This is the Mobile Tech Podcast, brought to you by worldpodcasts.com. Now here's your host, Tech Girl Miriam Joire, brought to you by Qualcomm.
Hi and welcome to the Mobile Tech Podcast. I'm your host Miriam, and today is Monday, June 1st, 2026. We're in Taipei for Computex, and I've got Don McGuire here of Qualcomm. Hi, Don. How are you?
I'm good. Good to see you.
I'm very excited to have you on because there's been a lot going on with compute, and I feel like people just forget about compute that Qualcomm is so involved in. You guys have been doing this—I've been coming to the show on and off for over a decade, and I remember being here when you did 835.
And then, you know, 7cx and 8c and 8cx, and here we are now with the X series and X2, and now you've got this Snapdragon C. So, I want to talk about all that.
Absolutely. Yeah, let's do it.
Fantastic. So, I think there's some confusion out there about the positioning for Snapdragon C, right? It is not Oryon-based, it is Kryo-based.
And so, can you give us an equivalency of where this fits in, like if you were to compare to a mobile chip? The reason I'm asking this is because the elephant in the room is MacBook Neo, right? And I want the audience to understand, because a lot of people are making the parallel between mobile chips and compute for laptops nowadays because of that.
Yeah. And it's not necessarily correlated, right?
So I would start by saying, look, we're always looking for places in our portfolio to expand the market, right? And so we've gone all the way up to X2 Elite Extreme, where we've got some amazing devices from ASUS and other partners coming out, and the performance is just industry-leading. It continues to be industry-leading, even through this week. But then we look for opportunities in the market, and one of the opportunities that we saw was that some of our competitors and some of the options out there have sort of gone away in this $300 to $500 range. That's where historically Intel Celeron or Pentium and some of the AMD processors have lived, and they've pretty much abandoned that part of the market. So we were able to develop Snapdragon C. It's a very efficient SoC solution for laptops in that price range.
And so it's not really Neo compete, right? It's really about that entry level $300-ish—that sort of entry-level student, young student, streaming, doing basic things on your computer, first-time computer buyers, those types of people. That's where C lives. And like you mentioned, it's not Oryon-based, it's Kryo-based. So we are taking a mobile architecture and we're enhancing it for the C platform, and then putting it into some very cost-effective designs. And we've got partners that—
I saw the Acer Aspire Go 15 yesterday. They actually brought it out for me as a pre-production unit so I could take photos of it being in the glass case. Honestly, I think this is a great price point for a lot of people. My question to you is, you know, I've been using the MacBook Neo because I'm a Mac user and I wanted to see if it could hold up to my workflow with essentially last year's binned version of their smartphone chip. So for them, it would be the equivalent for Qualcomm of taking an 8 Elite base and converting that into something, then putting it into a laptop. The performance I'm getting there is good—it's matching my MacBook Air M2 from four years ago roughly, you know? But I can't see myself going lower than that. So I'm kind of wondering where, when you talk to the OEMs, they see this Snapdragon C fitting in. Because for me, the fact that I can still do DaVinci Resolve 4K video edits on a Mac is kind of amazing. Do you think that'll be possible with something like Snapdragon C?
Well, actually where we see sort of Neo compete is really with our Snapdragon X platform, right, which is our entry level on the X series. It is Oryon-based, and it's got some serious performance and is actually much better.
And the price points will be similar.
And there's also devices right now available, right? From HP and others that are much better value.
The OmniBook 14, that's one of my favorites.
They're much better value. They actually do more, they perform, they're touchscreens, and they're fully capable laptops with NPUs and AI capabilities that the Neo doesn't have. Right? And so if you're looking for value for money and performance in that price range, there's a lot of devices that exist, and that's really where X lives.
Got it. And of course, we have X Plus and we have X Elite which goes up to X2 Elite Extreme. So the question to you is more like, because that's one of the things that Qualcomm is so good at—you have building blocks that you build your chips around, right? You could take a Kryo-based architecture like the 8s Gen 4 for mobile and say, "Hey, we're going to prioritize and add more NPU, and we're going to add more decode chips for streaming." So now you still have a super-efficient chip, but it's much more cost-effective. I'm wondering how much of that modularity you were able to apply to Snapdragon C to make it still usable? Because, as I said, I have a hard time going below Snapdragon X in my mind in terms of my use cases, but at the same time, now we're looking at $300 and we don't want people to suffer, right?
