Tantra’s Mantra Podcast – Episode 37

Listen & Subscribe On

Transcript

Prakash Sangam:
Hello everyone, welcome again to Tantra’s Mantra where we go behind and beyond the tech news headlines. I am your host Prakash Sangam, founder and principal at Tantra Analyst. Today’s subject is quite an interesting one. It’s about a major shift happening in the PC industry. Of course, I’m talking about Windows on Arm.
As we all know, Qualcomm has been making processors for Windows for many years now with a handful laptops using Snapdragon 8CX platform. Actually, I used quite a bit of them and very extensively, probably I was the only analyst who used Windows on Arm laptops exclusively for more than a year as my daily driver.And I’ve written quite a bit about them as well.
But that’s all set to change with their new Snapdragon XElite platforms.They announced it last year and we’re expecting the devices very soon in the market. And they also recently announced XPlus, so they’re expanding the platform as well.Both of these are based on the architecture developed by Numia, the company that Qualcomm acquired in 2021.
The benchmarks of XElite show a lot of promise, beating Apple’s M3, and probably we’ll see how they do against M4 that was recently announced, and of course beating Intel’s Meteor Lake by a large margin.
And with Microsoft Build coming up in a couple of weeks, the expectation is there will be a lot of interesting announcement and news coming out of that event. Of course, Microsoft will talk about the platforms and Windows on Arm platform and so on. And the expectation is a lot of PC OEMs, including Lenovo, Dell, HP, and others announcing their products there as well. I think it will be starting from there and it will be an onslaught of these announcements as we move in the year.
I’m sure you’ve seen and are reading the leaks and the rumors about these devices as well. So to discuss this topic, not from a specific performance or anything like that, but more on an industry-wide discussion, I’m delighted to invite back my good friend and fellow analyst Leonard Lee from Next Curve. Leonard, welcome to the show.
Leonard Lee:
Prakash, thanks for having me back. Yeah, really looking forward to this discussion. And for those of you who don’t know me, I’m Leonard Lee, I’m the executive analyst at Next Curve. And if you’d like to find out about Next Curve, check us out at www.nextcurve.com. And with that, let’s get into it.
Prakash Sangam:
Absolutely. Thanks for the details. I highly encourage you guys to check out Next Curve website and all the good work Leonard is doing. He generates quite a lot of content. So, all of it is really worth checking out. I highly recommend check out the stuff.
So let’s get started. So we have tons of things to discuss. I’m sure we can spend hours on it.
So I’ll try to be a little bit concise on the market impact aspects of it, not do details about the benchmarks and comparison and so on. So I’ll just talk about Windows on Arm for so many years. And I think this is the moment for Qualcomm, right? Where they have the real opportunity to append the PC industry and make a mark. I think XElite has the qualities and the characteristics of making Windows on Arm mainstream and make a reasonable impact. What do you think?
Leonard Lee:
Yeah. I mean, I think there’s really two inflection points that tap into Qualcomm’s legacy. Like we were mentioning before, Windows on Arm, they’ve been working on this for a long time. But the other thing that they’ve been working on, and it caters to this pretty robust hype that we’re looking forward to, which is the AI PC.
Qualcomm’s been, they’re just right along there with Apple and having introduced NPUs into their smartphone devices and probably ahead of everyone else in the PC space and NPU. And, you know, they call it Hexagon. People probably don’t know that Qualcomm actually has a deep legacy in on device endpoint AI. A lot of that is due to their IoT, but most notably their smartphone legacy.
They’ve been working all this stuff for a long time. And so this, I mean, that’s, I think, this dual convergence of critical technology trends that I think are going to work really well for Qualcomm regardless of what the benchmarks claim. But I agree with you. You know, the benchmarks themselves, you know, there’s a lot of debate, arguments and, you know, shots being shot across the bow, right? So and, you know, I always I always consider these benchmarks really sort of apples to oranges comparisons, right?
It’s really difficult to get a good baseline, you know, benchmark. But what’s I think pretty clear is they have they have some pretty serious stuff that their brain table with the Nuvia technologies, which is now manifest in their CPU cores that are dubbed Oryon. So yeah, I mean, I think it’s a very exciting time for Qualcomm and a time for contemplation for everyone else in the PC space, I think.
Prakash Sangam:
And I fully agree with you. And specifically this point about AI on devices. Yeah, I mean, they started working on AI, when it was not as sexy as it is right now. So I mean, they called it DSP at that time, Hexagon DSP. And then they had to rename it as NPU. I mean, obviously, it has increased the performance by magnitudes in the newer versions. But they have a good legacy there, specifically on the on device.
