Qualcomm's decision to remove integrated modem from Snapdragon 865 SoC is puzzling

Shawn Knight

TechSpot Staff
Staff member

As we mentioned in our initial coverage and as Ars Technica recently expanded on, Qualcomm’s Snapdragon 865 SoC doesn’t feature an integrated 5G modem chip in the main package. In fact, there is no onboard modem at all. Instead, an extra chip is needed for 5G and 4G operation of any kind.

Why is this a problem, you ask?

Allow Qualcomm to explain. In this 2012 press release on the benefits of integrating a 4G LTE modem into its Snapdragon S4 processor, the company said, “typically, the more chips that are involved in building a device, the more challenging it is to conserve battery life while maintaining performance.”

In short, “consolidation means good things for your battery.”

With the Snapdragon 865, Qualcomm disregarded that line of thinking. Not only will flagships running the Snapdragon 865 use a lot of power when tapping into 5G on this separate chip, they’ll do the same when using the much more commonplace 4G LTE since Qualcomm moved the 4G modem off the main SoC, too. That’s a head-scratching move, until perhaps, you consider that it’ll mean that Qualcomm will sell even more chips now (all 865s will be accompanied by a separate modem chip now).

Worse yet is the fact that an additional physical chip leaves less space for other components like bigger batteries and will cost consumers more money.

The industry’s solution, as we’ve already seen with early 5G-enabled handsets, is to simply go bigger. The Galaxy S10 5G is a perfect example as the 5G variant measures a hulking 6.7 inches diagonally. That extra size is needed for a bigger battery and to help dissipate heat, the latter of which is still apparently an issue for early 5G phones.

As Ars notes, it feels like Qualcomm is trying to rush 5G just like it did with 64-bit. 5G will still be in its infancy next year so the hurry isn’t really justified – not enough to make next-gen flagships perform worse on 4G, anyway. Given the way the stars are aligning with the very real limitations of mmWave technology and the threat of bigger, hotter and more expensive handsets, 2020 may not be the year of 5G after all.

Galaxy S10 5G image courtesy BGR

Permalink to story.

 

Uncle Al

TS Evangelist
OR .... maybe they have done their homework and figured out that 5G could be a pink elephant ....
 

neeyik

TS Evangelist
Staff member
It’s puzzling because Qualcomm could have easily retained an on-die 4G modem but used a separate chip for 5G. This would make sense given how iffy the 5G situation is right now. But to completely remove an embedded modem of any kind is hardly progressive thinking for a modern SoC design.
 

ferrellsl

TS Booster
It’s puzzling because Qualcomm could have easily retained an on-die 4G modem but used a separate chip for 5G. This would make sense given how iffy the 5G situation is right now. But to completely remove an embedded modem of any kind is hardly progressive thinking for a modern SoC design.
It's doubtful that they could have retained a 4G modem on the 865 given it's higher transistor count and additional shader cores while sticking to the same 7nm die size as the 855. They had to squeeze the extra functionality in somewhere and the logical choice would be to drop the transceiver. What isn't progressive about offering a stronger CPU/GPU for a computing platform?
 

ferrellsl

TS Booster
I think you answered your own question....

"and will cost consumers more money"
No, I haven't answered it. I'm still wondering why the author thinks it's puzzling in light of the engineering and all the added features that were incorporated into the 865 while sticking to the same 7nm die of the 855. Squeezing all that added horsepower onto the same die meant that something had to be removed and the transceiver was the logical choice. Here is the list of upgrades that were added to the 865 not found in the 855. And of course it costs more. You didn't expect all that for free did you?