No, we don't. We want to create affordable solutions, but that people can use. And so that's really where C lives. We don't think people are going to be disappointed if you're looking to buy a $300 PC, right? A two-in-one, clamshell laptop, whatever it is. We think that you'll be really excited about the performance.
Because you've been able to modularize the chip?
Whether it's GPU, we've been able to create a value proposition that we think fits within the price point and within the use cases that people are going to want at that price point.
And from a marketing perspective, could you just have taken an 8s Gen 4 and told the Windows manufacturer OEM partners, "Hey, use this as a laptop chip and call it the same"? Would that make sense?
It might make sense if you're only looking at that vector of it. But you know, we have to make it work on Windows, right?
So there's some changes required.
Yeah, there's changes required, there's customization required. We don't want to just recycle for the sake of recycling. We have to make it work for the use cases and for the users. And so there's things we had to do, obviously, to make Snapdragon C work for that value use case.
I was wondering about that, right? Because there's also a marketing angle—you don't want to use the same name potentially, right?
No, absolutely not. Exactly. It makes it smarter.
Because there is bespokeness to each chip that we make. But you know, we want to scale, right? We are now, as you mentioned, the history of us and compute—
I have been such a proponent of Windows on ARM and ARM-based compute for lightweight use, bare bones, and also thin, light, and portable. And the fact that you guys have 5G built-in is another option that is available to Qualcomm products. That to me has been very exciting. And I hate to say this, but as a Mac user, when they went ARM, I was like, "Oh my god, this is exactly what I've been wanting." I still have my Microsoft Surface Pro X, which has the SQ1 chip that you guys collaborated with Microsoft on.
SQ1, thank you. I still have that, and you know what? With Windows 11, it still runs really great, which is proof that Kryo-based is totally fine. So I'm super gung-ho. I'm so glad I got to see the evolution of this. Now my question to you is, as an aside, Google Book—you guys mentioned you're going to be a part of that?
I'm curious: Snapdragon C for that?
No, they're looking at our entire portfolio.
Yeah, because there'll be different price points for Google Book. And so, you'll see X2 Elite-based products, you'll see X Plus-based products, and Google really wants to have a full range.
Got it. Right, this isn't the reincarnation of Chromebook, right?
Right. This is a brand new category. And because it's heavily based on Gemini, it's going to need some powerhouse. And as AI becomes more agentic, with those workloads, it's really important that not only are you going to generate those tokens in the cloud, but you can bring workloads locally because of cost. We're here in Taipei, and we've had this long journey of compute. Today, Cristiano is going to be on stage with his keynote talking about '26 as the year of the agents in agentic AI and the workloads. You'll see some of the comparisons of the tokenomics of agentic AI. One of the things that people are going to start to realize is that as you use agents to accomplish those tasks, you're going to generate tokens, and those tokens cost money.
Eventually—and this is why you see people stacking up four Mac minis and trying to do inference locally—they don't want a huge bill at the end of the day from OpenAI or somebody like that. So they're trying to bring the workloads on-device or bring them closer to on-prem, to you. With X2 Elite Extreme, X Elite, and X2 Elite, we're going to show the differences between cloud-based and hybrid on-device-based workloads and the money that you can save.
And the time that you can save by actually putting workloads where they are most efficiently done. It's a small segment of the market right now, but as more and more people get comfortable with AI and using AI, and then using agents to do things for them, this is going to become important to them, right? Because affordability is always an issue, unless you're up there just buying several-thousand-dollar machines for business purposes. So it's a pretty exciting time.
No, I think it's great. A random side question on this before we continue on other topics: what does the C stand for?
Right? It's like that makes sense. Bread and butter compute, right?
Naming is tough, you know.
Yeah, I know. That's why we decided to go with that.
Do you think that this is more of a long-term platform that you'll continue evolving parallel to X, or is this just right now kind of solving an existing problem for the meantime?
Our intentions are always, when we introduce something into our roadmap, to have it live on, unless the market changes, right? And then we might have to change it to keep up with market conditions.