And then, as you said, it bodes well for AI PC. And then it looks like AI PC has legs, right? We both were at an Intel conference last month. And then if you look at the shipment of AI PCs, I know that it seems to have good traction. And then kind of all indications there that it will start another upgrade cycle, especially in the enterprise space, right? I mean, I think the jury is still out with all that, right?
Leonard Lee:
You have shipments, you have OEMs purchasing the next generation of processors. They just happen to be loaded with an NPU, which is, I think, what largely constitutes if a processor is an AI PC processor or not, right? I think a lot of people in the industry will remind folks who are, let’s say, casual observers of the industry that GPUs have been there for a long time.
You can run a lot, pretty much anything on a GPU. It’s the NPU that brings a lot of energy efficiency and optimization to the table for certain types of AI workloads, where you’re going to get an additional benefit in terms of power-efficient AI compute.
And so this definition of AI PC, Microsoft, I guess, people have just allowed them to define what it is.And I think the minimum technical requirement or specification for processors is that it has 40 tops of NPU compute. I don’t know if that really matters, but that is what I think you’re going to be seeing coming to market regardless.
Inevitably, we’re going to see more of these processors with NPUs and them being in the market. I still have a lot of questions on what is the immediate use case for these things. I mean, it’s more of a build-it-they-will-come kind of mentality, but I think that’s another discussion that we can have another day. I mean, maybe you can share some insight, because I know you were at Adobe, I was at NAB. I go to a lot of different types of industry shows.
I don’t just focus singularly on, let’s say, the semiconductor industry. I have a pretty broad range of end-market research that I do and exposure to different industries that consume, especially the high end of silicon. But you do the same thing, right? I mean, I know that you were at the Adobe event, so hopefully we can share, sprinkle some additional color.
Prakash Sangam:
So I think in terms of use cases, we are quickly talking about it. There are traditional use cases, not Gen.EI, that have been traditionally there and being enhanced with the AIPC. No question about that. There is definite need for TPU in the devices, but the Gen AI that everybody is talking about, I don’t think there is a good roadmap right now to move that kind of workload on the device.
Most of it is happening on the cloud, and they have this notion that doing it all on the cloud is not sustainable from power processing and other aspects, They have to bring it to the device, but how do they do it? What is the framework? When?  I think that’s still out there. It’s a journey.
Leonard Lee:
The journey, yeah. Definitely plans out there for bringing Genia and the device in a consistent way.
Prakash Sangam:
But this is a fast moving industry, and we’ll see how that turns out.
Okay, so getting back to the comparison and so on. So, and I agree with you, benchmarks only explain you so much. There are many other things that make or mark any device or any processor and so on. So, but just double clicking on the performance. So, you know, XElite by many benchmarks, it’s meeting or beating M3 and M4 just came out. We don’t know a lot of details about it, but purely on the claim tops perspective, I think M4, they claimed 38 tops.  And I think by definition, XElite, which has 45 tops,  beats on TPU TOPs number itself. And it beats Intel Meteor Lake by large margin.
And I know that Intel’s latest, Lunar Lake would be a good comparison to XElite. So, and all of these benchmarks are good, but if you really look at the competition, the primary target for Qualcomm would be Intel OEMs, right?
Leonard Lee:
Oh yeah. I think people missed that point. Yeah, I totally agree with you. And I think where the comparison to Apple really makes a lot of sense is simply because Apple with their Apple Silicon portfolio has really shaken the world. 
And what Qualcomm is doing is they’re proving something that Arm has game in the PC market, right?
I think, and that really is that big proof point that we’re looking forward to in the second half of this year as these devices armed with the X series portfolio of processors come to market.
Prakash Sangam:
Yeah, absolutely. And kind of Apple Silicon is a proof point that what the best you can do with Arm. And of course, Apple, because of its vertical integration, can do many things much faster than an ecosystem like Windows. But the comparison is inevitable, but really in terms of market competition, primarily, at least in the beginning, will be with Intel-based Windows laptops, right?
It seems Qualcomm is targeting primarily the enterprise market, and even otherwise, too, usually for a player to be really recognized as a major player, you have to have many models with different configurations, different price points, and so on, which was not the case so far. I know you had a few PC SKUs, not configuration variability, and so on.
But if you listen to many of these players, look at the leaks and the rumors, it seems that it’s not just one or two, but it looks like many OEMs with many different models are being planned to be out second half of this year, mid-second half of this year with XElite. That also gives a bigger importance to Windows on Arm this time.
