Nvidia unveils RTX Spark Superchip for PCs
- Nvidia unveiled its RTX Spark PC chip and RTX Spark Superchip at Computex in Taipei on June 1, pitching local AI capabilities for laptops and desktops. - Nvidia said the platform pairs an Arm CPU with a Blackwell GPU and up to 128GB of unified memory, targeting AI models on-device. - Nvidia said Microsoft, Dell, HP, Lenovo and Asus systems using RTX Spark are due later in 2026.
Nvidia used its Computex keynote in Taipei on June 1 to introduce RTX Spark, a new PC chip it said is designed to bring AI workloads directly onto laptops and desktop computers. Chief Executive Jensen Huang said the chip was part of a joint effort with Microsoft to “reinvent the PC” for the AI era after three years of work. Reuters reported the launch as Nvidia’s push into a market long dominated by Intel, AMD, Apple and, more recently, Qualcomm’s Arm-based Windows systems. ### What exactly did Nvidia announce in Taipei? Nvidia announced both the RTX Spark GPU and the broader RTX Spark Superchip platform at Computex, framing the launch around AI PCs rather than conventional gaming notebooks alone. Reuters said the company presented the chip as a way to move AI capabilities directly onto personal computers instead of relying only on cloud data centers. (finance.yahoo.com) Tom’s Hardware reported that the RTX Spark Superchip combines an Arm CPU, a Blackwell GPU and up to 128GB of unified memory. Nvidia’s own Computex roundup said the platform is aimed at slim Windows laptops and compact desktop PCs and brings together CUDA, RTX, DLSS, TensorRT and other Nvidia software tools on one PC-focused design. (finance.yahoo.com) ### How is this different from the AI PCs already on the market? The 128GB unified-memory figure is one of the clearest ways Nvidia is trying to separate Spark from other Windows-on-Arm systems. Tom’s Hardware and PCMag said the shared memory pool is meant to let the CPU and GPU work from the same large block of RAM, which Nvidia says can support local AI models up to 120 billion parameters. (msn.com) Reuters said Huang described the product as part of Nvidia’s effort with Microsoft to build PCs around AI agents. Nvidia’s GeForce post used similar language, saying RTX Spark is intended for “personal AI agents” and local AI applications across the Windows ecosystem. (msn.com) ### Which companies are expected to ship these machines? Microsoft, Dell, HP, Lenovo and Asus were among the hardware partners identified in coverage of the launch. KitGuru said those companies would ship RTX Spark laptops later this year, while Reuters and other reports described the chip as targeting both laptops and compact desktops. (finance.yahoo.com) CNBC reported that Nvidia’s entry into PCs would put its Arm-based chip into new laptops from Dell, Microsoft, HP, Asus and others. That places Nvidia more directly into the Windows client market rather than limiting its PC presence to discrete graphics cards. ### Was this only about AI, or did Nvidia also make gaming announcements? (kitguru.net) DLSS 4.5 Ray Reconstruction was announced alongside RTX Spark at Computex. KitGuru reported that the feature is due in August, and Nvidia said it will improve image quality in path-traced and ray-traced games and later come to Blender 5.3 in the fall. (cnbc.com) Nvidia also said more than 1,000 RTX-enhanced and accelerated games and apps are now available. The company used the keynote to package Spark, DLSS 4.5 and local AI software updates as part of a broader Computex product push. ### What comes next after the Computex launch? Later in 2026, partner systems from Microsoft, Dell, HP, Lenovo and Asus are expected to begin shipping with RTX Spark, according to KitGuru and other reports. (kitguru.net) Tom’s Hardware also reported that Nvidia outlined a three-generation roadmap for RTX Spark systems, naming Rubin with LPDDR6 memory and a later generation called Rosa Feynman. (nvidia.com) May 31 and June 1 announcements from Nvidia also set two near-term milestones: DLSS 4.5 Ray Reconstruction is scheduled for August, and Blender 5.3 support is slated for the fall, according to Nvidia’s Computex post. (nvidia.com) (kitguru.net)