Menu
Your Cart
NVIDIA DGX Spark AI Supercomputer – GB10 Grace Blackwell, 128GB LPDDR5X, 4TB NVMe
NVIDIA DGX Spark AI Supercomputer – GB10 Grace Blackwell, 128GB LPDDR5X, 4TB NVMe
NVIDIA DGX Spark AI Supercomputer – GB10 Grace Blackwell, 128GB LPDDR5X, 4TB NVMe
NVIDIA DGX Spark
Label-3 %
NVIDIA DGX Spark AI Supercomputer – GB10 Grace Blackwell, 128GB LPDDR5X, 4TB NVMe
NVIDIA DGX Spark AI Supercomputer – GB10 Grace Blackwell, 128GB LPDDR5X, 4TB NVMe
NVIDIA DGX Spark AI Supercomputer – GB10 Grace Blackwell, 128GB LPDDR5X, 4TB NVMe

NVIDIA DGX Spark AI Supercomputer – GB10 Grace Blackwell, 128GB LPDDR5X, 4TB NVMe

31,000.00
29,999.00
VAT INCLUSIVE

Customers also bought

NVIDIA • DESKTOP AI SUPERCOMPUTER

NVIDIA DGX SPARK AI power on your desk.

NVIDIA DGX Spark belongs to a new class of computers designed to build and run AI, enabling developers to prototype, fine-tune, and inference large AI models locally, then move workflows to the data center or cloud.

NVIDIA DGX Spark AI Supercomputer
1 PFLOPTheoretical FP4 Tensor performance using sparsity
128GBLPDDR5x Unified System Memory
273 GB/sMemory Bandwidth
4TBNVMe M.2 Self-Encrypting Storage
NVIDIA GB10 Grace Blackwell
NVIDIA GB10 SUPERCHIP

Grace Blackwell in a desktop system

1 PFLOP

Experience up to 1 petaFLOP of AI performance at FP4 precision with NVIDIA Grace Blackwell architecture, a Blackwell architecture GPU, and a 20-core Arm CPU.

GPU — Blackwell Architecture
CPU — 20 Core Arm: 10 Cortex-X925 + 10 Cortex-A725
Tensor Cores — 5th Generation
RT Cores — 4th Generation
DGX Spark unified memory
COHERENT UNIFIED MEMORY

Unified memory for large AI workloads

128GB

128GB of LPDDR5x unified system memory with a 256-bit memory interface and 273 GB/s of memory bandwidth. The official specifications state that DGX Spark can run AI development and testing workloads with models up to 200 billion parameters.

128GB LPDDR5x Unified System Memory
256-bit Memory Interface
273 GB/s Memory Bandwidth
AI Models up to 200B Parameters
DGX Spark system oneDGX Spark system two
NVIDIA CONNECT-X NETWORKING

Connect two systems. Work with larger models.

405B

High-performance NVIDIA Connect-X networking enables two NVIDIA DGX Spark systems to be connected together to work with AI models up to 405 billion parameters.

ConnectX-7 Smart NIC
10 GbE — 1× RJ-45
High-performance networking for dual-system workflows
Compact NVIDIA DGX Spark
COMPACT DESKTOP FORM

Advanced AI in a compact desktop form

1.2 KG

A compact 150 × 150 × 50.5 mm desktop system weighing 1.2 kg, with 4TB NVMe M.2 self-encrypting storage and NVIDIA DGX OS.

150 × 150 × 50.5 mm
System Weight — 1.2 kg
4TB NVMe M.2 with Self-Encryption
DGX OS
FULL OFFICIAL SPECIFICATIONS

Technical specifications

All values below are taken from the official specifications for product NVDGXSPARK-PB.

PART NUMBERNVDGXSPARK-PB
ARCHITECTURENVIDIA Grace Blackwell
GPUBlackwell Architecture
CPU20 Core Arm — 10 Cortex-X925 + 10 Cortex-A725
CUDA CORESBlackwell Generation
TENSOR CORES5th Generation
RT CORES4th Generation
TENSOR PERFORMANCE1 PFLOP¹
SYSTEM MEMORY128GB LPDDR5x Unified System Memory
MEMORY INTERFACE256-bit
MEMORY BANDWIDTH273 GB/s
OSDGX OS
STORAGE4 TB NVMe M.2 with Self-Encryption
USB4× USB Type-C
ETHERNET10 GbE — 1× RJ-45
NICConnectX-7 Smart NIC
WI-FIWi-Fi 7
BLUETOOTHBT 5.3
AUDIOHDMI Multichannel Audio Output
DISPLAY1× HDMI 2.1a
NVENC | NVDEC1× | 1×
DIMENSIONS150 × 150 × 50.5 mm
WEIGHT1.2 kg
¹ Theoretical FP4 performance using the sparsity feature. Specifications are preliminary and subject to change.
No reviews yetBe the first to share your opinion about this product.
تواصل معنا