RTX Spark Surface Laptop Ultra Wins on Local AI but Falters on Price and Gaming, Reviews Say

The Surface Laptop Ultra starts at $2,599.99 for an 18-core model with 24GB of memory and 512GB of storage. A 20-core, 128GB, 1TB configuration tops out at $5,899.99, and lead times for that model already stretch to late December. The 64GB to 96GB of GPU-accessible memory that lets 35B-parameter models run well only comes with the 128GB build, while a 32GB or 48GB Surface will not fit the same models.

The reviews acknowledged the RTX Spark laptop's power but agreed it is not without weaknesses. Zac Bowden of Windows Central said, 'It's a beast, but not at everything,' while Tom's Guide called it 'not a MacBook Pro killer but something much weirder and wildly powerful.' The praise centered on local AI models: every outlet that tested them put the Surface ahead of Apple's M5 Max and AMD's Ryzen AI Max+ 395, the most powerful mobile chips for laptops. Gaming performance drew little praise, with some reviewers calling it merely okay and others describing it as a weak spot.

Cost is the main sticking point. Tom's Hardware warned that 'sky-high DRAM and NAND price hikes' will be a headwind for the platform in the near term, and it noted that the same money still buys 'a lot' of cloud tokens from OpenAI and Anthropic. Windows Central said comparable devices offer just enough and cost less.

The 20-core N1X CPU pairs 10 performance cores with 10 efficiency cores and targets heavy multi-threaded work. In single-core tests it edged the Dell Pro Precision 5 16S, scoring 558.5 against 534.8 in Cinebench 2026. But StorageReview said single-thread results sit 15 to 30 percent behind the x64 field, which it called 'just okay.' In multi-core work the Surface was 34 percent ahead of the Razer Blade 16, but it did not compete with Apple's latest Max chips, consistently trailing the M4 Max and M3 Ultra. Under a heavy all-core CPU workload using an OpenBLAS double-precision solve, the device lost power repeatedly, though most buyers will never run anything like it. In real-world work it won nearly every multi-threaded CPU test, compiling LLVM in 206 seconds against 327 for the next-fastest system. Its Arm64 SVT-AV1 encoding managed about a fifth of what the x64 systems produced, and memory latency of 251 ns was the worst in the group. Day-to-day use feels fine, and even emulated apps feel native, but for single-threaded, browser-bound work the x64 field is likely the better buy.

The 6,144-core Blackwell GPU is the real engine of the chip, with fifth-generation Tensor Cores built for AI workloads and up to 96GB of memory to draw on. In local AI tests it was the fastest at agentic tasks, finishing about 25 percent quicker than the M5 Max MacBook Pro. In Stable Diffusion 1.5 INT8 it scored 11,294 and substantially outperformed the RTX 5080, 5070 Ti and 5060. For models that fit in 24GB, however, the Razer Blade 16's RTX 5090 generates tokens 2.3 times faster, and video rendering trails even the RTX 5060.

Gaming is a secondary strength. Performance is similar to an RTX 4070 or 5060, and Cyberpunk 2077 held 60fps at 1440p with frame generation. Shared memory bandwidth and input latency are limited: shared memory bandwidth hit 300GB/s, below the 384GB/s on a mobile RTX 5060. Gears of War: E-Day averaged 77.7 fps with 1% lows of 39.9 fps and 83.5 ms of latency. Control Resonant reached 109.2 fps, but only with 3X frame generation. TechRadar's roundup said that if AI is not the user's priority and low latency and top performance are, a separate gaming laptop with a dedicated RTX 50-series GPU would still provide a better option.

Editor's Summary

Microsoft's Surface Laptop Ultra, powered by Nvidia's RTX Spark, drew praise for local AI performance but criticism over high pricing, uneven gaming results and single-core gaps. Early reviews position it as a specialized AI laptop rather than a general-purpose MacBook Pro competitor.