Meet the Microsoft Surface Laptop Ultra: NVIDIA RTX Spark and What We’re Testing

Microsoft’s Surface Laptop Ultra goes on sale October 16, and it’s one of the most interesting Windows laptops in years. It’s the first Surface built on NVIDIA’s new RTX Spark platform. That platform packs an Arm CPU, a Blackwell RTX GPU, and up to 128GB of shared memory into one package in a 15-inch laptop that weighs about 4.4 pounds.

We’re bringing the Surface Laptop Ultra into the Plugable lab, and we’ll be putting it through hands-on compatibility testing with our docking stations, display adapters, networking and storage products, and our TBT5-AI eGPU enclosure. This post covers what Microsoft and NVIDIA have announced, why the platform matters, and the questions we plan to answer. The results will come in follow-up articles.

Surface Laptop Ultra at a Glance

The Surface Laptop Ultra targets developers, AI and data science users, and creative professionals who want workstation-class GPU performance without carrying a workstation. Here are the confirmed specifications from Microsoft:

What Makes RTX Spark Different

Most Windows laptops with serious GPU power use a separate CPU and discrete graphics card, each with its own memory. The GPU is limited to its own video memory, which usually tops out at 8 to 24GB on laptops. That limit is a big reason large AI models have been hard to run locally on a laptop.

RTX Spark takes a different approach. The Arm CPU and Blackwell GPU share one pool of LPDDR5x memory, connected by NVIDIA’s NVLink-C2C interconnect. In a 128GB configuration, the GPU can use most of that memory. NVIDIA says RTX Spark systems can run AI models with up to 120 billion parameters locally. That opens the door to on-device work that used to need a desktop GPU or the cloud: running large language models, fine-tuning, local coding assistants, and processing private data that shouldn’t leave the machine.

It’s also NVIDIA’s first big step into Windows on Arm. The Surface Laptop Ultra pairs Arm efficiency with an RTX GPU that supports NVIDIA’s AI software tools and modern gaming features. For applications that aren’t yet native to Arm, Microsoft’s Prism emulation runs x86 apps.

Connectivity: USB4, HDMI and Magnetic Charging

For anyone building a desk setup, the port selection is promising. The Surface Laptop Ultra has three USB-C ports, all rated for USB4 at 40Gbps, and all three accept 140W charging. One of them includes Microsoft’s new Magnetic Connect, which replaces the older Surface Connect port with a magnetic USB-C connection. It also includes a full-size HDMI 2.1b port, a USB-A port, an SD card reader, and a headphone jack.

Microsoft lists support for up to three external 4K 60Hz displays over DisplayPort 2.1.

One detail stands out for docking. Microsoft’s specifications describe these ports as USB4, not Thunderbolt. The two are closely related: USB4 is built on the Thunderbolt 3 protocol, which Intel contributed to the USB Promoter Group in 2019, and Thunderbolt 4 is built on USB4. The difference is in the guarantees. Thunderbolt is an Intel-certified standard that requires certain features, while USB4 leaves some of them up to each manufacturer, including the PCIe tunneling an external GPU relies on (more on that below). That’s why Thunderbolt docks generally work with USB4 laptops, but testing is the only way to know which features this laptop actually supports. For a deeper look, see our guide to USB4 vs. Thunderbolt 4.

The Compatibility Questions We Want to Answer

A new CPU architecture, a new GPU vendor for Windows on Arm, and a new port design add up to a lot of open questions. Here’s what we’ll be looking at:

• Docking stations: How do Thunderbolt 4, USB4, and USB-C docks behave on this USB4 host? Do full-featured Thunderbolt docks work as expected, including their downstream ports and Ethernet?
• Multiple monitors: Can a single-cable dock drive more than one extended display? Does the RTX Spark display engine support DisplayPort MST, which many Windows docks use to run two or more monitors from one connection? How do the HDMI port and USB-C ports count toward the three-display limit?
• DisplayLink: DisplayLink drivers already run natively on Windows on Arm with Snapdragon systems. We’ll confirm how DisplayLink docks and adapters perform alongside an NVIDIA GPU on this new platform.
• Charging through a dock: The laptop supports 140W charging and needs 115W or more for fast charging. Most docks provide around 96 to 100W. Will a 100W dock keep the battery charged during heavy GPU or AI workloads?
• Networking and USB accessories: Do Ethernet adapters, storage, and other USB peripherals work with the drivers built into Windows on Arm, or do any need extra drivers?
• External GPUs: Can the Surface Laptop Ultra’s USB4 ports run an external GPU at all? More on that below.

We aren’t making any compatibility claims yet. We’ll share results once we’ve tested on real hardware.

Exploring External GPUs with the Plugable TBT5-AI

The Plugable TBT5-AI is a Thunderbolt 5 enclosure for a full desktop GPU. It’s designed for running AI models locally with tools like Ollama, LM Studio and Microsoft Foundry Local. Pairing it with the Surface Laptop Ultra raises some interesting questions:

• Do the laptop’s USB4 ports support the PCIe tunneling an eGPU needs?
• Will NVIDIA or AMD desktop GPU drivers load on Windows on Arm alongside the built-in RTX Spark GPU?
• If it all works, how do AI workloads split between the laptop’s unified memory and a desktop card’s dedicated VRAM?

While we’re not ready to share results yet, we expect the TBT5-AI to work well over the Surface Laptop Ultra’s USB4 ports. We’re excited to dig in, and we’ll publish what we find.

What’s Coming Next

Over the coming weeks, we’ll publish:

• Docking station compatibility results for the Surface Laptop Ultra across our Thunderbolt, USB4, USB-C and DisplayLink docks
• Multi-monitor testing: how many displays, at what resolutions and refresh rates, and with which docks
• TBT5-AI eGPU findings: what works, what doesn’t, and what it means for local AI on Windows on Arm
• Practical recommendations for building a Surface Laptop Ultra desk setup

Check back soon for results. If there’s a setup or accessory you’d like us to test with the Surface Laptop Ultra, let us know. Your questions help shape our testing.