Loading...
Loading...
NVIDIA unveiled the RTX Spark superchip, pairing a Blackwell-series GPU (up to 6,144 cores) with an ultra-efficient up-to-20-core CPU and as much as 128 GB of unified memory. With FP4 AI throughput near 1 petaflop, native CUDA support, and RTX features like RT cores, DLSS, Reflex, AV1 encode, and NVIDIA Broadcast, the SoC targets thin laptops and compact desktops for local model development, on-device agents, creative workflows, and gaming. OEMs including ASUS, Dell, HP, Lenovo, Microsoft, and MSI are lining up systems, signaling a push to mainstream high-performance, power-efficient, AI-capable personal computers; availability via sign-up.
RTX Spark brings near-petaflop AI throughput and native CUDA to thin, power-efficient PCs, enabling local model development and real-time AI features. Tech professionals must plan for new hardware capabilities, software tooling, and performance profiles in client devices.
Dossier last updated: 2026-06-01 07:45:23
Microsoft and NVIDIA unveiled Windows PCs accelerated by NVIDIA RTX Spark at NVIDIA GTC, positioning them as the most powerful and efficient thin-and-light Windows laptops yet. RTX Spark delivers up to 1 petaflop of AI performance, up to 6,144 Blackwell RTX cores, up to 20 Arm-based power-efficient CPU cores, and up to 128GB unified memory. Microsoft said it optimized Windows with workload profile scheduling and the Microsoft Power and Thermal Framework to maximize performance and efficiency on the heterogeneous architecture, plus DirectX 12 and Windows ML integrations to enable neural rendering, ray tracing, and native TensorRT support. The collaboration aims to fuel local agents, creators, AI developers and gamers with high performance and power efficiency.
Microsoft announced the Surface Laptop Ultra, a high-end 15-inch notebook co-engineered with NVIDIA and optimized for RTX Spark and Windows, aimed at creators, developers and AI builders. It pairs an NVIDIA Blackwell RTX GPU with up to 128 GB of unified memory, full CUDA support and claimed 1 petaflop of AI compute (NVIDIA-sourced theoretical FP4 TOPS with sparsity), enabling local execution of very large models. The laptop emphasizes thermal design, acoustics, repairability, a 2,000-nit mini‑LED PixelSense Ultra display, extensive ports (HDMI, USB-C/A, SD, headphone), large haptic trackpad and all-day battery in pre-release testing. Microsoft positions the device for intensive workloads like 3D rendering, long compiles and multi-model AI workflows; availability is later this year and specs may change.
Microsoft unveiled the Surface Laptop Ultra, a high-end 15-inch laptop engineered with NVIDIA to target creators, developers and AI builders needing desktop-class local compute. The device pairs an NVIDIA Blackwell RTX GPU with up to 128GB of unified memory, full CUDA support, and claims up to 1 petaflop of AI compute—enabling local execution of models up to ~120B parameters. Microsoft highlights a mini-LED 2,000-nit PixelSense Ultra display, large haptic touchpad, broad port selection (HDMI, USB-C, USB-A, SD, headphone), refined thermal and acoustic engineering, and all-day battery life. Positioned as a repairable, performance-first Surface, the Laptop Ultra will ship later this year as a pre-release product with specs subject to change.
NVIDIA announced the RTX Spark superchip, a combined AI and RTX graphics SoC targeting ultra-slim laptops and compact desktops. The chip pairs a Blackwell-based GPU with up to 6,144 cores, an up-to-20-core ultra-efficient CPU, FP4 AI throughput up to 1 petaflop, and up to 128 GB of unified memory, enabling local development and inference, creative workflows, and gaming with features like RT cores, DLSS, Reflex, AV1 encode, and NVIDIA Broadcast. CUDA runs natively, positioning RTX Spark as a platform for personal AI agents, content creation, and high-performance gaming on power-efficient systems from OEMs like ASUS, Dell, HP, Lenovo, Microsoft, and MSI. NVIDIA is collecting sign-ups for availability notifications.
NVIDIA announced the RTX Spark superchip, a combined Blackwell-series GPU and ultra-efficient CPU designed to bring high-end AI, creative workloads, and gaming to thin laptops and small desktops. The chip packs up to 6,144 Blackwell GPU cores, a 20-core CPU, FP4 AI performance up to 1 petaflop, and up to 128 GB unified memory, promising native CUDA support for local model development, accelerated creative apps, and real-time ray tracing with DLSS and Reflex. NVIDIA positions RTX Spark as a platform for on-device agents, local prototyping and inference, and continuous desktop AI services, with OEM partners including ASUS, Dell, HP, Lenovo, Microsoft, and MSI planning systems. Availability is via sign-up; the announcement signals a push to mainstream high-performance, power-efficient AI-capable PCs.