Features:

GB10 Grace Blackwell Superchip (1petaFLOP AI performance Blackwell GPU, 20-core Arm CPU)

• Accelerates AI, Data Science, compute, rendering, and visualization

Workloads 128GB coherent unified system memory

• Work with large AI models of up to 200 billion parameters, fine-tune models of up to 70 billion parameters
•  ~100GB of system memory available for user workloads

ConnectX-7 networking

• Connect two EdgeXpert systems together to work with models of up to 405 billion parameters
• A maximum of 2 DGX Spark systems can be clustered together

DGX OS & NVIDIA AI software stack

• Seamlessly move workloads from EdgeXpert to DGX Cloud or any accelerated data center or cloud infrastructure

Flexible deployment configurations

• Configure as an AI Workstation or network connected personal AI cloud

Compact, power efficient design

• Easily fits on any desk, powered by standard wall outlet

Specifications:

• Architecture: NVIDIA Grace Blackwell
• GPU: NVIDIA GB10
• CPU: Arm 20 core, 10 Cortex-X925 + 10 Cortex-A725
• CUDA Cores: Blackwell Generation
• Tensor Cores: 5th Generation
• Tensor Core Supported Data Formats: TF32, FP16, BF16, INT8, FP8, FP6, FP4
• RT Cores: 4th Generation
• Tensor Performance: 1 petaFLOP AI performance (FP4, Sparse)
• System Memory: 128 GB LPDDR5x, unified system memory
• Memory Interface: 256-bit
• Memory Bandwidth: 273 GB/s
• Storage: 4 TB NVME.M2 with self-encryption
• USB: 4x USB 3.2 Type C (up to 20Gb/s)
• Ethernet: 1x RJ-45 connector 10 GbE
• NIC: ConnectX-7 Smart NIC
• Wi-Fi: WiFi 7
• Bluetooth: BT 5.4
• Audio-output: HDMI multichannel audio output
• System Weight: 1.2 kg
• Display Connectors: 1 x HDMI 2.1, 4x DP1.4a via USB-C
• NVENC | NVDEC: 1x | 1x
• OS: NVIDIA DGX OS
• Dimensions (L*W*H): 151 L x 151 x 52 mm (1.2L)

Warranty: 3 Years

Note: Features, Price, Specifications are subject to change without notice. Images used for representation purposes Only