AI at Inflection Point Drives Demand for Storage and Memory

Jensen Huang, CEO of the 2023 Nvidia Global Technology Conference (GTC), said that artificial intelligence (AI) is at an inflection point and that generative AI will open up new opportunities for a wide variety of diverse data and many applications. Said it was creating waves. He said his Chat GPT, a generative AI, has generated over 100 million users in just a few months.

Below is a summary slide from his presentation on AI, highlighting various new hardware and software products and some of the company’s partnerships to deliver various AI services, especially via the cloud (such as big data centers). is showing. In his talk, Janssen described his digital twin, professional image and digital for the design of genomic analysis and new biochemicals, semiconductor manufacturing, product design, automotive design and manufacturing, and robotic automation in factories and warehouses. It showed applications in medical sciences such as video creation and editing.

Jensen unveiled new products and services to support more sophisticated and powerful AI in more industries. These new services use AI to improve productivity. One example is that human language can be used to develop practical programs that generative AI can translate into software. This means that anyone can do some programming without extensive technical training. Jensen said NVIDIA wants to reshape itself as a provider of software-driven services and provide acceleration libraries for many applications.

AI training in particular consumes large amounts of digital storage for modeling data and memory to support processing that data. NVIDIA’s GRACE CPUs for AI and cloud workflows include 1 TB of memory. The company’s GRACE Hopper CPU Superchip is designed for large-scale AI and high-performance computing (HPC) applications, delivering 10x better performance than previous devices for applications with terabytes of data It is designed to contain 96 GB of High Bandwidth Memory (HBM) close to the processor chip. You can see this chip connected to the GPU below.

Runs all NVIDIA software stacks and platforms, including the NVIDIA HPC SDK, as well as AI and Omniverse. The product includes the company’s BlueField 3 digital processing unit (DPU) and 4^th Generation NVLink.

At the GTC DDN, we announced A compatibility.³I storage appliance with next-generation NVIDIA DGX to support large data models and AI training models that require fast throughput. According to DDN, “Available as part of DDN’s A.³An 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 network technologies. Validated with NVIDIA QM9700 Quantum-2 InfiniBand and NVIDIA SN4700 Spectrum-4 400GbE switches. ” To double the IO performance of a DGX H100 system, you need a high-performance storage solution that can support that performance. The DDN AI400X2 storage appliance is shown below.

In addition to these on-premises deployment options, DDN said it is also announcing a partnership with Lambda to offer scalable data solutions based on NVIDIA DGX SuperPODs with more than 31 DGX H100 systems. Lambda plans to use this system to allow customers to reserve between 2 and 31 of his DGX instances backed by DDN’s parallel storage and a full 3200 Mbps GPU fabric. This hosted offering provides rapid access to GPU-based computing without the promise of large-scale data center deployments.

NVIDIA GTC demonstrated NVIDIA’s continued support for AI modeling and inference infrastructure, as well as new software and service offerings. This infrastructure requires large amounts of storage and memory to train and run these models. DDN showcased its latest storage appliance for the DGX H100 system.

Follow me please twitter or LinkedIn. check out my website.