Nvidia DGX Spark GB10 128 GB LPDDR5x-SDRAM 4 TB SSD DGX OS Workstation Gold

1. Grace Blackwell AI Superchip Delivering Up to 1 PFLOP of AI Compute

At the heart of the DGX Spark is NVIDIA’s GB10 Grace Blackwell Superchip, combining a Blackwell GPU with a 20-core Arm CPU (10 Cortex-X925 + 10 Cortex-A725 cores). The system can deliver up to 1 petaFLOP of FP4 AI performance, making it capable of running advanced AI workloads directly on a desktop without requiring cloud resources. This enables developers to prototype, test, and deploy AI models locally with access to the full NVIDIA AI software ecosystem.

2. Massive 128GB Unified Memory for Large AI Models

Unlike conventional PCs that separate system RAM and GPU VRAM, the DGX Spark features 128GB of coherent LPDDR5X unified memory with 273 GB/s memory bandwidth. This allows AI developers to work with models far larger than those that would fit into a consumer GPU’s VRAM. NVIDIA states that a single DGX Spark can run AI models of up to 200 billion parameters, while connecting two systems together can support models of up to 405 billion parameters.

3. Compact AI Workstation with Enterprise Connectivity

Despite its small 150 × 150 × 50.5 mm footprint and 1.2 kg weight, the DGX Spark includes enterprise-grade connectivity. The 4TB Founders Edition provides 4TB self-encrypting NVMe storage, 10Gb Ethernet, Wi-Fi 7, Bluetooth 5.4, ConnectX-7 networking at 200Gbps, four USB-C ports, and HDMI 2.1a output. This makes it suitable for AI labs, edge deployments, research environments, and developers who need powerful local AI hardware without the size or power requirements of a rack-mounted server.