Train Trillion-Parameter Generative AI Models
Experience the capability of generative AI through cutting-edge AI supercomputers. The NVIDIA DGX™ series combines elite superchips into a comprehensive GPU network, designed for handling large terabyte-scale models. It excels in areas like recommender systems, generative AI, and graph analytics, offering significant shared memory capacity for scalable performance across expansive AI models.
| Component | Specification |
|---|---|
| CPU + GPU | 72x NVIDIA Blackwell GPUs in Grace Blackwell Superchips |
| Total Fast Memory | 30.2TB |
| Interconnect | 72x OSFP single-port NVIDIA ConnectX®-7 VPI with 400Gb/s InfiniBand |
| NVIDIA NVLink Switch System | 9x L1 NVIDIA NVLink Switches |
| Component | Specification |
|---|---|
| CPU | Dual Intel® Xeon® Platinum 8480C processors, 112 Cores total |
| GPU | 40+ NVIDIA DGX H100 system with eight 80 GB H100 GPUs |
| Memory | 640 GB of aggregated HBM3 memory |
| Networking | Single-node bandwidth of at least 40 GBps to each DGX H100 system per GPU @ 900 GBps with fourth generation of NVIDIA NVLink |
NVIDIA GB200 NVL32
The next generation of DGX SuperPOD is purpose-built to deliver extreme
performance and consistent uptime for superscale generative AI training and
inference workloads. Built on NVIDIA’s own internal cluster designs, the full-stack
resilience capabilities—available for the first time in enterprise AI infrastructure—
allow enterprises to focus on innovation rather than operational complexity.
| AI Performance | GPU-to-GPU Bandwidth | Shared GPU Memory |
|---|---|---|
| 1,440 | 1.8 | 13.3 |
| PetaFlops of FP4 | Terabytes per second (TB/s) | Terabytes (TB) |
Accelerating Automotive
NVIDIA DGX systems are pivotal for automotive AI development, offering powerful AI computing for training deep neural networks essential in autonomous vehicle technology. Providing scalable, high-performance solutions for processing extensive datasets, optimizing data loads for safety and efficiency. Ideal for developers, NVIDIA DGX systems streamline the transition from data center innovation to practical application in self-driving car technology.
Powering Smart Cities
NVIDIA DGX systems are transforming edge computing and smart city development by enabling real-time, efficient processing of data from a myriad of sensors. The platform offers unparalleled AI computing power at the edge. Its compatibility with hybrid cloud and multicloud IoT environments ensures seamless application deployment and management. Optimized for enterprise-grade security and performance, NVIDIA DGX is essential in advancing smart healthcare, retail, manufacturing, and urban management.
Fueling Innovation
NVIDIA DGX systems are at the forefront of driving innovation from classrooms to startups, where brilliant minds of today are solving the toughest challenges of tomorrow. These systems provide students and emerging entrepreneurs with the high-performance AI computing necessary to explore and develop advanced solutions across various fields. By offering access to the same cutting-edge technology used in industry-leading AI research, NVIDIA DGX systems enable classrooms and startups to experiment with complex data models, foster creativity, and accelerate learning and innovation.
AMAX as your NVIDIA DGX Solution Provider
At AMAX we specialize in designing and building on-premises compute solutions to empower and accelerate your AI and HPC workloads.