https://videocardz.com/newz/dell-confirms-xps-laptop-with-nvidia-n1x-at-computex
Hopefully, we’ll see software updates arrive faster as more SM121 systems get deployed in the field.
Maybe this also explains why DGX Spark updates have been so slow — most of DGX Spark engineering resources may have been focused on N1X.
I think it will be similar to DGX Spark pricing by taking out the 200GbE QSFP, we shall see today.
I just hope they don’t block the Spark from being able to use Windows. Obviously the Mellanox will likely be broken but don’t really care about that on Windows. CPU & GPU path is pretty much high likelihood if not guaranteed to be “operable” if not blacklisted. PCIE paths to SSD I don’t see any reason why it wouldn’t unless they really adjusted the PCB that much.
There are very useful AI applications that can be done there without having to deal with the annoyance of Wine & Proton.
Windows ARM does not seem to be a highly usable system; Linux has better support for ARM and offers more tools
Based on the leaks chips are designed for a 45W to 80W power envelopea lot lower than the GB10 inside the Spark. That power envelope is for the CPU+GPU and not GPU only. Hopefully the systems housing the N1X won’t get as hot as the Sparks. Tomorrow (6/1/26) the embargo is over and we’ll find out more about the N1/N1X.
I’m not sure what you are arguing about. Linux has a smaller subset of applications to begin with - “better support” on Linux is mostly just derived from that, if you’re in the supported territory it’s fine outside of it it’s unsupported which is to be expected.
Windows provides inline x64 emulation (worse performance than Apple’s prism of course) though for a general user that already makes it leaps and bounds past Linux (where you don’t get the breadth of applications Windows has because of legacy not to mention needing to set up the roundabout FeX setup to emulate x64).
45W to 80W probably expected since you are working with ~100W battery. That’s part of the reason I want to be able to use the Spark for testing. The N1/N1X will be power neutered.
I was thinking of picking up a second GB10 system in order to split workloads (not run bigger models). The N1X, I have to admit, intrigues me. Having portable local inference would be kind of nifty. Price point is going to be an important detail for me, though. If they’re coming in at Macbook Pro 128GB/4TB prices, though (roughly $6M), I’ll just grab that Acer Veriton GN100 4TB device.
Seems to be slotted in below the performance power of the DGX Spark. Will be interesting to see if they’ve resolved any of the memory bandwidth issues we have, as they are advertising 1 petaflop of AI compute, when we cannot even get near this on our more powerful DGX Spark. Or are they expecting 4/ts to be useful on a laptop…
I think memory bandwidth is still the same in N1X. This is the mobile version of DGX Spark so I do not have high hope regarding the performance. On the bright side, we will have more units based on SM121 so it will not be abandoned and we will have software update. As long as software stay updated, hardware will be useful.
I’m glad my Asus is within the return period; I’m packaging it up and will be returning it and waiting things out.
It’s not the laptop necessarily that inspired this - just the fact that in 2028 the Vera Rubin with LPDDR6 will be out and that’s what the Spark should have been. If consumer RTX sparks come out in the ~2k range, then I’ll grab one of those as i don’t need the awkwardly expensive ConnectX interface as these things just don’t scale well anyway.
Also, I was slightly enthusastic about deepseek flash on spark but after doing the math, it costs more in electricity per million tokens than the raw API costs and i work on open-source software so i don’t need the privacy for those tokens to drive the excuse to run slow and steady locally. I can use qwen3.6 35 a3b on my GPU to do any “stuff i don’t want on the cloud” which is mostly just nothing thse days as work has mandated ai/controls for everything work related now too
It is interesting how quick they’re pivoting to Windows.. . Windows on ARM isn’t as bad as some people say - the Surface pro X is a nice piece of kit.
So for me, it’s less the laptop - maybe more about the rebranding to RTX desktop - but seeing that its impossible to ammortize how much of a rip off this current gen is if Vera Rubin with much faster ram is just around the corner or cheaper RTX desktop variants will be slightly faster to compete in gaming and end up running the small models better anyway
It is difficult to predict the future. The safe bet is using the cloud based AI to meet your current development needs. Local AI really shine when you need the security and privacy. OpenAI/Google/etc… can change policy at anytime. Local AI will stay the same once you set it up to meet your requirement.
It really comes down to degree of control over your development environment and deliverable. Many people think Hermes agent are great but I only use it to experiment. When I want a dependable tool, I use Google ADK to create it. This way, I know exactly what tool will do and pin point the error when it failed.
they market as Spark but used Cloud Model for Demo, sorry but 1PF will also be the next lie, only the bigger install base will be good but how many people will feel scammed if they dont get what they expect ?
EDIT: there wasnt only Laptops, also Desktop have been on the stage… i wait for the people thinking they run Claude on that :D
No need to predict the future, Nvidia laid out the plans for refreshes in 2028 and 2030. The marketing slides show Vera Rubin LPDDR6X in 2028 and the RTX Spark Desktops if they don’t ship with ConnectX, they will be a few % faster (because they use slightly higher timing ram) and quite a bit cheaper.
Why no LINUX laptop? :-(
I expect we will see linux on these in no time. My hunch is that Nvidia doesn’t like being in the consumer business anymore and is happily letting Microsoft lead the overhead of that but Ubuntu and RHEL and other Linux variants will be around.
The main CPU and GPU will run Linux, obviously, but the big question is if the manufacturers will surround that with supported devices (wifi, networking, [touch]screen, HID, USB).
The thing which might make me potentially interested in this would be if one of the laptops configurations repurposes the PCIE lanes from CX-7 to Thunderbolt. That would in theory allow an eGPU with something like a 3090 to run a hyper fast draft/speculator model.
There’s a RTX Spark desktop to go along with the notebook. See details at NVIDIA RTX Spark — Slim Laptops & Small Desktops
I already have a DGX Spark derivative, I am also interested in a RTX Spark laptop or better tablet with some network connector like
to use the reduced powered AI RTX laptop on the go alone, whereas at home I would like to connect it to my DGX Spark derivative to run a more powerful cluster :-).
A cluster example could look like the AMD Strix Halo cluster via PCIE x4 in this video.
What do you think about this? Should or will this be possible?
CM
Any idea or guess on when the DGX Spark will be refreshed? (2 years from launch… Oct 2027)? The RTX Spark desktop will be missing the ConnectX-7 NICs?
This would be a cool dev laptop on which to run Ubuntu. I’m not really interested in running Windows. However, I foresee some big problems with this idea:
The embedded controller in laptops is usually a black box. This is a problem because you can’t control things like fan curves, power states for the GPU or max battery charging level from the Linux side, which is critical on gaming laptops or laptops that are plugged into AC power for extended periods (otherwise you destroy the battery). Also, sleep/resume might not work.
Devices like touchpads, LCD backlights, WiFi, and sound might not have any drivers.
Sound: Even if the sound chip is supported, sometimes the laptop uses speaker amplifiers connected via i2c that have no Linux driver, so you get no sound out of the laptop unless you use headphones. This was a huge problem with ASUS laptops for a while.
Assuming the manufacturers are aware of these problems and have selected common laptop components that are Linux-compatible, these laptops would be a great option for development on-the-go. Taking a portable DGX Spark on the plane would be pretty awesome.
I think they expect you to run linux stuff under wsl in those N1X windows machines.
