Procyon® Is Ready to Benchmark NVIDIA RTX Spark

A new class of AI PC hardware has arrived, and Procyon® is ready to measure it.

NVIDIA RTX Spark pairs a Blackwell RTX GPU with a Grace CPU, connected via NVLink-C2C, delivering up to 128 GB of high-bandwidth unified memory. This architecture will arrive in retail hardware starting October 2026, with optimized, premium Windows on Arm devices launching from leading global manufacturers, including ASUS, Dell, HP, Lenovo, Microsoft Surface, and MSI.

What Procyon now tests on RTX Spark devices

Procyon® AI benchmarks are enabled on RTX Spark hardware, covering:

• Procyon® AI Computer Vision 2.0 – Real-world tasks like object detection and image grouping
• Procyon® AI Image Generation – Using diffusion models
• Procyon® AI Text Generation – across multiple large language models (LLMs)

This means IT teams, reviewers and technical evaluators can benchmark how RTX Spark-powered devices handle real AI inference workloads, on the same hardware NVIDIA and its partners are bringing to market now.

The Procyon® AI Text Generation Benchmark also introduces a new default execution path for both x64 and arm64 NVIDIA devices: llama.cpp with a Compute Unified Device Architecture (CUDA) backend.

AI text generation support for macOS

Evaluation capabilities now extend to Apple ecosystems with full support for Procyon® AI Text Generation Benchmark on macOS. The benchmark delivers model parity with the Windows version, testing how systems handle large language models (LLMs). Built for Apple Silicon, this execution path fully leverages Mac graphics processing units (GPUs) and supports readiness for the latest macOS environments, giving reviewers and IT teams a consistent standard for cross-platform comparison.

Still a PC, not just an AI accelerator

RTX Spark is built for demanding local AI work, but it still needs to function as a working PC. The Procyon® Office Productivity Benchmark measures real-world tasks in Microsoft Word, Excel, PowerPoint and Outlook, and has been fully validated to work as expected on RTX Spark devices alongside the newly released Procyon® Essentials suite and its multitasking workloads.

This reflects Procyon's broader philosophy of testing with real applications whenever possible to deliver insights that mirror how systems perform in everyday use, not just under synthetic AI workloads.

Day 1 readiness for the next hardware wave

New AI PC hardware categories tend to arrive faster than the tools to properly evaluate them. Having Procyon® AI benchmarks suite ready for RTX Spark from the earliest devices means: .

• Reviewers can publish accurate, real-world performance data from day 1.
• Original equipment manufacturers (OEMs) can validate their configurations against a consistent, vendor-neutral standard.
• IT teams evaluating this hardware for technical or creative staff don't have to wait for benchmarking tools to catch up.

The Procyon® benchmark suite is part of ULTRUS® software from UL Solutions. Explore the full Procyon® suite, including the AI Computer Vision, AI Image Generation and AI Text Generation benchmarks, at benchmarks.ul.com/procyon.