RTX Spark: el PC d’IA de NVIDIA serà assequible?

I gotta show you the most beautiful part which is video games. It is also the closest to our heart. G-Forces and Videos, greatest marketing campaign. You could raise your lobster in here. Don't care, give my 32 gigabytes of DR5 for $90 back, please. Slap is wasn't affordable price tag, then we'll talk. I paid $6 for butter today. If you thought that in video was out of touch, you know, they were kinda at this level before, but now, Nvidia's out of touch level is getting off the charts it's insane. All of the CPUs until now were created for people. We were the users. Compute is revenue now. Compute is profit. Look at his stock price. I announced... I announced a product. Look at his stock price. Recently at Nvidia's keynote at GTC or it's technically basically just compitex. A long time ago Nvidia used to be a GPU company. Video was a gaming company and video was a GPU company in the past, but if you take the transcript of this, it is 38 pages because the keynote is like two hours long. So I'm gonna scroll all the way down. He said AI 95 times. Now doesn't even include all the times. It was just on screen shown to everybody. But the craziest part about this is that Nvidia did announce a consumer product, something that consumers, normal people apparently are supposed to care about us humans. All the CPUs of the past, we built for humans. This CPU is built for agents. But us humans are supposed to be spending thousands of dollars apparently to buy one of these to run our agents. In video things that this reality, everybody can afford it. I could totally imagine that someday there's actually an AI supercomputer in your house.

You have to have it in your house just like you have a home theater in your house. You have steriles in your house. You have game consoles in your house. Where do they think this money is coming from, dude? Most people are struggling. Computer prices are insane. And then they come out with this and like everybody's gonna have one of these in their home. And what's funny about this announcement is this is a consumer-facing product. And in the good old days when they just started talking about AI at 10, that was like two to three years ago, they still used to front load their presentations. by showing consumer tech, like gaming chips and everything. But nowadays things have freaking change. I think it was like an hour, 20 minutes into it before they even started to mention anything about consumer products. So that's that out of touch level. It's kind of going off the charts. Okay, but what is this thing? Because honestly, there are certain aspects about this new device that Nvidia is making that are honestly pretty cool. Nvidia RTX Spark is what they announced. New laptop chips. and they're partnering with a bunch of different companies to work on this. One of the big ones is especially Microsoft because they're going to get these to run on Windows. You might be wondering, is like, okay, Nvidia's already been in the laptop space in the past. I have a gaming laptop that literally has an Nvidia GPU in it with what's the big deal here. Well, this is a completely different system because this is an Nvidia SOC, a system on a chip, which means the CPU and the GPU are smooshed together. and they're just one big system. And this is kind of relatable to like what Apple does or AMD has been doing even Intel has been doing on their chips. SOCs

are really popular because they can be very power efficient, but they also have downsides of not being very modular, and if it breaks, your system's pretty much just screwed. There is leaks of this chip already called the N1X. I think this is a code name for the Nvidia chip. This is the exact same thing that ended up releasing here on the RTX Spark. This is an Nvidia GPU combined with an Nvidia CPU. Nvidia has been making CPUs in data centers for quite a while, but on this product here they're partnered with MediaTek, an ARM processor company, and that CPU is 20 cores. There's a lot of cores going on there and apparently they're ultra efficient. I think 10 of them are high power and 10 of them are low power. So to seem 20 cores off the rip doesn't make me think that this is gonna be a cheap processor by any means, but you know, maybe I'm proven wrong. Maybe this is gonna be a super budget friendly processor, especially with up to 128 gigabytes of memory. But then the GPU on here has 6,144 CUDA cores. And if you're to relate that to desktop GPUs, this is also based on the black wall architecture. It's equivalent and Cuda Core account to the RTX 5070 on desktop. The thing about this is an RTX GPU that is capable of gaming. They say right here it's game ready with DLSS and Reflex and everything. This is double a 7 by the way. The new double a 7 game I'm looking forward to playing it. I look a little bit like him.

But it seems like a lot of the examples that we've seen of this thing performing in playing games, they're trying to hide the performance of it. They don't allow anybody to show FPS numbers. That's a big, big no-no. Right here, they just say it runs at 1440p over 100 frames per second. But we know in video. And video likes to show things like a frame generation as real performance. And video likes to use DLS as upscaling as the same thing as all true performance. Here it is, running at 1600p resolution with ray tracing and DLSS 4.5. This is playing so smooth. So I wouldn't really be surprised. You know if Nvidia's faking these numbers a little bit. This is supposed to be like a consumer product, but you're not seeing these in any gaming devices right now. The desktop RTX 5070 draws somewhere within like 250 watts, and I don't think that they're gonna be able to cool that many watts. Especially if you've looked at a lot of the devices that this is coming in, it seems like they're gonna be powered with just a USB-C cable, which usually maxes out at 100 watts. So it's gonna be severely power limited compared to the desktop 5070. I expect maybe like 65% of the performance 70% of the performance and one of the major limiting factors it might be because it's on an arm base CPU. We've seen how other arm-based Windows processors end up running. I'm not 100% sure how this is going to perform, but at least we have seen that it has been working and what's kind of cool is we did see Fortnite running, which does use anti-cheat. So they're able to get Windows on arm running through anti-cheats as well. Our check spark supports games running on direct-tox, open-GL, or Vulcan.