Leonard Lee:
Yeah. No, I totally agree with you on that. And the interesting thing is that with a single skew, when you look at both the XElite and the XPlus, each of them cater to different tiers, power tiers. You have, let’s call them an economy or efficiency and then a performance tier. And so those are different TDPs. And that lends toward a diversity that Qualcomm can support with its OEMs without having, let’s say, a hundred different skews to support all the different variations that an OEM customer might want.
And so, I just wanted to highlight really quick the introduction of XPlus. Because we were originally introduced with the top tier XElite, right? And that kind of blew everyone away. One of the things I think, you’re making the comparison with the M series. There were comparisons to M2 as well as 3.
But the thing is, when you look at the details, the XElite starts off with a really unusual configuration, which is an all-performance core configuration, where you have 12 performance cores. And when you look at what Apple is doing, it’s basically your traditional E core, P cores. You have a mix of that depending on which particular skew that you’re looking at. But when you look at the X series, we’re just looking at performance cores, even with the newly introduced Plus. And the only difference being between the Snapdragon XElite and the Plus is that the XElite has 12 cores, right? Two that can turbo boost, right?
And then you have Plus that has 10, but then both of them have 45 tops of NPU. So, you know, obviously Qualcomm feels that both of these processor classes need to support, you know, the minimum requirements of what is perceived to be AI PC computing, right? And, you know, they said basically both are for premium tier.
Prakash Sangam:
I mean, it’s like sub tiering within the premium tier, right? And it will be interesting to see the pricing, right? Obviously, the question will be premium tier. But I think what will be interesting to see is how they are priced vis-a-vis Intel based counterparts. I mean, if you historically look at it, they were the same or slightly higher with whatever few models Qualcomm previously had with its 8CX.
But with Orion and XElite, it’s a totally different ballgame. So yeah, it would be some inkling, but it would be really something that I cannot speak on publicly. But I think that is one thing that everyone will be looking at.
Leonard Lee:
Yeah, yeah. I mean, the way I looked at it, and we’ll probably have to follow up with our friends at Qualcomm at some point about this, is that the Plus was really for high, high mid-tier kind of thing. It was more of a portfolio expansion move, which is interesting because when XElite did come out, and Intel announced the Core Ultra, it was a little bit of a head scratcher because the Core Ultra came in kind of light in terms of the NPU.
I mean, the estimates are 11 tops, which is low this Microsoft standard for what is minimally required for a processor or a laptop to be, or a device to be classified as an quote-unquote AIPC. So, you know, I thought this was more of Qualcomm, you know, just expanding the breadth of devices that they could go to market with early on. Because, you know, initially, yeah, I think a lot of us were thinking, you got to start from the top. This is quite literally a new category, right? They’re coming to market with probably the top, top of tops, if you will. So why not just go wide?
Because, you know, there was the announcement of Lunar Lake, right? They’re going to come to market with 100 total or what Intel calls platform tops. And then it’ll probably have close to 45, 40-ish NPU tops of AI compute. They’re kind of approaching the market in a different way. You know what I’m saying? It’s almost like a reactive thing. Because, you know, quite honestly, I think the Core Ultra probably comes in a little bit light compared to competition even AMD is coming to market with, right?
Prakash Sangam:
Yeah, I agree. And I think probably when the Core Ultra design started, AI was not that much of a hype, maybe. So it was probably hard for them to change midway. So that’s the reason, you know, the Lunar Lake is coming very quickly after a Meteor Lake, right? So it’s not a standard time frame, right? So it’s a course correction for them after Meteor Lake.
Leonard Lee:
But the fact that a lot of these companies are pivoting very quickly based on media headlines and just hype sentiment is a testament to the fact that this is an NPU game. This has a lot less to do with CPU, which I think folks were used to considering or benchmarking or prioritizing in terms of a compute IP that informed their purchasing decision.
Now, all of a sudden, it’s an NPU, and it has different implications in terms of how it can be designed in and scaled out, right?
Prakash Sangam:
Yeah, correct. And one thing interesting was the NPU capability in terms of tops, at least, between XElite and XPlus are the same. So I think that says that in the AIPCs, obviously, the NPU will play a key role, right?
So as I mentioned, I have used Qualcomm 8CX-based laptops quite a while, and the biggest challenge, the biggest by far was app compatibility. And I think with Windows 11, that has improved a little bit, and then Microsoft has been consistently giving this messaging that we will make sure the 64-bit emulator performance is almost similar to running natively any app on the processor itself and so on.
