nov. 12, 2018, there was a strong, well built, and much publicized aura of awesomeness around Intel’s first attempt at 5G, the XMM 8060, only to evaporate with the news of shiny new but yet elusive XMM 8160. Let’s look at what happened, why it happened, and what it means to Intel’s sole smartphone OEM, Apple.
Intel announced their XMM 8060 exactly one year ago, in Nov 2017, and played their part in the busy 5G news cycle by announcing successful progress of their technology development. In Aug 2018, they announced inter-op with Ericsson over 5G-NR mmWave link. The picture that accompanied the press release really piqued my interest. The 5G user device in that picture was the size of a fridge with racks and cabling all over. Having worked in tech for long, to me, their set-up was akin to a proof of concept design, where you use racks of generic hardware and rudimentary “lab-grade” software to prove that your technology works, a far cry from commercial form-factors. Usually it takes two to three years, if not more, to reduce the size from a fridge to a small chip that goes in to a device that one could hold in their hand. So, as late as August, 2018, the writing on the wall was that it was impossible to have a portable device based on Intel 5G modem before 2020, if not later.
The recent revelation of not commercializing XMM 8060 confirmed the fears that Intel was behind competition in 5G. Obviously the “fridge” that was shown in August 2018, was far from a productizable solution. Presumably Intel fabed XMM 8060 silicon, but it was not clear whether these inter-op tests used that actual silicon or not. In such case, announcing the new XMM 8160 solution with a claimed shortened timeline was a clever attempt at putting a positive spin on the state of affairs. To compensate for the delay, they seem to have included the multi-mode support with a single chip solution embellishment. Applying my believability yard stick, I will not believe any of it unless I see something in real flesh and blood (Silicon in this case) in real form-factor devices. Intel’s consistent failure to deliver a competitive solution, in a timely manner, going back to 3G, 4G, and Gigabit LTE makes it harder to give them the benefit of doubt. There was some hope that 5G being a new generation, Intel could leapfrog. But alas, it is not to be.
One could point out any number of reasons why Intel fumbled on its first 5G attempt. I am sure Intel is evaluating it very closely as well. I could identify at least two possible flaws that might have contributed to it. First, in the rush to get early lead on 5G, Intel invested too heavily on 5GTF, the pre-standard version of 5G championed by Verizon. When, to their and many other’s surprise, 5G-NR timeline was accelerated, they had to pivot and realign their strategy. For a behemoth like Intel, such pivot definitely costs crucial development time. The other reason is more fundamental and a chronic challenge, especially when solutions for new generations of technology are being built. Unlike every other major modem player, Intel seems to lack their own smartphone test platform, which incorporates all the components of a real smartphone, including AP, RF Front End (RFFE), Antenna Subsystem, Power Management and other subsystems, in a smartphone form-factor. I am saying this because they have never shown such a platform publically, and also they no longer sell APs and other subsystem solutions for smartphones. Having such a platform simplifies and substantially expedites the development of modem solutions. Also it allows quick and efficient optimization of modem performance to fit it in the challenging power and heat envelopes, as well as the shape, size and other form-factor constraints of a smartphone. Without such platform, you have to jostle with test instruments, simulators etc., which is a big handicap.
So, what does this cancellation of XMM 8060 and advancement of XMM 8160 mean to Apple? Well, Thanks to Apple’s secrecy, we don’t know. We can only speculate and make educated guesses. Did they have early 5G iPhone plans? Were those plans shattered by Intel’s inability to productize XMM 8060? Even with XMM 8160, Apple will be faced with much constrained and crushed schedule to integrate it into their iPhones. It is worth noting that from modem perspective, designing a 5G smartphone is a far different and much complex endeavor than previous generations, mainly because of mmWave spectrum. You can read more on why and how in my article here.
The two most important things worth highlighting regarding mmWave are—absolute need for tight integration between modem and RFFE, and proper placement of antennas. The complex 5G techniques like beamforming, beamtracking and beamsteering (as well as massive MIMO for sub 6-GHz) require very close communication and deep integration between the modem, RFFE and antennas. This integration essentially determines the mmWave performance of the smartphone. Unfortunately, there are no standard interfaces or APIs to accomplish this. The interfaces are proprietary when both the parts are coming from the same vendor. Additionally, the location of the antennas plays a crucial rule in the performance as well, which needs deep expertise to perfect it. I think because of these reasons, the other merchant modem provider, Qualcomm, is providing not only Snapdragon X50 modem and RF modules but also smartphone reference designs to their OEMs. These take the modem complexity away and help OEMs to focus on other aspects of their smartphone design. This is quite evident from Qualcomm’s many demonstrations of form-factor test devices, as well as definitive announcements from a large number of their OEMs about impending launch of commercial devices in 2019. Since other leading vendors such as Samsung and Huawei design their own phones, it is not a major issue for them.
Additionally, Intel is only providing RFICs and note full RF modules, along with their modems. This puts the onus on Apple or any other OEM to integrate them with third-party RFFE providers such as Qorvo, Broadcom and others. Apple indeed has the wherewithal to accomplish it. But they will probably need more time than supporting traditional cellular bands. Also, Intel has to work very closely with these third-party RFFE vendors and might even have to open APIs for perfect integration.
I recently wrote an article delineating my reasons, including these, on why iPhone 5G will not appear before 2020. One might wonder, did Apple ever think of defying their tradition, and adopting 5G sooner, and being among the first? That is not a far-fetched idea, if you consider the global awareness and traction of 5G, and the excitement it could have brought to their uninspiring iPhone portfolio. If so, did their legal tussle with Qualcomm, and most importantly, inability of Intel, their sole modem supplier, to deliver on XMM 8060 play a role in them giving 5G a pass? Again, with Apple’s penchant for secrecy, we will probably never know.