-----CPU-----
Kryo 585 CPU cores with 20% boost over the older Kryo 485 cores found in the 855

-----GPU-----
Adreno 650 GPU with 20% performance boost over the 855

-----ISP-----
Spectra 480
Up to 200MP photo
64MP/dual 25MP with zero shutter lag

-----Video-----
8K at 30 fps
4K at 120fps
720p at 960fps unlimited recording
Dolby Vision, HDR10+, HDR10
10-bit H.265

-----Connectivity-----
Bluetooth 5.1

-----AI-----
Hexagon 698
Qualcomm Sensing Hub

-----Charging-----
Quick Charge AI
 

neeyik

TS Evangelist
Staff member
It's doubtful that they could have retained a 4G modem on the 865 given it's higher transistor count and additional shader cores while sticking to the same 7nm die size as the 855. They had to squeeze the extra functionality in somewhere and the logical choice would be to drop the transceiver.
I couldn't (quickly) find a labelled die shot of the 855, but here's one of the Kirin 990 SoC:



It's not the Snapdragon 855, but the setup is reasonably similar. One can see that the suggested 4G modem is quite sizeable as part of the overall die area, but it's roughly the same as the CPU cluster and about half that of the GPU.

Now the GPU in the 865 is going to be a fair bit bigger than the Mali above and that in the 855, due to the 50% increase in ALUs and ROPs, but the layout of CPU clusters, GPUs and cache are such that they're quite easy to scale in size, and not have the overall die balloon in size.

The 855 is only 74mm2 in area - Apple's A13 is 99mm2 and 8.5 billion transistors, and it has a 4G modem. The Kirin 990 has 10.5 billion transistors and an embedded 5G modem. In other words, competitors have gone larger, but still retaining a modem.

What isn't progressive about offering a stronger CPU/GPU for a computing platform?
The 865 will only work with Qualcomm's 5G modem - they're not, as far as I'm aware, offering any 4G LTE support with it. So any phone/tablet designers who want the simplest, most compact, 4G LTE layout won't be offer such a product with the 865. Why design and release a new, and clearly better SoC, but then limit its viable market?
 
  • Like
Reactions: Godel

ferrellsl

TS Booster
Why design and release a new, and clearly better SoC, but then limit its viable market?
They aren't limiting a viable market, they're actually expanding it by opening up the 865 to tablets, Chromebooks, gaming devices etc...It makes no sense to offer a 5G or even a 4G transceiver for those devices and a discrete 4G/5G transceiver can easily be added to a cell phone. Many cell phone makes already do this and it gives them the flexibility of adding/updating their existing designs designs without having to go back the drawing board.
 
  • Like
Reactions: TempleOrion

neeyik

TS Evangelist
Staff member
They aren't limiting a viable market, they're actually expanding it by opening up the 865 to tablets, Chromebooks, gaming devices etc
A SoC with a modem is just as viable in those markets, though; the modem doesn't have to be enabled. But where it is available, that SoC is viable for LTE tablets, Chromebooks, gaming devices, etc. So the Snapdragon 855 would be fine for WiFi only or WiFi+LTE models of these; the Kirin 990 5G would also be fine. These would also be single chip solutions for the markets, whereas the 865 is always going to be a twin chip solution, for markets desired mobile comms. That seems, to me at least, to be a limit of some kind; I can see how it could be considered to not be a limit too, though.

Related to this, I incorrectly said in my last post that "they're not, as far as I'm aware, offering any 4G LTE support with it" - this isn't true, as the X55 chip does offer LTE support.

Many cell phone makes already do this and it gives them the flexibility of adding/updating their existing designs designs without having to go back the drawing board.
Do you have a specific example of this? It would be interesting to see what SoC is being used, where the device is also using a discrete modem chip.
 
  • Like
Reactions: TempleOrion

m3tavision

TS Evangelist
11 days ago, at AnandTech:

5G and Modem Integration
IC: Current 8cx solutions rely on external 5G modems (external to the main SoC) in order to support 5G connectivity. We’re still in a 4G/5G transition phase – would you expect an integrated 5G SoC/modem as a single piece of silicon in this sort of form factor?

AK: Absolutely. If you go back to the 4G days when it was launched, we took all of the 3GPP specifications for 4G that were defined and we put it in a bit of silicon. Over the course of a couple of years, you hone in on the features and the functionality that are most useful – you figure out different partitioning and architectures for your modem to make it smaller, and then that gets moved onto the main SoC. That’s exactly what is going to happen with 5G as well. As you’ve already heard, we are going to have integrated 5G SoCs coming in 2020. The life of a standard mobile PC is in the region of 2-3 years, and you’re not going to have a new mobile SoC every year, so the next generation after 8cx is where you should look for an integrated solution.

