Tom's Hardware Tests Nvidia RTX Spark in Surface Laptop Ultra, Finds Local AI Strengths
Early testing of the new Windows on Arm platform shows strong local AI performance but weaker CPU benchmarks compared to Apple silicon.
Key points
- The RTX Spark Superchip features 20 CPU cores and 6144 CUDA cores in Microsoft's Surface Laptop Ultra.
- The headline notes the platform impresses in local AI, though gaming is not its strong suit.
- CPU benchmark results from the MediaTek complex trailed significantly behind Apple's M5 Max chip.
Nvidia CEO Jensen Huang announced the RTX Spark platform at Computex 2026 with the goal of reinventing the personal computer for the agentic AI era. Tom's Hardware has completed initial testing of this new hardware foundation using a Microsoft Surface Laptop Ultra equipped with the top-end RTX Spark Superchip configuration.
The test system included 20 CPU cores, 6144 CUDA cores, and 128GB of LPDDR5X-9400 RAM. Tom's Hardware spent a week evaluating the device to determine its capabilities in the context of Nvidia's vision for an active, autonomous computing companion. The evaluation focused on broad platform experience and competitive performance rather than specific system attributes like battery life or build quality.
Context and Background
Nvidia partnered with Microsoft, Dell, HP, MSI, Lenovo, and Asus to develop this new Windows on Arm platform. The goal is to enable secure and contained execution of local AI agents directly on the device. Tom's Hardware traveled to San Francisco last week to gather details on the RTX Spark platform before returning with the Microsoft device for hands-on analysis in their labs.
The testing methodology compared the Surface Laptop Ultra against two other high-end systems. The primary competitor was a 14-inch MacBook Pro featuring Apple's M5 Max SoC, which includes an 18-core CPU, a 40-core GPU, and 128GB of LPDDR5X RAM. A secondary comparison used AMD's compact Ryzen AI Halo system with the latest ROCm stack and llama.cpp builds, as a mobile Ryzen AI Max+ 395 system was not available in the labs.
Who Is Affected
The RTX Spark platform is intended for work and creation, according to Nvidia's announcement. The Tom's Hardware headline states that the platform impresses in local AI, but gaming is not its strong suit. The body text confirms that the hardware foundation is in place for this vision, though specific performance nuances vary by workload.
Performance testing revealed that the MediaTek Arm CPU complex may be the platform's weakest point. While the system did not feel slow in regular use, Geekbench results showed that both single-core and multi-core CPU performance greatly trailed Apple's M5 Max. This indicates that while the GPU and AI features are advanced, general CPU-heavy tasks may not match competitors.
Tom's Hardware stated that embargoes prevented the release of system-specific results such as battery life, display quality, or build details in this initial report. The outlet
What to do and how to stay safe
- Sudden role changes in output signal a successful override of system instructions.
- Hidden metadata in images or documents often carries invisible injection payloads.
- Immediate isolation of the affected session prevents lateral movement within the AI workflow.
Prompt injections exploit the model’s ability to follow instructions by hiding malicious commands in input data. Isolate affected sessions immediately and implement external guardrails to filter inputs before they reach the model.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
