DigestAI news desk

Cut through the AI noise.

Listen
Hardware & Compute12 min read

Nvidia’s RTX Spark Superchip outperforms Apple and AMD in local AI benchmarks

Nvidia’s RTX Spark platform, announced at Computex 2026, pairs the GB10 Superchip with Windows on Arm to enable local AI agents. Tom’s Hardware tested the Surface Laptop Ultra with the RTX Spark Superchip, featuring 20 CPU cores, 6144 CUDA cores, and 128GB RAM, against a MacBook Pro with Apple’s M5 Max and AMD’s Ryzen AI Halo. The Surface Laptop Ultra finished agentic AI tasks 25% faster than…

4 sources

Key points

  • RTX Spark Surface Laptop Ultra beats MacBook Pro (M5 Max) and AMD Ryzen AI Halo in local AI benchmarks by 25-50%
  • Gaming performance is weak due to CPU bottlenecks and limited memory bandwidth, even with DLSS Frame Generation
  • Surface Laptop Ultra priced at **$5,899** with long lead times, raising cost concerns vs. cloud AI alternatives

While the RTX Spark excels in local AI performance, its gaming capabilities remain limited. The platform struggles with CPU-heavy games like Gears of War: E-Day due to the overhead of Prism x86 emulation and the Grace CPU’s performance. Even in less CPU-dependent titles like Control Resonant, memory bandwidth constraints and input latency issues persist. Nvidia’s vision for the platform—secure, autonomous AI agents—remains unfulfilled due to unready Windows features. The Surface Laptop Ultra costs $5,899 with 1TB storage, and lead times extend to December 2026, raising questions about its value compared to cloud-based AI alternatives.

The story so far

4 episodes →
  1. Nvidia’s RTX Spark Superchip outperforms Apple and AMD in local AI benchmarksthis story
Full story from Tom's Hardware · by Jeffrey KampmanOpen source ↗

We test the RTX Spark Superchip in Microsoft's Surface Laptop Ultra — Nvidia's Windows on Arm platform impresses in local AI, but gaming isn't its strong suit

Tom's Hardware · 15 October 2026

Loading the full article…

This text was published by Tom's Hardware and written by Jeffrey Kampman. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Coverage and discussion

4sources
Topics · follow one to build your own front page
NvidiaMicrosoftAppleAMDDellHPRTX Spark SuperchipGB10 SuperchipQwen 3.8-27BQwen 3.6 35B A3BM5 MaxRyzen AI Max+ 395Jensen HuangJeff Kampman

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Hardware & Compute

All →

Related stories

hot ASML warns 2027–2028 chip tool output may fall short of AI demand new Google rolls out Universal Cart for US users in Gemini and Search hot Character.AI files for Chapter 11 bankruptcy, Disney offers $20 million for assets new Apertus 1.5 adds 4T tokens to 8B model, 2T to 70B, with multimodal and long-context upgrades hot OpenAI's GPT-6 Astra: Changing Work and Research new Nvidia pledges $250M in compute for UK AI agent safety research hot OpenAI’s AI agents breach Hugging Face in July test new Nvidia’s RTX Spark Superchip outperforms Apple and AMD in local AI benchmarks