Two driving forces are driving the present expertise market: insatiable demand for hardware, and the provision chain shortages making it troublesome to provide sufficient in amount to fulfil each order. Even with these two forces in motion, corporations should push and develop subsequent technology applied sciences, as no competitor desires to sit down on their laurels. That features Qualcomm, and as a part of the Tech Summit in late 2021, I sat down with Alex Katouzian, Qualcomm’s GM of Cell, Compute, and Infrastructure to speak in regards to the points confronted in 2021, the outlook for 2022, the place the relationships lie, and the place innovation is headed in terms of smartphone and PC.
We’ve interviewed Alex a few instances right here at AnandTech, and whereas he has been properly educated in coping with the likes of our pointed questions, we will inform he’s being open and sincere, telling us as a lot as he can (and typically a little bit bit additional). Alex has been at Qualcomm for 20 years, working his manner up via wi-fi cellular applied sciences, product administration, and taking up the MCI enterprise unit in 2017. Previous to that Alex was engineering multimedia and wi-fi chipsets, showcasing that his complete profession has been about placing next-generation connectivity into as many devices as potential.
 Alex Katouzian SVP and GM Cell |
 Ian Cutress AnandTech |
Ian Cutress: Right here at Qualcomm’s Tech Summit, you might be asserting the brand new Snapdragon eight Gen 1, with new advertising, and it’s a brand new chip on Arm v9. I sort of need to step again and take a look at the 12 months as a complete – Qualcomm launched the Snapdragon 888, and that is been in quite a lot of flagships, accompanied by quite a lot of them with Qualcomm 5G RF options. I do know you are an worker, and I do know you are most likely biased, however objectively what grade would you give Qualcomm for the 12 months?
Alex Katouzian: Nicely our Snapdragon 888 is prime in premium tier options. Once you take a look at the capabilities that had been obtainable to our clients and ecosystem via the applied sciences we supplied, we topped the chart by way of graphics functionality, total efficiency, AI functionality, and digicam efficiency – there are a number of telephones on the 888 which might be topping out on DXOmark. I’d say our Elite Gaming functionality was superb, we’ve executed that in cooperation with the builders and sport engine builders. All of them had been very cottoned on. So total, I’d say a fairly excessive grade on the general efficiency of a premium tier telephone.
IC: Would you say you’ve got had any pinch factors this 12 months, that get solved subsequent 12 months?
AK: Clearly the pandemic introduced on some provide chain points, and the severity of them fluctuated all through from the top of 2020 via into 2021. However by being a fabless firm, it allowed us the liberty to have the ability to transfer from fab to fab, not solely on the newest digital elements of expertise, but additionally on the previous analog expertise we have had the flexibility to maneuver from fab to fab as properly. We alleviated quite a lot of the provision points based mostly on the flexibility to have engineering assets obtainable to maneuver elements and make room for shipments to occur. Though we nonetheless went via some provide crunches, and we most likely will undergo some provide crunches in 2022, we alleviated quite a lot of that, and we’re capable of get out our merchandise.
IC: Does shifting from fab to fab like that disrupt the distinctive relationships with every foundry?
AK: The connection between a fabless semiconductor firm like us and the foundry needs to be very tight. However these had been unprecedented instances, and the ripple results had been nice as a result of the provision chain normally takes about 18 months to 2 years to return to fruition. Once we had that shutdown interval of say, 9-10 months, each facility within the provide chain was shutting down. They needed to restart that, to recalibrate gear, to rent workers, and any enlargement plans had been on maintain. Then the surge in demand for client digital gear was so excessive that it takes time for it to catch up. Given these ‘unprecedented instances’, I do not suppose that a foundry goes guilty an organization like us to attempt to diversify in such a manner that it permits us to ship merchandise. Should you look additional down the provision chain, we’re suppliers to our clients, and so they’re gonna should diversify of their provide as properly, and somebody needed to make that alternative.
IC: Aha, so goes each methods, each up your chain and down it.
AK: Precisely!
IC: So once I got here to my first Qualcomm Tech Summit, the venue had specialist gear put in for 5G as a result of we had been nonetheless very early, very nascent in 5G. The rollout on the time was extra centered on industrial purposes, however a very good quantity of the speak was on mmWave versus Sub-6. The place the US is specializing in mmWave, the remainder of the world was centered on Sub-6. If feels sort of odd, however we’re sort of nonetheless having these discussions! What has modified from Qualcomm’s perspective in these discussions?
