Microsoft's Surface Laptop Ultra Arrives as a Flagship for NVIDIA's RTX Spark Platform
Microsoft's Surface Laptop Ultra is one of the flagship devices for NVIDIA's RTX Spark platform, and Engadget's review describes it as the most powerful Surface device yet.
The platform's processor is an Arm-based system-on-a-chip (SoC, a single chip that combines the CPU, GPU and other blocks) that can be configured with up to 20 CPU cores and 6,144 GPU cores. Engadget reports that the base Surface Laptop Ultra configuration carries 18 CPU cores and 5,120 GPU cores. The most powerful RTX Spark configuration of the machine costs $3,700 and ships with 32GB of RAM. The CPU and GPU share a unified memory pool, meaning both draw from one block of system memory rather than from separate pools, and the Surface Laptop Ultra can be equipped with up to 128GB of it.
Pricing in the same review starts at $2,599 with 24GB of RAM in the stock configuration. Engadget also refers to the base configuration at $2,600. The machine has a 15-inch display with a 3:2 aspect ratio, the proportion of width to height, which gives more vertical workspace than the 16:9 panels common on Windows laptops of this class.
On graphics, Engadget reports that the RTX Spark chip delivers gaming performance similar to an RTX 5070 laptop GPU, a discrete mobile graphics card. That comparison is the reviewer's, and it places the part against hardware that many developers and creators already own.
Microsoft's own Surface landing page frames the lineup in its marketing language. The Surface Laptop Ultra is described as "The performance laptop made for the world makers," and Surface as a whole is called "The ultimate Windows device." The page also says that on Copilot+ PCs, improved Windows search can find files by typing a description. Surface Pro is pitched as "Power that adapts to how you work," and Surface Laptop as "Sustained power, beautifully refined." Microsoft's Surface page sets out those positioning statements.
In my view, the number that matters most to developers and power users is not the core count but the memory ceiling. Because the CPU and GPU address one unified pool, the 128GB configuration is not split into a system partition and a separate VRAM budget. Large model weights, big datasets and heavy multitasking therefore draw from the same allocation. Whether that architecture holds up as cleanly as a discrete-GPU design under sustained load is a question for detailed benchmarking.
The broader context here is that Microsoft is building its premium Windows hardware around a partner's Arm silicon, with Surface acting as the reference design. The price spread from $2,599 to $3,700 suggests the range is meant to cover both mainstream productivity buyers and those who want the largest memory and GPU allocations. For the people who build software on these machines, the useful questions are how the Arm toolchain, driver stack and x86 compatibility layer behave across real workloads, and whether the unified memory model makes local inference and simulation work more practical than on discrete-GPU laptops.
