5:03
So Nvidia in1 in1x this is PC's Apple moment right? I don't think so. It's something it is a moment and it is historical and it is interesting and for Microsoft I think this is going to be what they hoped Qualcomm would be although the Qualcomm devices are for a different class of user. There's N1X and N1. Think of DJX Spark GB10 and a laptop. Now when we first got our DGX Spark I'm immediately a crazy person. We put Cyberpunk on it, but it's not enough to run Cyberpunk, which is meant for x86 on ARM through emulation. I also got DLSS frame gen and DLSS upscaling working on
it, and it worked really well. And so now we have Jensen telling us that it's going to be 5070 class graphics in a mobile platform. The devices that I went hands-on with, they are actually really high quality. They're really well built. Interestingly, the models from Surface, the Surface Ultra and some of the other models um were pretty similar. Asus coloring outside the lines a little bit, but 128 gigs, GB10 basically in a laptop format and Adobe good Adobe support. Adobe has always been a secondass citizen on PC, but we saw demos of the Adobe Suite Premiere running at a breathtaking speed on the laptop. Now, still it's a hardware encoder. So, one encoder, one decoder, but they say it's a fast encoder, but still
more horsepower is more better. I immediately wanted to do power benchmarks. They politely said, "Oh, please don't run that." You know, power config and that kind of thing. I uh wanted to copy the video drivers to my flash drive so that I could set up Windows for ARM on my DGX Spark. These are RTX Spark laptops and they said no. Um yeah, but there's going to be a 64 and a 32 gig configuration. And one of the promises was all day battery life. That's going to be an uphill battle. That's going to be the promise that they probably fudge a little bit on. I may end up eating my hat on that one. Uh eating my words, whatever you want to say. But I think they might be able to do that on the 32 gig version. But the
64 or the 128 gig version running full agentic stuff locally, I don't know. The other fun thing is that they changed Windows. Windows has been fixed for all of the bugs that AMD encountered on stricks Halo. So things like you can only allocate half the memory unless you go into BIOS and then you say, "No, I would like to allocate 96 gigs of memory." you're going to be able to dynamically allocate the memory between applications and whatever you got running on the GPU. So, that's nice to see that they dealt with that. I actually had a laundry list of questions from Mic for Microsoft on problems that I ran into on Stricks Halo and they fixed them. Not only that, I had a
follow-up conversation with AMD and I said, "Hey, did you know Microsoft is fixing all these problems that you had on Stricks Halo?" And they were on top of it. They said, "Yes, we've actually been working with Microsoft. We are also going to have day zero support for all the improvements in Windows to fix problems that Stricks Halo experienced first. So I think that where we're going to end up is an Nvidia class stricks Halo laptop. So like stricks Halo with the stuff that Stricks Halo goes with but on Nvidia's platform. and my HPG1A. Best case scenario, it's like six, seven hours of battery life. Get Nvidia get another hour out of that with 128 gigs of LP DDR5.
I doubt it there. You know, all day battery life I'm thinking is like 8 n hours. And again, maybe on the 32 or the 64 gig version, but I don't see it on the 128 gig version. The big commitment though from Nvidia is three generations of this. So, not just Grace Blackwell, but also Ver Rubin and then Fineman after that. And I think second or third generation is going to be where we really see what everybody's actually excited about. But for power users and developers and folks actually building things, this will be useful. NVFP4, a 4-bit format running really well. 4-ish bit format running really well on a portable platform. I think that's also going to be exciting and transformational. the rugpool. The thing
that's really interesting here is that the kind of the promises of co-pilot might actually be realized in having powerful agents that you're able to run locally and then that might actually trickle down to NPUs. So you build something on a laptop like this that could actually run on an NPU for other ordinary users. There might be a path there maybe. I don't know. I kind of doubt it, but at least for the NPU. Uh the other interesting thing is that Microsoft really wanted these to be co-pilot PCs. And so if you look at the engineering of it, there's not an NPU in GB10. Where's it at? What did they do to add it to the laptop chip? And if you look
at close-ups of the laptop chip, which I had in the Nvidia suite, um, it is a different chip. And you don't do a different chip unless you absolutely have to. So, they probably done something here for power or integration or anything else. And of course, the these laptops are not going to have Connect X7 or anything like that. So, the networking story here is completely different. but Windows on ARM and I think this is going to be the most amazing Linux laptop for the Nvidia ecosystem. But again, bootable USB, they were like, "Please, sir, uh, let's not do that." We're like 5 months away from launch. A lot might change between now and then. Nvidia is kind of known for that. So, you're going to need
several different large truck sized grains of salt depending on several different flavors of salt depending on which umbrella of functionality that you're looking at for this. But still, in a lot of ways, this hardware already immediately ticks a lot of boxes that were promised with a Windows Qualcomm, but now we're Windows on ARM Nvidia Edition. Oh, and the other thing with all day battery life. So, I did test Premiere like Premiere doing the thing, unplugged the laptop. Premiere did not slow down doing the stuff that depended on AI agents, agentic workflow, whatever you want to call it. However, the power usage also did not go down. And the thermal system here, I'm just eyeballing it. Looks like
about 80 watts, give or take. the amount of air flow through the fans and the level of fan noise and the amount of power that you're going to use lose from fans running at that speed probably 150 watts 99.9 watt hour battery 1 and a half two hour battery life but again they've got 5 months to fix it so we'll see so yeah if I had to call it right now a competex in 2026 I would say that it's 60 to 75% Nvidia marketing halo and 35% % Okay, maybe. I don't think it's going to be Apple N5 all day battery functionality. I just don't think it's going to be that functional in 5 months based on what I see here today. But I do think as a power user product for developers and people that are building things and people that need to be able to run 128
billion parameter models locally, it's going to be a good experience. It's probably going to be an experience really similar to what you can already get from 128 gig stricks Halo laptop but with Nvidia's ecosystem which may be the holy grail from your point of view or it may be a different set of headaches again depending on your point of view. So yeah, that's my quick hot take on the N1 and the N1X. I don't know. I feel like I'm kind of the odd man out here because everybody has been slathering praise on it and we've got five months to go and there's a lot that remains to be done before it would be worthy of that praise just for
the state that it's in now today. Well, you do know Nvidia that I could run Windows on ARM with the video driver. If you gave me the video driver for DGX Spark, it's going to be really similar to RTX Spark. You don't have to do any engineering there. I already know the PCIe part of it works if you do x86 emulation and pass through a GPU through the M.2 slot for x86. Just let's try Windows on ARM on GGX Spark. Like come on. Like when's the driver pack for that coming? And as long as I'm talking about Stricks Halo, there is one other fly in the arm for Nvidia and that is AMD's 192 gig version. The next iteration of Stricks Halo, which is uh what is it?
Medusa, you know, Gorgon Halo is going to probably have been launched by then. Maybe. I mean, AMD's already confirmed that they have 192 gig configuration, and we know they can ship it in laptops, and we know that workstation laptops like the HP G1A have sold like hot cakes. Might not be good news for Nvidia. We'll see.