AK: I feel increasingly more, the carriers are realizing that ‘a layered cake impact’ of their community, spreading out, turns into a extra essential factor to do in crowded environments or congested areas. mmWave roll-out, although typically it a little bit bit tougher as a result of the quantity of base station towers, it’s important to have much more of them, although they’re a little bit bit tougher, carriers are realizing that it really is a lot better to have mmWave in congested areas. Then, as you go additional out to extra suburban areas, that Sub-6 can cowl and much more, perhaps probably even falling again to LTE.
However that realization, and the deployment of the gear – bear in mind, we’re on the finish of the second 12 months of the industrial launch of 5G, and for each G, the transition takes between 10 to 12 years. So the rollout of mmWave goes to begin to occur extra. I feel extra international locations are going to begin to deploy: we’ve got deployment in Japan; Korea is slowly beginning to come up; China’s beginning to consider the very fact there’s numerous bandwidth obtainable there for them. Then the sectors that mmWave can get utilized to are a lot wider than what we had earlier than, in several sectors of say financial system, manufacturing, retail – after which you’ve gotten different technique of deploying mmWave to personal networks by way of small cell. These are all serving to. I feel once you take a look at industrial, once you take a look at manufacturing, once you take a look at automotive, there’s quite a lot of room to develop for mmWave, and that enlargement goes to be from the ecosystem.
IC: For this occasion, we’re studying rather a lot about Snapdragon eight Gen 1. As you’ve got mentioned, and because the numbers present, you guys are a premium smartphone chief, persevering with from the earlier technology. In my thoughts, I am juggling up the truth that as a premium chief, you need to preserve the market share, however you additionally want the whole lot to work to keep up unit gross sales. So on some stage, you’ve gotten play it secure to be sure to have that product working appropriately – however on different ranges, it’s important to take dangers with new options, and subsequent technology applied sciences which will or could not work long run. Evaluate that to all of your rivals which might be making an attempt to catch up, and might take these dangers. How precisely do you steadiness that? Is there a proper technique to speak about how that’s balanced?
AK: I feel the way in which to take a look at it’s that the technique at Qualcomm has been to take the danger on the expertise on the biggest channel that you’ve got, after which re-use that expertise in merchandise in progress sectors which might be adjoining to our largest channel. So cellular is clearly the biggest section we’ve got at Qualcomm – so you will note the danger taking in advancing expertise in mobiles first, although once you’re a cellular first surroundings, energy dissipation needs to be prime of thoughts. Nonetheless, to illustrate I need to apply AI functionality to cellular, energy is the primary standards. Then I can determine my efficiency to energy ratio, and I can scale from there. But when I need to take a look at automotive, and I need to do autonomous driving, efficiency is the primary standards, energy is the second standards, perhaps even the third. Nonetheless, we’ve got the flexibility to scale with our expertise roadmap, so after we take the prospect and the danger of placing sure applied sciences in cellular, we begin to consider what the purposes are so we take that danger, we put it in, and see if we will develop that market. Then we will determine the best way to scale the bottom design right into a market that wants efficiency first, and energy second. In order that’s how we take a look at it throughout the board.
It is the identical precise software on CPU – in the event you bear in mind, just a few years in the past, we had our architectural license with Arm. We introduced in on our personal CPU, and we took benefit of that, and we sort of tried to develop the market. At the moment, we did not have adjoining companies that we had been actually rising – we had been fascinated by them, however we weren’t rising them, however the software is precisely the identical. Now with Nuvia approaching board, we will do the identical factor you realize – a ‘one expertise roadmap’ that may scale. We’ll first have that first occasion of that in a PC chip sampling subsequent 12 months, attending to merchandise in 2023. Then we’re going to consider the best way to deliver it into cellular, automotive, after which infrastructure software as properly. So we’ve already thought of the best way to do the facility efficiency scaling for various companies we’re going after. Identical with graphics, similar with digicam capabilities, similar with video capabilities. Every of those applied sciences take the danger of it being launched in cellular first, scale it, take into consideration different markets that we go to after which we alter our efficiency and energy factors.
IC: It is nice that you just mentioned video capabilities, and talking in regards to the digicam, as a result of one of many factors that got here out of the displays and a few of our briefings is that Qualcomm is pushing 8K HDR, and extra superior 4K HDR recording. However the silicon does not have an AV1 decode unit, and a few folks have mentioned that it’s since you guys aren’t a part of the AV1 Alliance. Are you able to shed some extra mild on the selections there?
AK: So after we plan our applied sciences out, typically on units such because the premium tier, we’ve got to consider issues two or three years prematurely. The timeframe that we have to introduce these units may be very important based mostly on the OEM launches, and the wants of the carriers. Given these circumstances, we have not been capable of accommodate AV1. We’ll most likely accommodate sooner or later, however up till now, we simply have not been capable of do it. It’s not that we do not need to do it – we simply have not had the flexibility to plan it in in time to attempt to make that occur.
