AI workload growth is driving structural increase in storage and memory demand at datacenter scale.
At the 2023 Nvidia Global Technology Conference (GTC), CEO Jensen Huang announced that artificial intelligence is at an inflection point, with generative AI generating a new wave of opportunities across various diverse data and applications. He highlighted that Chat GPT, a generative AI platform, generated over 100 million users in just a few months. Jensen showcased AI applications spanning medical science including genomic analysis and designing new biochemicals, semiconductor manufacturing, product design, automobile design and manufacturing, and digital twins for designing factory and warehouse robotic automation as well as creating and editing professional image and videos.
Jensen announced new products and services designed to support more sophisticated and powerful AI across more industries. These offerings provide productivity boosts through AI capabilities, including the ability to develop working programs using human language that generative AI can convert into software—enabling individuals to program without extensive technical training. NVIDIA announced its intention to reinvent itself as a provider of software-driven services and acceleration libraries for numerous applications.
The company's new hardware reflects the substantial computational and storage demands of AI training. NVIDIA's GRACE CPU for AI and cloud workflows includes 1TB of memory, while the GRACE Hopper CPU Superchip for large-scale AI and high performance computing (HPC) applications is designed to provide 10X higher performance than past devices for applications using terabytes of data. This superchip includes 96GB of high bandwidth memory (HBM) positioned close to the processor chip and incorporates the company's BlueField 3 digital processing unit (DPU) and 4th generation NVLink, running all NVIDIA software stacks and platforms including the NVIDIA HPC SDK, AI, and Omniverse.
At GTC, DDN announced compatibility of their A3I storage appliances with the next generation of NVIDIA DGX to support AI training models requiring large data models and high-speed throughput. According to DDN, "Offered as part of DDN's A3I infrastructure solution for AI deployments, customers can scale to support larger workloads with multiple DGX systems. DDN also supports the latest NVIDIA Quantum-2 and Spectrum-4 400Gb/s networking technologies. Validated with NVIDIA QM9700 Quantum-2 InfiniBand and NVIDIA SN4700 Spectrum-4 400GbE switches." The DDN AI400X2 storage appliance was presented to address the doubled IO performance requirements of DGX H100 systems.
DDN also announced a partnership with Lambda to deliver a scalable data solution based on NVIDIA DGX SuperPOD with over 31 DGX H100 systems. Lambda intends to use these systems to allow customers to reserve between two and 31 DGX instances backed by DDN's parallel storage and the full 3200 Mbps GPU fabric, providing rapid access to GPU-based computing without commitment to large data center deployments. The 2023 NVIDIA GTC demonstrated the company's continuing support for AI modeling and inference infrastructure alongside new software and service offerings, with the recognition that this infrastructure requires significant storage and memory to train and run these models.