The other thing I want to ask about is, you mentioned Cristiano. I ran into him in LA at a Mercedes event recently, because I cover automotive and other parts. You have a big announcement this week at Computex here on data centers with Dragonfly. Can you briefly touch on that? Because AI is everywhere. This show is full of AI. Everywhere I go is AI. When I went to Mercedes and ran into Cristiano on a rooftop pool in LA, we talked about AI. So tell us a little bit about Dragonfly.
Yeah, so we're using the Computex keynote today to introduce our new data center brand, Qualcomm Dragonfly. There'll be more details on the brand, the architecture, and the roadmap for our data center solutions at our Investor Day coming up on June 24th in New York City, so you should tune in. But we're using this opportunity to tease the brand. It's a beautiful, amazing brand, and we're very proud of it. I would make an analogy of, in a sea of darkness of data center brands, we are going to be the light.
We're going to be the Jedi Masters of the data center. You'll see a little bit of a tease today and a reveal of an amazing physical manifestation, and then it'll whet your whistle for Investor Day on June 24th.
Fantastic. Super excited about Dragonfly. Tell us real quick: where do you position it? Because I don't know much about data centers, but I presume that Nvidia is a big player in there, Intel's a big player in there, AMD is a big player in there. Are you putting yourself in that realm?
Yeah. You'll hear this in Cristiano's keynote. We're talking about AI and agentic AI in the context of the compute continuum, and that goes from milliwatts all the way to kilowatts. It just so happens that we play across that entire continuum, from the smallest wearable device all the way through to a data center rack on the inference side. Where we're squarely entering the data center is in inference, and that inference side of the data center is becoming more and more important. People used to talk about training, training, training; it's really about inference, tokenomics, cost-effectiveness, TCO, and all those types of things. So we are making a really strong play across all three vectors: custom silicon, ASICs, agentic accelerators, as well as CPUs.
You know what gets me excited about all this is that Qualcomm started off as ARM—super power-efficient, as small as we can make it—and now you can grow, whereas the competition started big and can shrink. That gives you such an edge, and I kind of love that. ARM rules the world; I've been saying that for 20 years and people were not listening, and now they're listening because they're like, "Oh, MacBook Neo," and, "Oh, Snapdragon C," and, "Oh, Dragonfly." What else is coming up from Qualcomm in the next few months that we should be looking forward to?
It's going to be a busy summer. Right when we get back from Computex, we've got AWE, which is a big show for VR, AR, MR—all the Rs. And then from there, we go into our Investor Day on June 24th, as I mentioned. We've got a big summer coming up with some smartphone launches. Right on the heels of Computex tomorrow, we have Microsoft Build.
Oh yeah, it's tomorrow. Wow.
Cristiano will be making a cameo at Build tomorrow. We digital-twinned him, and he's just going to be in both places. No, I'm just kidding. But we need to do that one day.
That would be great when it's possible.
When it's possible, that would be great. No, he did a little pre-record with Satya. But yeah, it's just a really busy time of the year.
Yeah, and that leads directly into Snapdragon Summit that you guys are going to do at some point.
Snapdragon Summit is going to be amazing.
Fantastic. Well, I'm super excited.
Thanks so much for your time. I'm very stoked to see where Snapdragon C brings us. Folks, I've got some coverage on Computex on the Acer Aspire Go 15, which is the first Snapdragon C device announced. There's more coming. There's going to be some new Snapdragon X-based products announced. There is a desktop PC from ASUS we're going to see this afternoon, a little Mac mini-style thing with a Snapdragon X2 in there, I think X2 Elite. But anyway, check it out. Don, thanks again. It's a pleasure to have you.
All right. And folks, we're back with Domenico Lamberti. Am I saying it properly?
Close enough. Dom is here, MobileDom. We're here at Computex in Taipei, as I mentioned earlier in the show. Listen, a lot of stuff happened here. Some of it is exciting, some of it is a little odd, and I have lots of thoughts. I figured Dom and I can talk about this stuff for the next 40 minutes or so. What's your overall impression? This is your third Computex; this is my 10-plus, but I took a long break between 2019 and now. What's your impression?
This has been—I hate to say dull, because that's the wrong word, but slow. This is an in-between year for platforms, right? Usually most brands would take Computex as a time for a new motherboard chipset. So maybe AMD or Intel will be launching their next platform, so ASRock, MSI, or whatever will have 50 new motherboards.
Yeah, there'd be a collection of them.
