7:09
So, Nvidia just launched the new laptop chip. It's called the RTX Spark. It's a 20-core ARM-based CPU and it packs an RTX 5070 GPU that has the same core count as the desktop version. It goes up to 128 gigs of unified RAM and it's being used in very thin laptops, as thin as 14 mm and it still has all-day battery life and it runs the full Nvidia stack. So, CUDA, DLSS, ray tracing, all of it. Now, I've never had any doubt that if Nvidia wanted to make a laptop chip, they would make a good one, one that's energy efficient, powerful. There was never any worry, but what I am worried about is what they want you to use this chip for. Because if you watch their presentation, the first third of
it or maybe even more, they were talking about agents. So, we got to talk about this cuz this is the main pitch. If you know what an LLM is, like a large language model. So, let's start there. Large language models. So, ChatGPT, Gemini, Claude, things like that. They are like tools that you can type to and they'll type back or respond back, right? Agents are the next step of that. Instead of just responding back to you, they can actually do things for you. So, as an example, you would ask it to like draw an owl and then make it fly while changing the camera angle. And the agent will go and perform the steps required
across your actual apps, checking its work as it does it to complete the goal. It would kind of be like a derpy little intern that lives inside your computer that can do stuff for you. Now, keep in mind there are pre-existing tools today that can accomplish stuff like this, but they definitely require like a little bit of knowledge and they're tools that you have to kind of piece together that typical person wouldn't want to do. Now, Nvidia's bet, along with Microsoft, is that this will be the new era of PC and they're calling it exactly that. The idea is that these agents would run locally on your device instead of the cloud. And to do that, you would need a lot of unified memory. Because to have a
genuinely capable agent, you would need a fairly large model that's sitting inside fairly fast RAM. And like a regular laptop can't do that. Not even a gaming laptop because there's not enough of that super fast RAM in gaming laptops. So, a MacBook Pro could technically do because you can get up to 128 gigs of unified RAM there. Same with an AMD Strix Halo laptop. That has a lot of unified memory, but neither of those two devices have access to the CUDA software stack. And CUDA is the thing that virtually every AI tool and framework and model is built on and optimized for first. So, the RTX Spark with its 128 gigs of unified memory has access to this CUDA stack. It's great.
It's like purpose-built for this vision. However, this is where things get a little shaky. Because if you think about it, this is not Nvidia's mountain to climb. It's actually Microsoft because they're the company that has to build and implement the part of Windows that the agents would live in. Like this is baked into the operating system, would run from your taskbar. It needs access to your files, your apps. It has to have a lot of access to your system. And if you think of Copilot, most people are actively trying to remove Copilot from the system because they find it too obtrusive. Can you imagine like Copilot extreme? This thing would need more control, more access, and just more It's
more capable, right? So, it has to have the ability to do all these things. And I don't know if we're in the position where we actually entrust a company like Microsoft to have that kind of control. But, it would need it to pull this off. Now, if they nail it, it would be incredible. Genuinely incredible because now your computer could do these multi-step tasks that we find boring and time-consuming. But, at the same time, I also can't imagine Microsoft making a tool that's powerful like that, but wouldn't be obtrusive somehow. But, we'll see. Uh the second thing that the RTX Spark chip is supposed to be good at is with creative work. So, it is a 5070 with a lot of RAM and Adobe apparently has new
architecture for Premiere and Photoshop engineered specifically around the chip, which is kind of cool, but we'll have to see what that's all about. Now, the third thing is gaming and to be clear, it is the third thing because if you look at the list of laptops, these are not gaming laptops. You got like Lenovo Yoga, Dell XPS, ASUS ProArt. These are creator or productivity first laptops that can also game, but Nvidia's throwing some pretty interesting numbers out there. We will have to test all this stuff down the line. They might even be great for gaming. Now, in terms of where I stand on the whole RTX part, I think the hardware is very interesting.
Uh I think that pricing, if you actually are interested in this thing, the pricing is not revealed yet, but I imagine we should brace for it to be quite expensive because just the market is that way right now. Uh the other thing though is just like the vision of the new era of PC. When you think of the complexities around this, like forget about the technical element of it. That's already extremely difficult, but even if they could pull it off, the kind of social acceptance of having a tool like this baked into your operating system, that's a huge ask. I think in this day and age, like if that thing was available today, I don't know if I would want that on my system. Like it's neat, but when I've kind of pieced
together myself and I know exactly where everything is, like that feels a little bit more comfortable, but if this is built into the operating system and it can do stuff when you're not even there, I don't know. It's It is strange. Um if there was one company though, if there was one company that has the kind of clout and reputation and just size to pull it off as a partner, it would be Nvidia. They're super rich. They have a ton of connections with multiple industries, AI, gaming, like productivity, workflows, like so much stuff stems around Nvidia. So, maybe they could pull it off. Who knows? What do you think? Is this something that you would If you think about it, like what it's capable of and the examples they've
given, is it something that you would actually use and or want on your system? Okay. Hope you guys enjoyed this video.