IC: We currently have a bifurcated strategy with 4G and 5G, given that some markets have been quicker to adopt different parts of the 5G standards than others. Is that strategy of separating 4G and 5G silicon expected to continue long term, or will there be a right time to go for a single integrated 5G philosophy across the board?

AK: I think the good part right now is that we have a line, a product line of devices that span our entire portfolio. In PC devices, we have the 8cx at the top, but we also have a version of the 8cx called the 8c. It will be a little less powerful, but it targets parts of our line-up in-between the 8cx and the 850. And there’s another one coming, so you will have a whole line-up – some of them 4G, some of them 5G.

IC: Is the smartphone market ready for 5G in the mid-range?

CA: Yes. You have to look at all the stakeholders and why we’re so focused on scaling right now. There are three parts to this answer
.
First, think from the operator’s perspective. It’s very clear, you can argue, that you can have a conversation about how fast you are going to build coverage, how fast we are reforming the spectrum, but the trends are very clear: users are using more data, and the average data consumption per month is going up. If you look at 5G in Korea, there are 2 million subscribers, and the data from KT (Korea Telecom) is that average data per user has doubled. So from the operator perspective, you’re going to get to a point where it is uneconomical to give unlimited data plans on 4G, and there is a need to switch for 5G, as 5G offers you a lower cost per bit and gives the user to have a premium experience with less congestion. Operators want to move you onto 5G for a variety of reasons, but that one is key from a user experience perspective
.
IC: We’ve seen a few operators show resistance to providing affordable 5G plans. From my perspective, aside from the cost of current devices, that’s somewhat of a limiting factor.

CA: Depending on the market there may be some segmentation of price plans, and I think that network slicing is going to enable this. I feel that every new technology that hits the market is expected to have a price premium for that extra performance. But overall, the economics, from the operator perspective, there’s one element on the revenue side and that’s how it is priced to the consumers. But in terms of operating costs, 5G brings about a different cost equation. Especially if you can build out knowing how the customers will use data in the future.

It’s hard for me to say what the operators will exactly do, but I’ll tell you that we’re still at an initial scale-out. The thing is, having worked at an operator, it’s all about transitions from one generation of technology to the next. At the beginning of those transitions, you have a non-ideal situation, based on hardware and deployment. Even in Europe today, we still have operators with 2G. As operators have migrated from 2G to 3G then to 4G, you have to adjust that cost equation, paying maintenance, electricity, paying all those things, and in the beginning of a transition, you have to run multiple networks, and then by getting people on board helps smooth the deployment out to future sites. So I think you’re going to see the equation of the economics being amortised as more users sign up, along with CapEx (capital expenditure) and OpEx (operational expenditure). How they do that deployment, in a transition phase, depends on the technology, but ultimately it’s a drive to a lower cost per bit. 5G is designed for this, and the data usage model will change: customers might not have a fixed line into the home in the more, and use 5G for that last mile of cable. This is why scale is so important.

For part two of the question ‘is the smartphone market ready for 5G in the mid-range’, we should consider that we now have a very mature smartphone market. In a mature smartphone market like the United States, the number of people with at least a ‘good’ device is high, and people buy knew phones every 2.5-3 years. In Europe it’s more 3-4 years. So if you’re going to buy a phone right now, and you already know about 5G and heard that 5G is coming, you can end up with a device for 2-3 years that won’t support the fastest speeds or get the best user experience. Users who know about 5G will want a 5G phone with how fast 5G is being supported.

The third part of my answer is that the device ecosystem is mature ahead of the network: the devices are ready and the network is catching up. Unlike the transitions to 3G and to 4G, where the operators were first, here at Qualcomm we have been accelerating and driving the device adoption of 5G. So even if you never use the 5G modem in the device, it’s still a very competitive unit, such as the Samsung Galaxy S10 5G, from the industrial design, battery life, and all these things. With the device ecosystem being ready, the operators are getting there, and with more phones enabling 5G today, users aren’t going to want to be stuck on a 4G smartphone.

