NVIDIA Just Slapped Apple Silicon - RTX Spark

Nvidia finally did it. They just took the lid off of RTX Spark, a super chip that promises to take on device AI from an irritating button that I accidentally press on my keyboard to a deeply useful Agentic Collaborator. And while it would be easy to look at this demo that transforms a sketch and a mood board into a 3D animation and dismiss this entire concept as lap slops, guys, love it or hate it, this is a seismic shift in mainstream personal computing, and you're going to have to know about it. On the hardware side, there's no big surprises here. DGX Spark, Nvidia's pint-sized to desktop AI computer, was announced over a year ago and has been in user's hands since October. It's that, but in consumer mini PCs and laptops. So, we're looking at 20 ARM CPU cores, up to 128 gigs of LPDDR5X memory, and a powerful Blackwell-based GPU core with a little over 6,000 CUDA cores, which actually raises a pretty big question. Why did this take so long? I mean, if the rumor mill is to be believed, these things were supposed to be announced at Computex last year. Well, according to Nvidia, there was just a lot of work to do. Building the hardware, building the software, and building the partnerships to get it right. And it really sounds like they've done just that. In no particular order, here's what they're promising. Seamless gaming in Windows on ARM. Whether you're playing an ARM native game or using the Prism translation layer, you click play and you go. And that includes partnering with top game developers to ensure that DRM protections and antiche will not be an obstacle. Two is all day battery life in machines as slim as 14 mm with RTX 5070 class gaming performance even away from the wall. And third, of course, is AI. AI AI. Nvidia is extremely proud of the deep integration that they're building [music] into Windows in collaboration with Microsoft that will allow users to choose a model or models create boundaries like when [music] or if data is sent off device to the cloud and then turn them into a fast powerful local collaborator. And I don't know about you guys, but this is what I've been waiting for. I don't want to talk to OpenAI, but if my laptop can handle a 120 billion parameter model or realistically multiple smaller ones with a million tokens of context, I holy beans, guys. I can actually do something with that without sharing my innermost secrets with Sam Alman or having my wallet catch fire from an expensive subscription. So, let's dive deeper then and get hands-on with some of these shockingly sexy looking devices after this shockingly sexy segue to our sponsor, Asus, who jumped at the opportunity to feature their upcoming ProArt P14 and P16 laptops powered by the Nvidia RTX Spark. They boast all day battery life, vibrant 1600 nit HDR screens, and outstanding build quality. We'll have a closer look at these later. Let's start with the elephant in the room. This isn't the first time that Nvidia and Microsoft have announced that ARM on Windows powered by Nvidia is the future. Last time didn't go so well. So, what makes this different? Well, the obvious answer is 13 years and billions of dollars at stake. See, Apple isn't the only one who's realized that building a business that depends on Intel to consistently innovate [music] forever is dangerous. Another difference is that ARM has gone mainstream basically everywhere but personal computers. Your phone is ARM, your watch is ARM, and the data center you're being served this video from is using ARM. So software developers are more familiar with ARM and the tools that they use every day are better suited to compile for different architectures including ARM. The last one is that even for software that isn't compiled for ARM realtime translation layers which allow x86 software to run on ARM are everywhere now. Apple uses Rosetta 2 on the Mac and Valve's upcoming Steam Frame uses FEX. I've used both of them and they're awesome. And while neither they nor Microsoft's Prism are absolutely perfect, they've all come a long way. So most of your favorite applications are just going to run and you won't even realize that they're being [music] translated. The other big reason that I believe it's going to work this time is Nvidia. They don't always get things right on the first try, but if we look back at their track record on GPU compute, dating all the way back to the first general purpose computing GPU, the 8800 GTX, they've successfully predicted the direction the market was headed at basically every bloody turn. And this time, they're not even going out on a limb. On a hardware level, the CPU GPU comboed chip that's at the heart of the RTX Spark is very similar to another wildly successful product, Apple Silicon. Let's see. Efficient ARM CPU, check. Powerful GPU, check. Sophisticated interconnect and high-speed memory that enables them to work together at speeds far beyond traditional PCIe, check, and check. And like Apple, Nvidia is now a $5 trillion company with both the money and the clout to push their partners to support their vision. Here's how they laid it out. On the Windows side, that looks like new Windows primitives that deliver identity, containment, policy, and endto-end security capabilities. Then NVIDIA OpenShell allows users to define what agents can and cannot do, dynamically route queries to local models or the cloud based on the user's privacy policies, and even disguise personal information in queries that are sent to cloud models. Then there's the third parties. Hermes Agent and Open Claw have both committed to adopting these security and privacy layers in their new Windows apps. So the goal here is to take something that is already broadly adopted by developers and enthusiasts and then make the interface and the controls simple for regular users. So they can create agents that will execute tasks in Windows applications, reason through cross app workflows, generate images and video code plugins and apps and semantically search local files. One that's personally exciting for my line of work is the Adobe partnership. The video pipeline in Premiere is being upgraded to take advantage of unified memory and Photoshop is supposedly jumping from about 5% GPU accelerated to 100%. Whatever that means. Both applications are also hooking into the new Windows Agent Manager so that your agents will be able to interact with both of them. Now, this is the point where I had to stop the briefing for a moment and ask a