running even multiplayer titles with anti-cheat like Fortnite. I think that's pretty cool. I'm not 100% sure how they did it or if that's gonna work with every game. And even here's showing that like different creative apps like Adobe is optimizing specifically for this. Another thing with the power consumption they do quote all day battery life because probably as well because the processor, the CPU that's on it's gonna be pretty efficient. And this looks pretty cool. The processor looks like it could be really really capable. But it makes me wonder is like, who is this for? Because in video they want you to run AI agents on this. Maybe the AI agents will play the video games for us. Wouldn't that be fun? Microsoft and Nvidia are going to reinvent the PC. This is going to be the new PC. Where everybody's just gonna have an AI agent that can get into all of your programs and run them for you to do things on your computer for you. Apparently they could do things better than a human ever could. Today we're able to now present these Kudax libraries to agents who can use it much more effectively than even humans. And if you thought Nvidia drivers were revived coded, well now they're going to be vibe making their chips as well. One of my favorite use cases of agents is chip designers. People talk about AI reducing jobs, complete nonsense. In this running on your local machine just instantly brings up a ton of concern. This thing would need more control, more access. I don't know if we're in the position where we actually entrust a company like Microsoft to have that kind of control. And I think this sentiment is shared by a lot of people. Especially when you see things from Intel as well, I know this isn't Nvidia, but... Brain, Inspire, Computing, so that they can train AI's to model a human brain in reason like a human brain.

I'm starting to wonder what this whole human aspect is even going anymore. Whatever happened to the idea of a personal computer that's your computer. And video specifically compares the personal computer, the PC, to phones. This reinvention of the computer is as big of a deal as the reinvention of the phone into what we now know as the smartphone. And they've wanted to push this idea that everybody will be using these agents all the time. But in the future, you just pick up your phone and you chat with your agent. You're chatting with your PC in the future. That you have your own personal one just running at your house constantly. But do they seriously think that this is going to work considering how expensive these things are going to be? You see, Nvidia constantly talks about how this thing has up to 128 gigabytes of RAM, which means it's able to run large local models on it that are capable of doing a lot of things. But have you seen RAM prices? I mean, y'all are the ones causing it. So let's look at some things to get an idea of how much this whole revolutionary new idea of personal computing, how much this is going to cost because forget gaming and everything. But we're going to first look at the framework desktop PC. Now this can be specced out with AMD's AI 9 plus max 395 or whatever and this is like the really beefy integrated GPU. If you just spec out this device with 128 gigabytes of RAM, the overall price. To get this machine, which doesn't have a laptop screen or keyboard and stuff, it costs $3,500. And here's another one with the same processor in it, the AI Max Plus 395. It says 128GB of RAM, surprisingly, from ASUS here, is only coming out to a reasonable $3,000. Honestly, this one's a little bit cheaper than I thought it would be for 128GB of RAM.

just the RAM alone is probably like a thousand dollars probably more. And I mentioned it earlier but this core, this RTX Spark Core, is basically the same thing as the N1X processor. What's funny is we've already seen this in a device from Nvidia. It's called the DGX Spark. And this has 128GB of memory. It runs basically the exact same chip. This has already been released. 4,000 on a 700 dollars. But you can get the DGX Spark at a staples. And it's the exact same price almost $4,700. And I swear to God that they charge you for shipping, man. So yeah, everybody's gonna be able to afford this processor. And then when you take a little bit deeper, you see the devices that this is being put in. The new Surface laptop Ultra from Microsoft. And I'm like, I do like the trailer for this. It does look pretty freaking cool, but this is a very high-end premium device. I would be surprised if this one comes out to be kind of cheap. When we look at as well Dell, Has the new XPS line with the RTX Spark in it? And these guys, there's no way these things are gonna be cheap. Now the XPS, especially the high end ones, are very, very freaking expensive. Ace's has also been shown off the ProArt P16 and P14 that have these RTX Spark chips in them. Yeah, these are also very, very expensive laptop, especially when you start piloting on RAM, on RAM, on RAM, and it. So Nvidia seriously thinks that this whole AI PC that the PC is just built for the AI, why would it be built for the human?