But I think, especially in the enterprise space, app compatibility will still be a concern and an issue. I mean, if you’re an IT manager, if you’re deciding on buying thousands of laptops for your employees, even a couple of apps that are key to your IT ops are not running, then that might put some questions in your head whether this is the right time for you to take on this platform or not. Right? your views on that?
Leonard Lee:
Yeah. I mean, I’ll generally agree with you. I mean, we’ll find out, like you said, at Build, what degree of, let’s call it, endorsement and support and let’s say enablement Microsoft has provided Qualcomm this go, right? And like you said, they’ve been working with Microsoft for a really long time. But a lot of folks are talking about enterprise. I think Qualcomm has a great opportunity with creators, especially mobile.
I think there are a lot of folks who are, let’s say, social media creators or folks who are mobile, even in the broadcast industry, who might find this kind of powerful compute very useful in the field, right? Especially as you look at tools, like you were at Adobe, I was at NAB, and I checked out all the new stuff from Blackmagic Designs. Everyone seems to refer to DaVinci Resolve for all the cool new creator AI stuff.
And indeed, a lot of the capabilities that these ML capabilities, classical stuff that the NPU can enable is pretty impressive. We’ve seen some of the demos and benchmarks, not only from Qualcomm, from others, where the NPU actually makes a difference, especially in terms of power efficiency. And so I think that’s another audience to consider. Enterprise, I totally agree with you. It’s crowded already. It’s already entrenched. It’s going to be a tough market to squeeze into. And a lot of the silicon that’s coming into the market over the next six months is going to reach parity, comparative parity with each other fairly quickly, right?
And so, but I think because Qualcomm has demonstrated that they have promising chops on the CPU end of things, they have a viable competitive alternative, arm-based alternative that CIOs can contemplate.
Prakash Sangam:
All right, so in terms of competition, we talked about Intel and Lunar Lake, but if you hear the chatter and the rumor mill, it seems the Microsoft exclusive deal with Qualcomm on running Windows on Arm seems to be ending end of this year or sometime next year, which means the Windows on Arm concept itself will be wide open.
I know for every Arm player out there, which means MediaTek, Samsung, HiSilicon, if it gets its fab things in order, and there are some rumors about even AMD looking at using Arm and so on. So it’s kind of the same competitive profile in smartphones, and it’s good and bad, right?
Good, because Qualcomm has consistently won there. Bad, because so far they had exclusivity and they’re the only player in the game, but now they have to see competition, especially in the time where our Windows and Arm seem to be getting some traction. It’s an interesting dynamic.
Right?
Leonard Lee:
When it gets to desktops, I think a lot of the benefits that the mobile computing brings to the table starts to diminish a bit. I think AMD or X86 will probably continue to find at least a pretty solid defenses from the arm assault. But yeah, I think the portable or what some call the mobile PC space, that’s going to get really interesting, really fast, especially on the ultra mobile end of things, and on the creator side of it. Enterprise, I think that’s just going to be really a curious sector or battlefield.
That’s where, like you were alluding to, Intel is pretty well entrenched there. It might be more difficult, I think, even for Arm players to make inroads and gain ground there. If they do, it’s going to likely be still pretty hard fought territory. You know what I’m saying?
Prakash Sangam:
Yeah, of course. I’m interested to see what Lenovo, Dell, and HPE do with their enterprise laptops for this Ultra Premium. remember, it used to be called Ultra Mobile PC. I don’t think Qualcomm is right now focusing on the desktop, maybe at a later stage if they see success here, but that’s their battlefield, and that’s where they’re concentrated there, and you know, these three OEMs come up with multiple SKUs, multiple models, different configurations, matching what they do with Intel. I think that gives a very strong signal to the enterprise market, the CIOs and others. They are serious about it this time, and it’s worth taking a look at them.
And it’s not just the performance that matters, right? I mean, in the market development support, subsidies and others play an equally crucial role, and Intel has almost perfected this art, and they have a very close relationship with the OEMs, a lot of deals and so on.We haven’t seen a lot of, at least in the public domain, signs of Qualcomm trying to address that market need, if you will.
It will be interesting to see how Qualcomm will play that game and what will be Intel’s reaction and so on, right?
And obviously, as OEM, since you’re so close relationship with Intel, that’s your cash cow, you have to be balanced in how you hedge Intel versus Qualcomm, how much you basically promote your products and so on versus the other, right? They have to keep that balance all the time.