The last piece that you don’t see right away with one of these new devices is the 5G – there is a bigger difference between premium and the mid-range here. If I have a Samsung high tier device, or even a mid-tier device, and then I go buy the premium tier, what I will see is the premier tiere has a better camera, a higher end display, maybe a bigger screen with more screen area, more memory, better graphics, and the games probably play a little better. The services on both devices are still kind of the same – I still have WhatsApp, I still have Instagram. Some might perform better, some might be worse. With 5G, it’s all a little different. When it comes to the question of say, streaming 4K video, it’s a very binary result: you either have it or you don’t. If I want to stream some games, you either can or you cannot. It’s a very binary proposition. The evolution of social media with live broadcasting, such as Facebook Live, means that the entire developer ecosystem says that if the capabilities are there, users want to have it.

When you put all of this together, you get to the same conclusion that we got to, and I think that forms the key basis of our announcements this year. Mission one has been to get 5G launched, and get all the flagships supporting 5G. We have our flagship modem, and we have our second generation 5G that we will show at our Tech Summit later this year, and we’re going to impress you! But then that still isn’t enough – we need to scale 5G, and we need to scale it right now. The transition is going to be faster than we’ve seen with 4G, and that’s why you’ll see our 7-series and 6-series with 5G and we’ll keep going.

IC: Currently we’re seeing a mix of discrete 5G solutions and integrated 5G solutions, whether it’s in a smartphone or other devices. There has been some commentary regarding performance in both methods, such that some discrete options can offer 2x the bandwidth of an integrated solution. What is Qualcomm’s approach here?

CA: You have to think about 5G a little differently to 4G. I won’t comment on our competitors, as they’ll do what they want to do, but I can tell you that 5G gives you the flexibility to have differences in the performance as you design the modem. Despite this, there is one very important thing on 5G is average speed – if you remember when Qualcomm pioneered Gigabit LTE or even 2 Gigabit LTE, you had a significant improvement in peak speeds, but the average speeds don’t improve at the same rate. It’s very different with 5G – with 5G the average data rates go up and, what you’re going to see as we start moving our modems onto our 7-series chipsets is that we’re not going to compromise, average data rates aren’t going to be compromised, and latencies aren’t going to be compromised. You can still have different choices, as you’ll see from our portfolio, we’ll still have full support for millimeter wave, and full support for Sub 6 GHz. We are going to unveil all the details of our modem in the 7-series chipsets at our Tech Summit. But we have a plan, and you’ll see a pattern.




Essentially, this new chip is being replaced next year:
("As you’ve already heard, we are going to have integrated 5G SoCs coming in 2020.")
 

Wizwill

TS Booster
If you are a large-sized city dweller or live in a d tensely-populated suburban area and spent most of your online hours outdoors, , 5G makes perfect sense However, in either setting, if you work indoors in a modern building in that same mid-sized city, or live in a brick-faced house,in those suburbs you will probably be quite unhappy With the range and speed of your 5G-connected device.

A basic tenet of electronic physics is that higher frequencies tend to be attenuated by structural impediments more so than lower frequencies. Since those cities will have more cell towers as will the densely populated suburban areas, if you are marketing to those who live in those environments, 5G is the new golden goose.

God forbid you should try to utilize 5G devices in a rural area for checking maps or phoning.

I think someone IS going to realize the 5G king is wearing no clothes shortly. Qualcomm has been relatively prescient so far, so I think their decision not to include a 5G modem in their latest and greatest SOC is probably self-preservation.
 

dogofwars

TS Addict
Probably for market adaptability, it's not like there is a clear road map so they can produce the SoC in mass and adapt the amount of modem chip they produce depending of the market and rule and restriction.
 

RobJoy

TS Rookie
I think the move was smart.
They effectively removed the MAJOR heat generation from the SOC itself, so it now will not need to down-throttle as much, if at all.
Also it gives the manufacturers capability to position the chip behind an EM shielded and properly cooled enclosure. Win, win.
More speed at a minimal power drain delta.
Stop blowing this out of proportion, the power drain will still be minimal.
Qualcomm HAS actual engineers who DO know better than some couch developer wannabe journalists.
On top of that, they drive the costs down and give some manufacturers flexibility to opt out of 5G entirely, while still being able to use the sweet sweet speed of the new SOC.