We were the users, we were the renters, we live in a world counter by seconds. They think that this reality is coming when these devices just to be able to run it because you need that much ram in it is going to be that expensive like who is going to be able to afford this like who are these selling to? It's crazy it just feels like the normal person and I think we've been seeing this for years but the normal person can completely abandon. on this, like we used to just be gamers, like give us our freaking gaming GPUs. And you see like the comment sections on these, people are absolutely pissed off that just want freaking affordable computers. And the craziest part to me is in video is also actively making them more expensive. How you see in this freaking server, all of those chips that are by those, I think these are the various CPUs on their data center, all of that is RAM that we could have, but they're putting it in their servers. And then hogging up so much RAM has led to even budget and devices getting absolutely gimped. A lot of companies now are putting out competition to the Macbook Neo, and these are Windows devices that are stuck with eight gigabytes of memory. I don't know if you guys have experienced the nightmare hell situation that is Windows on eight gigs in 2026. You open up like a one browser window and your entire system just crap out, bro. Hmm, good luck. So these laptops that probably cost thousands of dollars from Nvidia are taking all of the RAM. They're using LP DDR5X. It's the same RAM that would be going in or more budget laptops as well. Not to mention too. Videos said this. The orchestration of all this runs on a CPU. This is the entire harness and the CPU's orchestrating all of the work. In the age of agents, the CPU is now a bottleneck to GPU utilization.

Directly affecting token throughput. So let's talk about their data centers CPUs. They're very stuff. All this agentic AI and everything is going to be more and more CPU heavy than it has ever been in the past. As AI is now starting to take over your entire PC they call it this aggregated AI or something like that. This this aggregated this distributed this heterogeneous computing problem is precisely the reason we built Our next generation which means that this is probably going to lead to a CPU shortage as well We've seen rumors about this Intel has been talking about in their presentations too Everything needs to run through the CPU the CPU manages what the agents can do and to me I'm like Shit CPUs really like the last beacon of hope in this space so you could still get a pretty affordable price on but they just gotta take everything don't they the supply chain we created for Vera Rubin is twice as large as Grace Blackwell. In a video set that this is in twice a production of what Grace Blackwell was so their last generation of data center chips. So that means that twice the production, things that is already limited like TSMC can only make so many of these chips a fabricator. That more of their production, this is where all of our GPU shortages are going. It's not just memory, but it's also all of the fabrication capacity is going to make data centerships instead of gaming and consumer hardware. It's just another slap in the face and then also wanting to say that they're making this consumer product and everyone's going to have it just absolutely blows my mind when you see all of the other priorities of the business. I mean, Nvidia now is not reporting gaming revenue as its own separate revenue source. Then Nvidia is actually completely abandoning.

The personal computer, which is disappointing because this is something I imagine like this whole AI trend or grift that Nvidia has done. It's not really a grift at this point. I don't think they're going back, but Imagine this didn't happen in this reality and video is just a competitor in the space. This would be a really really cool product. In videos making a CPU that competes directly with Intel and AMD, the two dominant players in the Windows CPU space. In videos also the only company that I think will be able to pull something like this off that they have enough muscle to push in and have developers also supported their chips. So, apart of this is like, oh this is actually really cool. and it will probably drive competition from Intel and AMD, but at the same time it's like Nvidia's someone doing this and I just I don't really trust Nvidia anymore. Am I crazy to say that? But I will say something that might help the situation. So if you were to want to get one of these RTX Spark. laptops, or even these home PCs that it would just sit at your home and run AI models all the time, that poor PC that AI slave, that the 128 gigabytes of memory is gonna be the big limiting factor. And if you were to just use this chip for gaming, not AI, I don't think you would need that much memory. But I have no clue what kind of SKU is there gonna be offering? Are they gonna be offering like a 32 gigabytes version of this? I would think that they would, but I'm not a 100% sure. And we can kind of break down what?

Might be possible there just haven't gone in the too much detail on this yet. My Cron makes LP DDR5X memory the stuff that's in the RTX Spark and this at least from them comes in these different capacities and 120 8 gigabit capacity So each chip on it has 128 gigabit which is 16 gigabytes of memory and there has to be 8 memory chips on the RTX Spark. They might have a 96 gigabyte one 64 gigabyte one 48GB, 132GB, or even a 24GB model. But the thing about this is what's really cool about the RTX Spark. It's a lot like Apple with their SOC's or AMD with their SOC's as well. The chip can run shared memory, which means this memory can be divided between the CPU and the GPU, your system. and your V-RAM. They can change dynamically based on which one needs more memory at the time. So if you wanted more V-RAM for your games, you could dedicate more to it. I mean, desktop graphics cards are still eight gigabytes of memory. In 2020, 6. So technically, if you had these and you could dedicate more of the memory to the GPU and possibly get away with doing more things on your system. But that does also limit about how much memory that you would need or would be reasonable. 32 gigabytes of memory. It's usually fine. You can dedicate 16 to the GPU 16 to the system, or maybe 12 to the GPU and then 20 to the system. But I think the sweet spot would probably be like 48 gigabytes of memory because you'd be able to dedicate 16 to the GPU and then stuff 32 for your system memory. And to be honest, a 48 gigabytes system still sounds like it'd be pretty expensive. We look at the prices on like max and stuff. And the others are still, they're getting nearing like $3,000. So maybe there'll be some reasons