There's nothing, because AMD is just redoing what they've done for the last year and a half. Nova Lake isn't ready for Intel yet. Nvidia has not got any 50-series refreshes. AMD's re-releasing a card from a few months ago that was China-exclusive. So in the big consumer computing realm, it's kind of been dulled down. And what's, at least for me, tried to take some of its place is big ARM vendors.
Yes. Obviously, we're going to talk about Qualcomm.
Qualcomm, MediaTek, Nvidia.
MediaTek and Nvidia, those are the two big ones. Which would you prefer to start with?
Probably with Qualcomm. Also, we should mention that outside of the ARM realm, there is Wildcat Lake, of course.
And that's actually very interesting because, as you know, I've been using a MacBook Neo for the last month or so for all my work, and it's recording audio right now. I use the microphone you're seeing right here for those of you watching on video. We finally have some competition that I think is meaningful—not the absolute $999 discounted at Best Buy for $699 kind of deal, and also not half-plastic, kind of metal, which is—I'm picky about the build quality and materials.
I think you brought up something very important that we should mention before we continue. I think you're right: it is a bit of a slow Computex. It is extremely hectic and busy with people, though. I think it's slow for PCs, for consumer stuff.
Yes, because the first three floors of the expo are data center, and AI is plastered everywhere. Look, I'm not naive, I'm not surprised.
But I really thought, this being the land of ASUS, the land of Acer, the land of HTC, the land of Foxconn, the land of TSMC, that we would be focusing on more consumer-based PCs. Now, don't get me wrong: data centers are obviously very important, AI or not, and ARM is eating the world, which I've been saying for 20 years, and it's now eating the data center world. Awesome, I'm on board. But I just feel like it does bring a different flavor to Computex, doesn't it?
Agreed. Again, this is my third Computex, but all the previous ones have definitely felt more consumer-focused, right? This one is different in such a way that it's not just that we're seeing server racks on the floor—we're seeing Jensen sign server racks in gold marker. We're seeing Delta showing off their new PDUs, power delivery units for racks. I remember my first Computex was when NVL72 launched. One of the things I saw walking around the show floor was a company—whose name I'm forgetting, it's not important—showing this water cooling setup which was gigantic. It was like three rack units wide and could cool 88 whole kilowatts.
And then two days later, Nvidia announced NVL72 kilowatt.
These power budgets are just ballooning.
It's crazy. But look, we're here to talk about the consumer-facing stuff we saw. Primarily for me, as mobile, it's going to be laptops that I'm focused on, but also I have an affinity for small, power-efficient computers, so maybe Mac mini competitors and things like that. The first thing I want to talk about, and the thing that—I'm three days into the show now, and the beginning of this podcast with Don McGuire of Qualcomm was recorded actually two days ago, so I've had time to digest our conversation about Snapdragon C series a little more, and I have mixed feelings here. My first thought is this: Neo here is about as low as I feel anybody should go.
Is that in a performance sense?
In a full computer usability sense.
I'm not talking about if you're just in Google Docs; a Chromebook at $200 might do the trick, even though the ergonomics of the keyboard, display, and trackpad are going to be compromised. So I'm including a bit of both. I'm talking about performance, but I'm also talking about the user experience as a whole. The fact that this thing has a good keyboard, a good bright 500-nit display even though it's not 120Hz or anything, a good trackpad, and arguably enough usable ports—it's bare bones, it does the job. And the highly integrated Apple ecosystem stuff in terms of software and hardware working together makes it possible for them to use an actual non-modified mobile chip. I mean, it's binned, but whatever. That to me is kind of the baseline, because if you go slightly below that, you start making compromises on build quality, materials, the trackpad, the keyboard, the display, and of course the performance as well. When they're saying to me, "Hey, we're making a chip that's targeting a $300 computer," my first question is: are these hypothetical $300 computers? Because with RAM prices and storage right now, are OEMs going to hit that $300 mark? And if you don't, how much worse is it going to be than a Neo that's going to be available refurbished from Apple for probably $399, right?
We were both in the same Q&A afterwards where someone said, "Are these going to actually hit 399?"
Yeah, and they said, "We hope so."
I understand; I'm not here to bash Qualcomm's perspective on this. They have no control over what the OEMs do, right? But it seems misplaced to me on Qualcomm's part to say, "Hey, Snapdragon X is what we want Neo to compete with. We want to make devices with Snapdragon X that are going to hit the same price point, build quality, and weight. We're trying to work with the OEMs to do that." And we might see some of that. In fact, we're seeing some of that with Wildcat Lake.