question that I'm sure is on all of your minds. Hey, what about your desktop users? Some of us already have discrete GPUs that are much more powerful, not to mention more expensive than even a mobile 5070. Will their onboard memory be a hard limit? From what I'm hearing, pretty much. Within the constraints of what can fit in your local memory, regular GeForce users will be able to use many of these new goodies [music] and sometimes faster thanks to their superior compute. But there's no magic bullet that is going to turn a desktop 5080 with its 16 gigs of VRAM into a 100 plus gigabyte GPU. Since we're off the rails, I also took the opportunity to bring up Linux support for all of this. And while Nvidia is not announcing anything right now, apparently the driver team is very focused on getting the Windows launch rate. The Envy folks that I spoke to fully recognize the building momentum around Linux desktops and they aren't ruling out anything for the future. The other big announcement has got to be next generation DLSS ray reconstruction. Love it or hate it, Nvidia is committed to ray tracing. And like DLSS 4.5, this looks like a major improvement in fidelity thanks to moving to a second generation transformer model. It's also coming to real-time previews in Blender and the results there look really impressive. A final side note is RTX frame gen is coming to Comfy UI. So your AI generated videos can use AI generated frames in between your AI generated frames. Some folks will be very interested in this. I suppose let's go look at some laptops. They were not kidding. This is thin. This is the P16 Protart from Asus. And sure, not the thinnest laptop I've ever seen, but RTX 5070 class GPU in there. How'd they do it? Evidently like this through a combination of an extremely slim 99.99Wh hour battery. And man, those are skinny heat pipes. And is that also a vapor chamber? Yes. And there's the chip itself. Man, it's amazing how small it is. All your power delivery right there. CPU, GPU, all fully integrated into a single package. And up to 128 gigs of LPDDR5X. Got your house on my wrist right now. And I guess it shouldn't surprise me given the heritage, but the IO is pretty robust, too. This unit's got HDMI obviously, and then three 40 GB per second USBC's, 3 and 1/2 mil audio, and an SD card along with a USB type A. Of course, the 16-in isn't the one that I would have my eye on. It would be the P14. Same IO. Love to see it. A touch thicker. Actually, I can't tell. No, it's not. It does get a slightly smaller battery capacity, but I think that's a price I'd be willing to pay for that much power in such a compact form factor. It is substantially thinner than my current daily driver, and that's already a lot of performance in this form factor. This is Stricks Halo. Then in the next room, we've got the Microsoft Surface Ultra. Also incredibly thin for the performance on tap. She's uh she's running a little toasty, but this has been sitting here running games. First time we came through here was over an hour ago. And with that in mind, it's not even loud, dude. Seriously, can you Okay, my mic's right here. Can Can you even hear it? Mhm. Can you? Yeah. Yeah. Yeah. Like barely, though, I assume. Barely. Yeah. That's cool. Given how early this is, I wasn't really expecting to be able to just interact with it. Unfortunately, these are the only three that we get to see today. But across the board, this is the closest I've seen to, let's face it, the gold standard, which is the MacBook Pro, but on a PC. Now, for the uncomfortable stuff. Nobody is talking about pricing. and Nvidia's DGX Spark, which if you'll recall is basically the same hardware, not exactly cheap, especially now that it's gone up in price, and I can't imagine that bolting a 10 OLED display, a battery, and a haptic trackpad to it is going to make it more affordable. Then there's the CPU. Nvidia and their partners are talking a lot about efficiency, but any mention of performance is conspicuously absent, which is weird to me. I mean, the Cortex X925 is a perfectly competent CPU core that's built on TSMC's 3nmter process, and serve the home found that it can be comparable to AMD's flagship Zen 5, at least in quick and dirty benchmarks. But Nvidia's silence doesn't exactly project confidence in this regard. And then there's the last deafeningly silent moment during my briefing when I wanted to talk about power consumption. DGX Spark is designed for about 140 watts, which by my math would give us uh less than 40 minutes of battery under full load on something like the Asus P16. So obviously then Nvidia is tuning the laptop variant for a massively lower power profile. But how much lower and how much performance is going to be left? A mobile 570 is rated for about 50 watts according to tech powerup with notebook check noting that it can be configured for as high as 100 watts. Well, Nvidia says that I can expect similar performance. But during our conversation, they also told me that I would be pleasantly surprised by gaming endurance on battery. And they had kind of like a smug expression when I asked, "Guys, are we talking like two plus hours away from the wall?" I mean, surely there's some efficiency gains from the tight hardware integration, but something isn't mathing here. I guess that's a mystery for later this fall when we get hands-on with shipping hardware and software. For now, we'll be left wondering, wondering when will Lionus do the Segway to his sponsor, Asus, who's working on a full lineup of devices featuring the RTX Spark. This one right here is a ProArt P16. And it features ASUS's Luminina Pro OLED display technology doing up to 1,600 nits peak brightness with panone certification and a nice little anti-glare coating that's really going to help cut down on reflections. It is, as you guys saw before, shockingly slim for all the performance that it has on tap. And that was no accident. While a large haptic trackpad would normally add significant thickness to [music] the device, Asus is using a new solution that is much thinner. Asus designed everything about these machines for the most discerning of creators. From the large batteries up to 99.9 W hours on the P16 to the super slim liquid metal cooling solution to even the surface finishes that are actually wow really fingerprint resistant. If you guys enjoyed this video and you want to see Nvidia's AI hardware at a much larger scale, check out the recent data center tour that we did. It was frankly mind-blowing.