no memory configurations out there, but again they're not gonna be very cheap. In judging by the fact these aren't really gonna be put in gaming devices, at least yet because the cooling is not gonna be very substantial. We see things like the DJX Spark can go up to like 150 or 200 watts, and that's using the exact same chip that's in this. So these laptops would be able to do more, but as you see, they're gonna be allowed to do more. Maybe they would sell a variation that would be more built for gamers. Don't know if Nvidia is about to say anymore, but maybe this one wouldn't have 20 CPU cores it'd be more cut down. But the major concern with these is are that you're gonna even work for things like gaming. Because Nvidia, yeah, we've seen it run games, but they haven't talked about performance numbers at all. You would think if the performance was very impressive in games that they would be bragging about it right now, even though they don't really care about gaming, that I think they would start to talk about numbers to get people hyped up. So does concern me if these things can even run games properly. Because the big thing about Windows on ARM is it is not Windows on X86. Games are built. for X86 Windows. Qualcomm Snapdragon, their chips still struggle a lot with gaming. So if you're to see the latest ones that came out with Snapdragon on Windows is the X2 series. In GPU score at least in 3D Mark, this is a synthetic benchmark which utilizes the GPU to its force potential. You can see that it's really not that bad. It seems to stack up GPU-wise to like an Apple M5.

or the new Intel Arc GPU that came out. But the thing is like it doesn't look that bad here. But then when you hop over to actual gaming, how does this perform in real programs? An in Cyberpunk, a 1920 by 1200, so about 1080p with quality upscaling. So it's upscaling from like about 720p up to that. It loses a lot of performance. Like it was pretty close to the Intel Arc GPU before. And now it's pretty far behind it. So this at least with the Qualcomm chips on Windows. It can have a lot of and this is led to what I think a ton of issues because this was as whole windows on arm push that Microsoft has already attempted and now they're gonna be doing it within video this time but these Snapdragon chips at least when they at first announced and launched in 2024 Snapdragon X failed Qualcomm sold 720,000 PCs and Q3 around 0.8% of market share and apparently their goal was to get 2% of the global PC market share. I don't know how they're doing now now arguably Nvidia doing their arm push as well, making their chip with media tech. I do think the Nvidia would be the one that would be able to pull it off. Especially judging by how long they've been working on this chip just to get it to run on Windows. They've had so many freaking issues. Now, Moore's Law is dead if you want to check out some of this stuff. He's gone into a lot of the history. of N1X, the same tip that's in RTX Spark, where it's been delayed after delayed, this chip was supposed to be announced last year. But that would also give me some confidence that Nvidia has worked out a lot of the kinks and is going to be introducing something that genuinely works pretty well. Now Nvidia has invested an Intel and is going to be making GPU.

Chipplets that would be going on to Intel processors as well. So maybe Nvidia is also hedging their bets to see if this whole arm push on Windows will end up working out. But Nvidia did announce one thing that is technically for gamers, though they are introducing it into other applications as well. But this is DLSS 4.5 Ray Reconstruction. So Nvidia is obviously been on this push for ray tracing for so long now. Been like what? eight years plus at this point, just trying to get ray tracing to be runable. And basically this ray reconstruction is a denoiser that tries to fix the issues with ray tracing. His ray tracing as a baseline is very, very noisy. And it looks like this one is just gonna look better. Generally with the super resolution, it ran worse using the new second generation transform model. So I wouldn't be surprised if we get worse performance, but better visuals. But at the end of the day, like, who else really using ray tracing? who can afford the GPUs that can do this stuff? But what do you think is in video's idea here? With this whole multi-thousand dollar laptop. And I think everybody's gonna be running this thing. I think it's absolutely insane right now. How do they expect everybody get a 128 gigabytes of RAM? Who the hell is this for? At the same time, the RTX Spark does still look kinda cool. But let me know what you guys think about it. I think this is the most out of touch thing. I've seen these AI companies do now. It is crazy. That's gonna be for me though. This is some goofy shit. I know I'm not gonna be by one. That's crazy man. That's so expensive. So I'm gonna see you guys in the next video. Hopefully you have a good one. And uh... peace.