NVIDIA DGX
Inspired by the demands of deep learning and analytics, NVIDIA DGX™ Systems are the essential instruments for AI research built on the new NVIDIA Ampere Architecture.
NVIDIA DGX Spark «NEW»
Powered by the NVIDIA GB10 Grace Blackwell Superchip, NVIDIA DGX™ Spark delivers 1000 AI TOPS of AI performance in a power-efficient, compact form factor.
- Up to 1000 AI TOPS of AI performance at FP4 precision with the NVIDIA Grace Blackwell architecture.
- 20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm
- 128 GB LPDDR5x, unified system memory
- ConnectX-7 Smart NIC
NVIDIA DGX Station «NEW»
The NVIDIA DGX Station is a high-performance desktop AI development platform powered by the Grace Blackwell Superchip and 784GB of memory, purpose-built to deliver exceptional compute power for large-scale AI training and inference.
- Built with the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
- 1x Grace-72 Core Neoverse V2
- Features a massive 784GB of coherent memory space
- The NVIDIA ConnectX-8 SuperNIC delivers up to 800 Gb/s to accelerate hyperscale AI workloads with ultra-fast, efficient network connectivity
NVIDIA DGX H200 «NEW»
Part of the NVIDIA DGX platform, DGX H200 is the AI powerhouse that’s the foundation of NVIDIA DGX SuperPOD™ and NVIDIA DGX BasePOD™, accelerated by the groundbreaking performance of the NVIDIA H200 Tensor Core GPU.
- 18x NVIDIA NVLink® connections per GPU, 900GB/s of bidirectional GPU-to-GPU bandwidth
- 7.2TB/s of bidirectional GPU-to-GPU bandwidth, 1.5X more than previous generation
- 10x NVIDIA ConnectX®-7 400Gb/s Network Interface
- Dual Intel Xeon Platinum 8480C processors, 112 cores total, and 2TB System Memory
NVIDIA DGX GB200 «NEW»
NVIDIA DGX™ GB200 lets enterprises achieve unprecedented performance and predictable uptime, dramatically boosting utilisation and productivity and increasing the ROI of their AI initiatives. It creates a new standard for AI performance, reliability, and scalability
- 36 NVIDIA Grace CPUs and 72 Blackwell GPUs—–connected as one with NVIDIA NVLink™
- 72x OSFP single-port NVIDIA ConnectX®-7 VPI with 400Gb/s NVIDIA InfiniBand
- 36x dual-port NVIDIA BlueField®-3 VPI with 200Gb/s InfiniBand and Ethernet
- Delivers 1.4 exaFLOPS of AI performance, 30 terabytes (TB) of fast memory, and 130 terabytes per second (TB/s) of bidirectional GPU bandwidth
NVIDIA DGX H100 «NEW»
NVIDIA DGX H100 delivers the performance needed for enterprises to solve the biggest challenges with AI. For organisations looking to scale this power across teams and workloads, NVIDIA DGX SuperPOD offers a turnkey infrastructure solution that brings together multiple DGX systems into a unified, high-performance AI factory.
- 8x NVIDIA H100 GPUs with 640 gigabytes of total GPU memory
- Features 6X more performance, 2X faster networking with NVIDIA ConnectX®-7 SmartNICs
- NVIDIA BlueField®-3 data processing units (DPUs), and high-speed scalability
- Dual x86 CPUs and 2 terabytes of system memory
NVIDIA DGX B200 «NEW»
With NVIDIA DGX B200, enterprises can equip their data scientists and developers with a universal AI supercomputer to accelerate their time to insight and fully realise the benefits of AI for their businesses. When paired with NVIDIA DGX SuperPOD, teams can scale effortlessly from experimentation to massive production-scale AI training and inference.
- Built with eight NVIDIA Blackwell GPUs interconnected with fifth-generation NVIDIA® NVLink™
- 1.4TB of GPU memory space
- 72 petaFLOPS of training performance
- 144 petaFLOPS of inference performance