IC: However you’ve received folks engaged on it?
AK: Oh, sure – we’ve got a number of video specialists. I imply, we contribute to requirements on video continually. We’re concerned with the place it is heading. We simply suppose that in time, we’ll have these capabilities – there is no doubt. We simply did not have the chance to get it in time schedule-wise, planning it in such a manner that it’s going to hit our timeline.
IC: Relating to relationships – Qualcomm and Microsoft. The Home windows on Snapdragon ecosystem is a protracted partnership, and we’re now seeing 64-bit emulation coming with Home windows 11, enabled with last-gen units and future units. How’s that relationship progressing? You’ve had Microsoft on stage at this occasion!
AK: Extraordinarily properly. Our partnership with Microsoft has at all times been strong, by way of a cloud-and-software firm cooperating with a technology-and-hardware firm, to supply an answer that is smart for the customers. They’re utilizing all these buttons, throughout telephones, throughout PC, and we’re collaborating throughout IoT, even AI capabilities, not simply in telephones but additionally in PCs, and probably in data-center collaboration as properly. Nonetheless, our relationship on the PC aspect has been changing into increasingly more refined as they be taught the capabilities and applied sciences that we’ve got. They’re additionally taking the danger of exploring how multimedia functionality can turn out to be increasingly more prevalent on the PC. As you realize, I feel you agree that cellular traits are coming to the PC and the belief that being able to be cellular and linked and having nice multimedia, together with digicam, together with audio, bettering AI, all of that’s changing into increasingly more essential, together with efficiency functionality of the CPU and GPU – a mixture of which might turn out to be very properly suited to a cellular workforce.
However let me offer you a little bit little bit of a historical past. Each enterprise Qualcomm has gotten into that has had incumbents. We have had three components that we have put in to satisfy KPIs (Key Efficiency Indicators). One is having a disruptive expertise that may make a distinction, two is a long-term funding, and three is a go to market accomplice. I can go way back to UMTS – we invested in UMTS for seven years earlier than it turned extra mainstream and to develop. Now, clearly, we had a disruptive strategy which was a extremely built-in answer with a protocol stack that was worldwide examined, and we may make extra OEMs horizontal versus vertical. So we had that disruption, we put in these seven years of funding, and our go-to-market companions had been the horizontal OEMs that exist at present. We did the identical factor for Wi-Fi, and we did the identical factor for RF entrance finish.
Once we are stepping into the PC market, we’re doing the identical factor. Now we have disruption, we’ve got long-term funding, and we have been doing this for about 5 years now. We simply invested in Nuvia, not only for the aim of the PC, however a chunk of it’s for the PC, and we’ve got our go-to-market accomplice in Microsoft for Home windows, and we’re working with Google Chrome based mostly units as properly. So these three components are met, and we’ll proceed bettering our units and capabilities. This additionally means working with Microsoft end-to-end, to use the entire underlying applied sciences to their OS to be optimized to run on Snapdragon units, after which drivers to show all of the applied sciences to customers.
IC: If I may get some readability on the connection between Qualcomm and Microsoft. There have been public discussions relating to what you name Home windows on Snapdragon (named due to the connection, though lots of people wish to name it Home windows on Arm). There have been reviews saying that, between Microsoft and Qualcomm, there’s been an unique settlement for these Home windows on Arm/Home windows on Snapdragon collaboration exclusivity is slowly coming to an finish. Is there something you possibly can touch upon that?
AK: I actually cannot touch upon that, however as I mentioned, the cooperation between us has by no means been this sturdy, and I feel for the foreseeable future we will work with them extra to extend the capabilities of Arm and/or Snapdragon based mostly options.
IC: So that you talked about Nuvia a few instances, a subject very pricey to my coronary heart. I’ve met with Gerard, Manu, and John pre-acquisition, and I used to be actually trying ahead to their server chip they had been going to provide. So I need to admit a little bit little bit of me died once you guys acquired them, however I need to applaud that it was a intelligent transfer in buying them. Given what Qualcomm CEO Cristiano Amon was saying on the time, saying about ACPC since its inception, it truly is the route it must go in. I do know Nuvia is extra the subject for subsequent 12 months’s summit, as you’ve got mentioned sampling in 2022, product in 2023. How is that acquisition going, and the way is the crew integrating?