I want to get to that in a minute. But my big thing here is, I don't think you can do less. From all the rumors I've heard—and first of all, they didn't give us any specs.
Still, three days in. Now, I understand when I interviewed Don, it was even before the embargo lifted, so he had briefed me one-on-one in advance. But the fact that they're not willing to share specs worries me, number one. Number two, we know a few things: we know that it is running Kryo and not Oryon.
As far as I know, that's not officially confirmed yet. We know it's definitely not Oryon.
Okay, that's the most official confirmation I've heard, which makes me sick.
Assuming it's Kryo, I'm thinking to myself: if you want to get close to an A18 Pro on a MacBook Neo or close to X with an Oryon core, you might want to use something similar to an 8s Gen 4, which is the Cortex-X4 or whatever that is, and optimize that for use with Windows. You need to deal with I/O changes and a bunch of other stuff. But we're not even sure that's what it is right now—they won't even tell us. So I'm thinking to myself, if this hypothetical $300 laptop exists, how compromised is it in those other areas: build quality, keyboard, trackpad, display materials, even if the performance is, say, 80% of a Snapdragon X? Are we still wanting to sell laptops at that price point? They exist right now and they're all crap, right? And I don't want that to repeat itself with Qualcomm on the box. Why would you do that to yourself, Qualcomm? At the same time, every time I brought up Neo, they're like, "Oh, you can't talk about Neo and Snapdragon C at the same time because Snapdragon X is what competes with Neo." And I'm like, well then why does Snapdragon C even exist? Why don't you let Intel Atom do its job, or whatever it is they're using—Celeron or Pentium for $220 laptops?
Yeah, I think they brought back Pentium or something like that.
That's right. As an aside, Intel marketing people, if you're listening, you squandered one of your best brands. Pentium was an incredible brand. Why didn't you keep that going as your core brand? Sorry, pardon the pun. Anyway, this is what confounds me with Snapdragon C. I don't understand the marketing. I understand that they can make a chip that's Kryo-based that meets 80% of the performance of the Snapdragon X, and there could be room for that. But then when they give us this $300 price point, and when they don't tell us any of the specs, and we're not even sure if it's an 8s Gen 4 equivalency, which is the higher-spec current Kryo, or if it's something even lesser, then I start to worry about performance. Then I look at RAM prices and storage prices, and I'm like, why? Why announce this here without giving us specs? That's kind of where I'm at with Snapdragon C.
My main problem with it, aside from the fact that we don't know anything about it, is they're actively refusing to tell us.
Which means it's either too early in the cycle—which I refuse to believe because they've got products in booths.
So it's far enough along that OEMs are making things. The reason they don't want to tell us is because there's going to be backlash.
The Acer Aspire Go 15, correct?
Yes, which is another problem: Windows laptop naming.
Well, we could have an entire podcast about that. In fact, I will have you back on just for that. But look, the thing I'm trying to say is I played with it, I touched it, I used it. They wouldn't let me look at things like the settings. In fact, in the demo area at the Qualcomm event at the W today, they had one of those and somebody went into the settings, and they took it to the back room—it disappeared. And the reference design developer kit they had locked down; you couldn't do anything with it. I think they disconnected the keyboard and trackpad on it internally.
I saw a post from—I think it might have been Tom's Hardware, do not quote me on this—someone saw one of these, was able to get Task Manager open, and the codename string was 8c Gen 3. Qualcomm apparently brushed it off saying it was just an internal modifier, blah, blah, blah. You have to put that in intentionally. Also, until I went to an X Elite laptop, I used an 8cx Gen 3 laptop for five years. I won't say it was great, but I really enjoyed it. It had something that has never been replicated on Windows since: proper sleep-wake. I could use that thing on a Friday, get it down to 50%, close the lid, and pick it up next Friday and be at 45%.
Maybe efficiency is going to blow us away. Maybe this is a two-day battery life type of deal with the right battery. But again, I'm worried that the OEMs are going to compromise what this chip can even do by shipping at $300. You cannot make a laptop that's good at this price. I am sorry. There is a reason why there are so many people out there that go to Best Buy or Walmart and buy a $200 to $300 Chromebook or Windows machine and have the worst experience of their life, not to mention all the bloatware that comes with it.
