






















For the first few decades of the company’s existence, everyone knew NVIDIA as the company that made graphics cards — a gaming juggernaut. Then, with the explosive growth of AI over the past few years, it became the world leader in AI. What many people aren’t aware of is that NVIDIA has also had a firm grasp on the creator market for years, with its GPUs consistently holding more than 90% market share.
The NVIDIA RTX Spark is a culmination of NVIDIA’s three core markets coming together in a single product that serves AI developers, content creators, and gamers with a single chip — something the company has never done before. Previously, NVIDIA has always depended on AMD or Intel for CPUs; with the RTX Spark, it still has MediaTek as a partner, but this time, both the CPU and GPU are on a single chip designed by NVIDIA.
The RTX Spark is very similar to the GB10 chip in the DGX Spark desktop platform that NVIDIA launched last year. I reviewed it at the time and found it extremely useful for running AI applications and for easy AI deployment. The Blackwell GPU in the RTX Spark has up to 6,144 CUDA cores and a 20-core Grace CPU built by MediaTek, with 10 Arm X925 performance cores and 10 Arm A725 cores. The RTX Spark will be configurable with up to 128GB of RAM, though I expect we’ll see configurations from 32GB to 128GB to control costs. The unified memory between the CPU and GPU should enable the platform to excel at gaming, creator workloads, and AI — likely reinforcing NVIDIA’s grip on its core audiences.
Beyond the hardware, NVIDIA is also partnering with Microsoft to advance RTX Spark with a native Windows experience for agents that includes enhanced security features, including NVIDIA OpenShell, which is designed to run agents more securely. Thanks to NVIDIA’s earlier work with DGX Spark, the RTX Spark also already has a lot of tools and software optimized for its performance and memory footprint. This means that NVIDIA has already established that the chip can handle compute-hungry tasks like very large 3-D scenes, 12K video, 120B-parameter models, and games played at 1440P at more than 100 frames per second.
For the DGX Spark, Adobe rearchitected its most popular applications, Premiere Pro and Photoshop, to deliver 2x faster AI and graphical performance. Adobe later announced that its Substance 3D Painter and Stager apps would get the same treatment for RTX Spark. At Computex 2026 in Taiwan, Adobe ran live demos of Premiere Pro powered by Microsoft’s new Surface Laptop Ultra with an RTX Spark chip inside.
At Computex, NVIDIA also gave a bunch of demos highlighting the performance of the RTX Spark. NVIDIA and its partners showcased a mix of creator, agentic coding, and gaming applications from some of the world’s biggest software players. For instance, NVIDIA demoed Epic Games’ Unreal Editor with an entire city loaded to memory — a 93GB file — running smoothly on the laptop. Similarly, Microsoft showed off SolidWorks running extremely smoothly even on battery, and that app is known to be quite performance-hungry. NVIDIA even unplugged the laptop to show that it does not throttle significantly on battery power. NVIDIA also partnered with Blender to show how the popular rendering tool benefits by being accelerated by RTX Spark. Microsoft showed off VS Code and GitHub Copilot working together and running smoothly on the Surface Laptop Ultra. NVIDIA also dedicated an entire room to gaming, showing off popular titles like Fortnite running smoothly, as well as newer titles like Pragmata, which looked very good. NVIDIA also gave DGX Spark demos that it says will translate to RTX Spark demos in the future, thanks to effectively having the same architecture. This raises the question of how soon the DGX Spark will run Windows, like the DGX Workstation and RTX Spark.
NVIDIA has also done a lot of work to squash any concerns that games might not be compatible with the RTX Spark, especially given how mature its graphics drivers already are on Windows. The real lift comes from ensuring that titles from developers such as Krafton (PUBG) and Riot (League of Legends and Valorant) use anti-cheat and game engines that are compatible with Arm, something NVIDIA announced at Computex. Qualcomm has already done some of the legwork for NVIDIA here, partnering with Epic Games to get Fortnite ready with Epic Games’ Easy Anti-Cheat software. In many ways, Qualcomm helped NVIDIA get to where we are today with anti-cheat and DRM providers, many of whom have worked with Qualcomm to ensure Arm compatibility well before NVIDIA debuted RTX Spark.
NVIDIA is following a slightly different go-to-market strategy from the one it used with DGX Spark. This time, it won’t have its own NVIDIA-branded product. It looks like the product will end up being Microsoft’s Surface Laptop Ultra, with both companies working closely to co-engineer it — but in the end it’s still very much Microsoft’s product. Microsoft also says that all of the Windows quality improvements it has been working on this year will be baked into the Surface Laptop Ultra at launch — and into any RTX Spark laptop running Windows. NVIDIA has also said that its usual laptop partners will join the fall launch of RTX Spark, including ASUS, Dell, HP, Lenovo, MSI, and Microsoft, with Acer and Gigabyte to follow. Microsoft also announced a desktop version of RTX Spark for its developer community, called the Surface RTX Spark Dev Box, at its Build conference held the same week as Computex.
There’s still no official word on pricing, but I do believe prices will vary wildly across the N1 and N1X chip variants of the RTX Spark, not to mention memory configurations. NVIDIA has not yet discussed the different configurations , so we will have to wait for those details to get a better idea of price and performance, which ultimately determine value. I expect most of NVIDIA’s offerings to be fairly premium, but it will ultimately depend on NVIDIA’s partners to show us what they can do and how they will design around RTX Spark. Nevertheless, I cannot wait to get my hands on a system for daily use and see how it stacks up against the latest from AMD, Apple, Intel, and Qualcomm.
此内容由惯性聚合(RSS阅读器)自动聚合整理,仅供阅读参考。 原文来自 — 版权归原作者所有。