AK: Extraordinarily properly, extraordinarily properly. The founders are very energetic, and I can personally say I most likely meet with them on a weekly foundation, as much as two or thrice per week. My colleagues which might be operating the PC enterprise inside my enterprise unit meet with them every day. We evaluate roadmaps on a regular basis, we evaluate methods, we’re very cooperative on the best way to go to market, not simply within the close to future, however in the long run as properly. The system structure, the chipset structure, are all being mentioned repeatedly – partitioning is being mentioned repeatedly, so it is a very tight integration of this crew, and that is precisely the way it additionally occurred up to now. For instance, we purchased two Wi-Fi corporations, one is Atheros. We combine these groups in inside the Qualcomm group and sort of make them one Qualcomm, after which we proceed that manner. That is precisely what’s taking place with the Nuvia crew as properly. And I’d say, on the server aspect, that these are opportunistic companies that we’ll go after. So we’ll see what occurs.
IC: Will the Nuvia core be known as a Kryo core, or will you’ve gotten any fancy advertising title for it?
AK: I have not thought of it but – it undoubtedly will not, or most likely received’t, be Kryo. However we have not thought of it but.
IC: Give me a name once you want a reputation!
AK: Hah, certain!
IC: Should you can maybe clear one thing up for me: is the Nuvia crew making a single core, or each a giant core or a little bit core? Or is it that they’re dealing totally with the SoC construction into which you add within the connectivity and the graphics?
AK: It’s each, all of the above. By that I imply that it is unattainable for us to place out a chipset answer as refined as this with out having the complete system being considered. Consider it this manner: the CPU by itself is a part of the ‘one expertise roadmap’, however so is graphics, and different issues. Then we’re actually fascinated by bringing a whole system answer to the PC and altering it in such a manner that you do not go after the normal designs. You realize, we talked about this earlier than, we’re searching for invoice of fabric financial savings, we’re design financial savings, we’re inner routing, we’re shell designs, and we’re thermals – the entire thing. So it needs to be a whole answer. In order that they’re undoubtedly concerned in the entire SoC design, and so they’re concerned in a number of completely different cores, the place it is smart. Whether or not it is large cores, little cores, or a mixture of what number of.
IC: Nuvia’s design will probably be Arm-based, and the Nuvia crew comes from individuals who constructed Apple’s M1 and expertise from Google. So will that chip compete primarily towards Apple as a result of it is ARM-based, or towards Intel as a result of it is Home windows-based?
AK: The units that we will come out with based mostly on the brand new CPU design and the brand new structure, will compete head on with Apple.
IC: So once you do your comparability charts, you will be evaluating towards Apple?
AK: Sure.
IC: Is that as a result of Intel is not within the image? Or is that simply because that’s the place the market is?
AK: The way in which we take a look at it’s that we’re not going after discrete designs: we’re going after an SoC and the structure that makes one of the best sense for the PC. Like I mentioned, cellular traits are coming into the PC, and I agree with you that efficiency smart in comparison with cellular it’s a lot greater, it has extra energy dissipation functionality, however we will attempt to make it rather a lot sleeker and have much more cellular based mostly. So our comparability is to an organization that may do each forms of SoC and convey that functionality to the PC, so we’re actually making ready ourselves for that.
IC: Qualcomm has a really environment friendly graphics structure in Adreno. It normally performs rather well within the cellular SoCs, however we’re presently at a time the place the discrete GPU market goes bananas, and the whole lot is promoting. I’d like to see a scaled Adreno GPU, and on condition that Intel can also be popping out with its personal, Qualcomm could possibly be a tremendous fourth competitor in that house. We’d like to get your feedback on that, however the important thing factor is: in the event you had been to do this, what coloration would it not be? AMD is pink, Intel is blue, NVIDIA is inexperienced, and so is an Adreno discrete GPU going to be gold?
AK: It is onerous to say as a result of we do not even have that product! However in the event you suppose alongside the traces of premium, yeah, you realize, it might be some shade of gold. However consider it this manner: we’ve got to enter the PC market with a really refined design that we have already labored on SoC-wise until now, however we will get a lift in terms of CPU functionality. We will scale our GPU functionality identical to we mentioned earlier than, and we’ve got the flexibility to scale throughout these units. So we undoubtedly can have way more performing GPU and CPU cores. Clearly additionally concentrating on energy dissipation, and people ratios have gotten to be proper, so if we choose a design level inside the PC, we’ll undoubtedly want these necessities. Then we will break off at discrete GPU if the enterprise is smart for it, however we undoubtedly have the aptitude to scale to that stage and the design functionality to provide one thing in that market. Plus, the onerous play is about having way more of an ecosystem. So as soon as we get into the market extra closely, and get extra video games developed on there for instance, and different purposes, then, the likelihood at all times existed.
IC: To me, it’s a no brainer! Wonderful, thanks a lot Alex on your time.
AK: Thanks Ian!
Now we have an upcoming interview with Qualcomm’s Miguel Nunes, Senior Director of ACPC, about the way forward for that line of merchandise. Say tuned!