You're going to have to subsidize the machine together.
And that's not going to change just because Qualcomm made a very power-efficient, decent chip with maybe 80% of X performance. Why would you go there?
When you're going down to these insanely low-priced products, your margins are essentially nil. Say we're using a 15.6-inch panel, just because that's one of the most common sizes and you can get them very cheap: that's going to be maybe 1080p if you're lucky, TN almost certainly.
And 200 nits at best, 250 on a good day.
Yeah. So you're going to have a plastic trackpad.
You're going to have a bad keyboard.
Horrific deck flex, correct.
You might get one USB 3 port if you're lucky.
It's probably going to be Type-C.
Yeah. But at this price point, you're probably also going to have a barrel jack for charging.
Oh god. Now that the EU is forcing it, one of those USB-C ports is going to have to charge the laptop.
With most laptops with a barrel jack, you can still charge via USB-C—
Probably at 18 watts or something.
It's malicious compliance levels.
Right, you're right. I'm not against the idea of making a more efficient chip, even if it's Kryo or whatever, but please give us some numbers so we can understand the marketing here. Give us some reason to get excited and not worried. That's how I feel right now about this.
Because we have nothing to go on, I'm effectively wondering if it is just a rebadged smartphone chip or tablet chip.
If it is, then you need to give us the analogy. Apple's not even giving an analogy; they're literally saying it's the chip, and we know it's literally the chip. You've already made an analogy on the X series. You've given us the analogy that the original X is essentially an 8 Elite, just basic Elite. Technically that was gen one, 8 Elite is gen two, and now X2 is gen three of Oryon. You immediately gave us an equivalency, with the caveat that of course you had to customize the chip because Windows and I/O, and I get all that.
That brings in my second biggest issue, and that's something as a user of a Windows on ARM laptop: Oryon is incredibly powerful, but the Adreno GPU drivers are horrific. That's one of the reasons—something we're going to speak about later with RTX Spark—that you don't need to have an incredibly powerful CPU if you can offload it to a capable GPU. Nvidia is coming in and saying, "We're still using Prism emulation, but everything that can be is now CUDA accelerated." Any NLE, any game, that worry is gone.
Yeah, Qualcomm doesn't have that advantage.
Qualcomm doesn't have that advantage.
I know, and that's a challenge. Here's the other thing that I'm worried about, and we're going to talk about Spark in a second: if you read the room—the room being the AI trend for better or for worse—if you read the fact that everybody's going ARM, including Nvidia... I have the theory, I joked about this, that 10 years from now, Intel chips will be one x86 CPU core for legacy with everything else ARM. They'll get to that eventually because it makes the most logical sense. But aside from that, unified memory: if they had announced a Snapdragon X chip or an X3 chip that was not faster than X2, not better necessarily, but had in-chip unified memory, then that's AI-forward, right? Because you're now saying you have 80 TOPS with X2 Elite Extreme that can be used by all the different parts of the memory, and unified memory is paying off like Apple has shown.
There is one big downside to that, though, and that is the cost of advanced packaging.
One of the reasons the X2 Elite Extreme costs as much as it does, even if you take out the fact that it's got 48 gigabytes of LPDDR5X-8533 or whatever it is, is that having the DRAM that close to the package requires advanced packaging. All of TSMC's advanced packaging is going to the AI data center people right now. It's so strained that companies are now actively going to Intel Foundry and seeing if EMIB is a viable alternative.
The more advanced packaging, the better, right? I would love to see an X2 standard or an X2 Ultra with on-package memory like Lunar Lake, but I just don't think we're going to see it purely because of the cost of every foundry process currently.
It's a good point I didn't think about. Well, that's a good transition. Qualcomm has done some cool stuff; I hope the C stuff ends up being okay. Let's move on to RTX Spark for laptops. This is going to be expensive, right?
We're talking $3,000 to $4,000 laptops at minimum.
If anything comes in under three grand, call me shocked.
Do you think this competes with MacBook Pro Max, or whatever the top one is?
The M5 Max or whatever. The SKUs are getting very confusing.
It's M5, M5 Pro, and then Max. Ultra only made it on the M1—
chips, right?
Uh, no. I think it's M2 and M3 got an Ultra, which was which is two Max dies fused.