Leonard Lee:
Yeah. And I think people forget how big Intel is, especially in client computing. You know what I’m saying? Even though they’ve had some challenges on the financial and market side of things, they’re still really huge. They have massive infrastructure that just simply can’t be discounted and they probably do a lot more software than people give them credit for. But at the same time, PC computing is changing. A lot of that, I know that a lot of folks in the PC world don’t want to hear it, but Apple’s driven a lot of that. And there’s going to be a big question now, like what you were saying earlier, Prakash.
So we have all this NPU compute now. What do we do with it? And there’s some hints as to, I’ve seen some generative AI applications and functions and tooling that look really useful. I mean, they actually help the creators. They don’t just do this single shot or double shot junk creation. It’s like literally, oh, wow, okay, this is helpful. Then running those iterations, right, of let’s say using a generative tool in going through what I call a sort of a generative creation flow, there’s iterations. You’re going to be racking up tokens pretty quickly. And so this thesis I think will prove out useful again for certain users. And I’m becoming increasingly convinced that it’s for the power creator, because the things that I’ve seen where it’s actually done something wow, is in the hands of people who are super pro.
Prakash Sangam:
Yeah, I agree. So in some of the examples I have seen, you talked about creating, right? It’s like text to image or text to video tools. When you start off the big initial creation, probably you need cloud to do that. And then small changes here and there, change the color, change the shadow here and there. It takes the same amount of time if you’re doing it on the cloud. So once you do the major creation, and then you download the full image, vector image on your laptop, and then doing iterations, changing a few things, and fine-tuning, if you will, it makes sense to do it on the device for many reasons. One, of course, how quickly you can do it. You don’t have to burn a lot of tokens, as you mentioned, do it all on the cloud, and efficient for even the service providers.
And if you’re a CIO, when you’re buying laptops for your enterprise, you’re looking at, if I’m buying now, can this still be useful as it is now in two, three years? Not just today. Every application out there is going AI, so you have to be considerate of how much AI, horsepower, whatever you’re buying now has to do the task not only today but of the future as well.
Yeah. I think in summary, from my point of view, I think it’s an interesting time. And it looks like a lot of stars are aligned for Qualcomm from ecosystem, from their performance, from what’s happening in the industry and the need for AI. They have a fair chance of getting a crack at it. What do you think?
Leonard Lee:
Yeah. I mean, I have to say that it will likely be a lot of hard work. And so if they’re able to execute, I’m sure it’s certain that they’ll get a piece of this AI PC pie. And if anything, AI PC, this hype has become a catalyst for really positioning Qualcomm in a really special way in the PC space. That arm can be used in highly performant PC computing. And I think they’re very well poised to prove that exact thing in PCs.
The thing that they’ve kind of struggled to do in the past, they’re bringing that NuVIA IP through Orion as well as their AI legacy, deep AI legacy, as well as ML, all that AI compression technologies that they’re into as well. Because if you think about what they do on modems, they do some insane compressed models and model compression. So all that stuff works out really well for them. And so across the edge, edge AI, I don’t know, they’re going to be hard to beat.
Prakash Sangam:
Yeah, the apps that they have for AI and so on, right? AI Hub.
And I think to just close it, I think there are two things that we have to watch out. First one is Microsoft Build. A lot of things I think will be announced and discussed there. And Computex, which is basically the event for a PC ecosystem. So I think it will be interesting to watch out for what will come out of those two events.
Leonard Lee:
Yeah, I think Computex is going to be really big this year.
Prakash Sangam:
We should definitely come back and discuss and maybe look at the same questions and what we talked about and how they turned out after those two events.
Leonard Lee:
Yeah, I’ll be there. I’ll be there for the whole week.
Prakash Sangam
That’s great, Leonard. As usual, always a pleasure discussing such interesting topics with you. It seems there are a lot of exciting things happening from Qualcomm side in the next few months. And maybe we’ll come back and discuss again sometime in the near future.
Leonard Lee:
It will be a pleasure. Thanks for inviting me on.
Prakash Sangam:
So, folks, that’s all we have for now. Hope you found this discussion informative and useful. If so, please hit like and subscribe to the podcast on whatever platform you’re listening to this on. I’ll be back with another episode putting light on another interesting subject.
Bye bye for now!
 If you want to read more articles like this and get an up-to-date analysis of the latest mobile and tech industry news, sign-up for our monthly newsletter at TantraAnalyst.com/Newsletter, or Read our Tantra Analyst Insights Articles.