At their annual Social Media Summit this year, Qualcomm’s biggest bang wasn’t a chip, it was a module. SemiAccurate was impressed by the chips shown off but the most impactful thing by far was a throwaway line during the first keynote.
FTC Disclosure: Qualcomm sponsored SemiAccurate’s travel to the 2019 Tech Summit.
There were technically two slides on the new module program, one that showed the front of the phones with the 765 and 865 modules in them and another with the back of the phones with the modules in place. Since the modules were renderings, not to scale, and not placed accordingly, there was little info be learned there. Luckily we tracked down those who know the details and found some very interesting things.
Renderings of the modules in question
You might recall that the first Qualcomm 5G modem, the X50, had it’s mmWave antenna offered in a module, not a bare chip. This all in one device had the antenna, PMICs, RF Front End, passive components, and much more in a tiny module.
This blew most onlookers away, no one was expecting such a small package so quickly, the QTM052 module basically enabled mmWave to exist in a modern phone form factor. A few months later, Qualcomm followed it up with an even smaller QTM525 antenna module for the second gen X55 5G modem. Needless to say there are a lot of chips in there, not just the antenna, and the size is stunningly small for such a powerful feature set.
That is what Qualcomm is trying to do with the 765 and 865 modules, replicate that level of integration for the entire package and motherboard. Before we get in to the details there, lets talk a little about why Qualcomm would make this and who would want it. First off the modules are optional and will trail the launch of the 765 and 865 chips, no one has to go the integrated route if they don’t want to, Qualcomm will happily sell you the same things they always have should any OEM want to do things themselves.
That said one of the biggest problems with phones of late is packaging complexity, for example Ifixit’s teardowns of the iPhone 11 series shows they all use stacked PCBs. This is done to lower board space but it has two pretty severe downsides, cost and size. Yes we know we just said it lowers board space but that is in the X and Y dimensions, something that is done to increase the battery volume. It also has the effect of increasing the Z-height by making the phone thicker, something that also adds volume for the battery but runs counter to Apple’s prime objectives in phones. That would be making things thinner so the kids in the coffee shops where people wear white can post things about how awesome their new phone looks even if it is demonstrably worse to use and breaks more often. Basically thicker doesn’t make the fashionistas in Cupertino happy.
Back to the less snarky bits stacked PCBs also add cost. They are initially more expensive to make due to added materials, more work needed for thermal management, and most importantly significantly higher yield losses. There is no real up side to stacking PCBs from a financial perspective, you only do it if you need to for other reasons, in this case battery life. The lessons learned from Apple and others are to avoid this tech unless you really have no choice, but it does work in volume.
So getting back to Qualcomm and modules, there are three for the Snapdragon 865 and two for the 765, not counting the optional mmWave modules. For the 865 there is a CPU module, modem module, and radio/transceiver/RFFE module. The 765 integrates the modem in the CPU so it is missing the modem module. All of them combine the major components of the subsystem with all the passives and related bits, you only need to run a few wires for power and signal to the module itself.
Why is Qualcomm doing this now? In addition to the cost, complexity, and height of the alternative, PCB stacking, 5G brings an explosion in complexity. Even if you don’t use mmWave in your designs, sub-6 has a lot more to it than LTE in both component count and engineering/certification complexity.
With the modules, Qualcomm is squeezing every square mm out of the subsystem and trying to optimize everything possible for area. You then take these self-contained modules and mount them on a PCB and hopefully end up with a smaller motherboard. Before you ask, “If they are mounting it on a PCB, isn’t that board stacking?”, lets just say that it isn’t for two reasons. Stacking PCBs tends to imply components on both sides of a PCB which are then stacked together. If you don’t mount components on both sides you could do the same for less cost by mounting on both sides. The Qualcomm modules just mount on a bare PCB.
Since Qualcomm strongly implied that they were getting density that wasn’t possible with bare PCBs, we asked what the substrate for the module was. Was it a PCB that was then mounted to another PCB? Qualcomm would not comment on this other than to say it was not a PCB on the module. This leaves an interposer/silicon substrate or something less common than the standard organic package, both of which should enable the density step Qualcomm is implying while bringing thermal, reliability, and energy use benefits as well.
That said interposers and the like aren’t cheap but they are very likely balance the price out through yield benefits, device assembly/manufacturing costs, design engineering savings, and possibly time to market advantages. In short they will probably cost a fair bit more than the bare components themselves but save significant amounts in other ways. OEMs will undoubtedly do the math and decide on cost and differentiation, but we have no idea where the break point is.
Then there is the biggest bang of them all for the modules, that line that says, “Operator certified”. Qualcomm said there are two operators that have currently certified the 865 and 765 modules, Vodaphone and Verizon, one operator really if you think about it. That said it was strongly implied that there are many more that will be announced in the near future.
What this means is that the long, painful, and expensive task of certifying a device for a given carrier can be more or less skipped if you use the modules. The list of tests that need to be done are only those that the device itself affects, lets say 90% of the work is done before you send the carrier the first prototype. Since the cost of testing and certification ranges into the six or seven digit arena, that is per carrier mind you, and needs a lot of expensive and specialized engineering support, this pre-certification is worth a lot more than most people realize. Even if you just consider the engineering bandwidth that can be used for other things, this pre-certification seems to make the modules a no-brainer to SemiAccurate.
We will once again stress that the modules are optional and the possibility exists to mix and match what modules you buy and what you do yourself. Other than differentiation there seems to be little reason to do things the hard way other than price. Just being able to plunk the modules down on a PCB, run a scant few wires between them, and off you go lets an OEM spin devices in a short period of time. That said if your volumes are high enough, it may be worth it to go your own route.
There is one other thing that modules bring to the market and if Qualcomm plays it’s cards right, it may mean modules explodes their market share.
Note: The following is analysis for professional level subscribers only.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- A big ARM server project shut down in silence - Mar 14, 2023
- A bit more on AMD’s Genoa memory issues - Mar 13, 2023
- Want to know when Qualcomm’s Oryon taped out? - Feb 16, 2023
- Things go bump in the night for AMD’s Genoa - Feb 15, 2023
- AMD’s Instinct MI300 is much more than a GPU - Jan 6, 2023