Yeah, fused. Exactly. So, M5 only has Max. So, you think it's going to compete with M5 Max?
no, I don't think it's going to match it.
So, okay. So, we know the CPU in Spark.
I mean, C9500 type or better.
Uh, slightly lower than that.
So, this is MediaTek who did the CPU.
So, for those that don't know, RTX Spark is also called Nvidia's GB10, which is what we saw in the DGX Spark Supercomputer last year. It's 20 ARM CPU cores and a custom Blackwell GPU that is roughly RTX 5070 equivalent.
It's 10 Cortex-X920 cores,
and I think it's 10 A720s. So, they are quite powerful. The X925 or whatever it is, is power at all costs.
It's that single-threaded performance. It doesn't matter the die area, it doesn't matter the power, which is why they paired it with the A700 series.
They need to rein it in somehow, but at the same time, you've got 20 cores, you've got a big-ass GPU, you've got a lot of I/O, you've got this massive chip-to-chip connection between the two, which is the new NVLink Fusion chiplet, which apparently caused some issues early on, which is why Spark was so delayed,
Um, but I know how the Cortex-X cores perform.
They do not have a single-core advantage against—it might get to M3 Max level.
So how does MediaTek fit in this? Because I thought if Nvidia already had 20 cores of ARM, where does MediaTek fit in?
MediaTek has an incredibly diverse ASIC business.
MediaTek is not quite to the level of a Broadcom or Marvell, right? But you will find MediaTek-manufactured ASICs in routers,
TVs. I'm very familiar with their business, and actually, we worked with them quite a bit. I'm just trying to figure out, what is the mobile expertise from the smartphone world, which they're really good at, are they bringing into this mobile chip?
I'm assuming it's power management.
Power management. That's my assumption. And do you think like MPEG decoders and stuff like that too that might be needed on a—you know, like all that AV whatever, all that stuff?
Why—the only thing I can think of is you could—
No, so there's no modem on—
which is also something we probably could have said about Snapdragon, is that none of them have modems on them.
Why would you use MediaTek's encoder/decoding engines when you've got NVENC,
which on the Blackwell GPUs can then plug into Premiere?
So this is purely power efficiency.
The only thing I can think of is that, A, it's MediaTek's ASIC business; Nvidia can just farm it out to them,
Right? Interesting. Well, I'm happy for them because I think, you know, they need a win.
And I think MediaTek needs more visibility beyond Chromebooks. I have the Lenovo Chromebook 14, which is running that—
Kompanio Ultra, and it's really great. You know, as you said, you've been using ARM-based Windows for a while. I started with the Surface X, which I still have and I've updated to Windows 11. I was telling Don this earlier in the show, it still kicks ass. I can use that; it's not as good as my Neo,
but that's my point. If that is that okay, I can live with that somewhat, but that's also really nice hardware.
Yeah, and that's the thing. It's a five, six-year-old machine.
$300, I'm like—it gives me the willies.
Yeah. The only downside, and I say only, I'm actually quite worried for MediaTek in this case. If it doesn't perform well, it's going to get put on MediaTek.
Yeah, there is that. But I feel like, look, I think with the specs that they've kind of shown us and the kind of RAM it has potentially, everybody knows that power management—I mean, it's not going to run 2 hours. We're hoping at least to run like half a day or something, eight hours or something.
Six, you know, that'd be good.
From what I've seen, the apparent power levels are 30 to 50 watts with bursts up to 80.
Okay. So, I think we'll be okay. I just feel like you're right. Nvidia is not going to take the hit if something goes wrong. If something goes wrong, Nvidia can say, "Well, the Blackwell GPU works fine. It works fine. We gave this to them."
Do you think AMD—as I joked, and maybe we can lump this together now and then move on to Wildcat Lake—do you think that I'm right that eventually Intel will figure out that they have to go back to ARM, like they did with XScale, to really get the efficiency and the scalability?
I don't think they're going to go ARM.
I think they're going to do RISC-V.
I think they're going to buy one of the RISC-V startups.
It's still RISC. When I say ARM, I should clarify that—
I don't count RISC-V because it's different. But I think to me and in my—you know, there's simple—
instruction, simple instruction RISC.
So I'm like, okay, yeah, that makes sense. I think they're going to buy a RISC-V startup.
