Windows ARM laptops

[xiaofan]

I think RTX Spark replaces the N1X/N1 code name and not the same as DGX Spark already, even though they are related.

DGX Spark, AI Workstation, not that really to the consumer market, runs Linux. US$4699 for 128GB/4TB version.
Edit: now got Windows on ARM version
https://marketplace.nvidia.com/en-us/enterprise/personal-ai-supercomputers/dgx-spark/

RTX Spark, consumer or business, runs Windows, initially more for laptops. But later it can be for AI Workstation desktop and mini PC, business/consumer laptop/Desktop.

Initially the price may well be on the high side, say from US$2000 to US$5000 (top end model with 20 core CPU, 128GB RAM, 4TB SSD), more for business and AI enthusiasts.

Later the lower end models may come down to US$1000 to US$2000 market segment say in 2027.

 

[stanlawj]

I'm not the only one who thinks RTX Spark laptops are going to be crazy expensive.

 

[joesph]

microsoft.com/ultra

 

[stanlawj]

- 15" Mini-LED touchscreen display
- Up to 2000nits of peak HDR brightness
- Large haptic touchpad
- Nvidia RTX Spark (20-core ARM CPU)
- Blackwell GPU (6144 CUDA Cores)
- Up to 128GB unified memory
- 1 petaflop of AI compute
- HDMI, 3x USB-C, 1x USB-A
- SD card slot | 3.5mm Jack

 

[stanlawj]

Will be DOA.
Too much power consumption especially to run AI agents.

 

[stanlawj]

https://wccftech.com/rtx-spark-vs-m5-in-clang-54-percent-faster/

RTX Spark Gives Decent First Impression Of Its Performance In Developer Workloads Benchmark, Beats M5 By 54%, Marginally Slower Than Base M5 Pro​

https://chipsandcheese.com/p/analyzing-nvidia-gb10s-gpu

Analyzing Nvidia GB10's GPU​

FP64 is critical for scientific computing. Looks like regular Nvidia RTX gaming GPU is just better.
RTX Spark GB10 will only be suitable for lower precision AI inference workloads.

 

[stanlawj]

Nvidia's Blackwell GB10 is built on TSMC 5nm process node. Runs too hot.
The next gen GPU will be a big jump to TSMC 3nm node. I will wait for RTX Spark Gen2 for it use TSMC 3nm node, estimated 2028 latest. In the meantime, Apple M5 Max is still much more useful.

 

[xiaofan]

Some more info from Microsoft Build 2026, other than Surface Ultra laptop.

New Windows on ARM64 machines
1) DGX Station (Desktop) -- mostly for Enterprise, AI workstation
2) RTX Spark -- laptops, mini PCs and desktops, for business and individuals

https://blogs.windows.com/windowsde...dows-as-the-trusted-platform-for-development/

• Introducing Surface RTX Spark Dev Box – purpose-built for developers powered by NVIDIA RTX Spark silicon, delivering up to 1 petaflop of AI compute paired with 128 GB of unified memory shared across the CPU and GPU. Comes with all of the above developer optimized Windows 11 experience so developers can build, test and run AI and agent workloads locally without setup friction or unpredictable cloud costs.
• Introducing DGX Station for Windows — the world’s most powerful deskside AI supercomputer for developing and running agents on Windows — powered by the NVIDIA GB300 Grace Blackwell Ultra Superchip. It is purpose-built to develop and run up to 1 trillion-parameter frontier AI models locally, as well as connect always-on, frontier AI agents to enterprise applications and workflows, coming in Q4 this year.

 

[xiaofan]

A positive review from LTT

 

[xiaofan]

I am actually pretty happy with my Acer Aspire 14 AI laptop (2025 model, at S$1000 plus a bit) with Qualcomm X1 26100 CPU, 32GB RAM and 1TB SSD.

I find that I use this one much more often than MacBook Neo.

Take note my use cases at home are more like a software testers even though I cannot really code, as I am quite involved in some cross platform open source projects. I do not play games or editing photos/videos, rather I am mostly using apps like Chrome, Windows Terminal, Visual Studio, Notepad++, WSL, MSYS2, Python, CMake, etc.

There are some issues but not a big problem.
1) Oracle VirtualBox does not really work as of now. I have to use another Acer Swift Go 14 laptop 2024 model (dual boot Windows 11 and Ubuntu 26.04) to use VirtualBox and a few Linux/BSD VMs. I also have a few lower end mini PCs to run Linux and BSDs.

2) Some devices got no Windows ARM64 drivers. Luckily not a big problem for me.

 

[xiaofan]

I set S$1500 as the budget for laptops and mini PCs (eg: Mac Mini). So I guess I will not really look at RTX Spark laptops or mini PCs any time soon.

One potential interesting thing is whether MediaTek will ship the ARM CPUs in the RTX Spark without the nice Nvidia GPUs, for low cost Windows ARM64 laptops, to compete with Qualcomm Snapdragon C series and Intel Wildcat Lake CPUs for lower end laptop market.

 

[dev_stg_prd]

A positive review from LTT

what's the use case for an average joe at home? unlikely no rite?

 

[xiaofan]

what's the use case for an average joe at home? unlikely no rite?

Indeed, initial launch will be more for enterprise, or tech enthusiasts who want to run local AI workloads.

But later there should be lower cost versions for average users with a 12 core CPU, 16GB/32GB RAM (up to 64GB), targeting consumer market

https://www.guru3d.com/story/mediatek-confirms-cheaper-nvidia-rtx-spark-chips-with-twelve-cores/

MediaTek has confirmed that Nvidia’s RTX Spark platform will include cheaper processor variants, expanding the lineup beyond the previously detailed high-end configuration. The confirmation was given at Computex, where the company stated that multiple SKUs are planned, including a physically smaller chip with up to twelve CPU cores. The twelve-core version is expected to use eight Cortex-X925 performance cores and four Cortex-A725 efficiency cores. Compared with the twenty-core model, this reduces both the CPU configuration and the expected platform power range. The listed specifications point to 2560 CUDA cores, 16GB to 64GB of LPDDR5X memory, and a TDP range of 18W to 45W.