Are You Ready for Smarter GPU Acceleration?

Dell Technologies and NVIDIA deliver secure acceleration for AI, data analytics and HPC to tackle the world’s toughest computing challenges.

When it comes to the power of server GPUs, nobody can get enough for AI and HPC for personalized medicine, conversational AI, recommendation engines and other data-driven applications. As a result, the demand for GPUs is soaring.

While data never sleeps, networking stacks can tax CPUs up to 30%, robbing them of the cycles needed to run applications. Smart NICs, with an on-board processor, can provide a wide range of CPU offload benefits and the market is projected to become a $600M market by 2024, or 23% of the total Ethernet adapter market, according to Dell’Oro Group.

NVIDIA BlueField-2 DPUs offloads task from the CPU

As emerging workloads present new challenges on the computational and processing capabilities, there are new approaches to intelligently off-load key tasks to auxiliary computing resources. Enter the new NVIDIA BlueField®-2 Data Processing Unit (DPU)), combining the advanced capabilities of the ConnectX®-6 Dx ASIC network adapter with an array of Arm processors and a high-speed memory controller, which enable enhanced and flexible software programmability. It offloads some network and storage tasks from the server CPU, addressing performance, networking efficiency, with cyber-security.

A good example of this approach may be found at the  Durham Intelligent NIC Environment (DINE) supercomputer. They use BlueField technology to directly access remote memory to improve the performance of massively parallel codes built for future exascale systems. Researchers now have a test-bed facility to develop new and novel computing paradigms. Additionally, NVIDIA’s Mellanox smart network interface card (SmartNIC) will also allow investigations into in-network computing.

PowerEdge Servers give rendering, AR/VR a boost with the upcoming NVIDIA A40

PowerEdge servers are increasingly used in Augmented Reality (AR) and Virtual Reality (VR) applications. These applications place tremendous demand on the compute capacity of the servers. In combination with the upcoming NVIDIA A40 GPU with NVIDIA Ampere CUDA cores, PowerEdge servers are able to accelerate rendering, augmented and virtual reality applications. The powerful DSS 8440 server plans to adopt NVIDIA A40 GPUs deliver deep learning and training performance at lower cost than competitors. These servers have been innovatively designed to have 25% more accelerators in a single chassis enabling 10% more Tensor FLOPS using the same rack density and it provides higher training efficiency (performance/watt) than competitive offerings when using the most common frameworks and popular convolutional neural network (CNN) models.

PowerEdge Servers and NVIDIA A100 Tensor Core GPUs provide smarter GPU acceleration

The need for AI and advanced analytics is growing from the edge to the hybrid cloud. Dell EMC PowerEdge R7525 servers combined with NVIDIA A100 Tensor Core GPUs deliver unprecedented acceleration at every scale for AI, data analytics and HPC.

At Dell Technologies, we support diverse accelerated workloads from edge to core to cloud. For example, Dell Technologies was among the first server vendors to certify the systems for NVIDIA EGX platform for edge computing. And we offer a wide portfolio of NGC-Certified servers that deliver a GPU-accelerated cloud platform optimized for deep learning and scientific computing. Additionally, we offer a broad selection of NVIDIA GPUs in the Dell EMC PowerEdge server family, and in Ready Solutions for Data Analytics, AI and HPC.

Dell Technologies and NVIDIA have a great relationship that will get even stronger as we roll out solutions that incorporate NVIDIA’s new EGX platform. Stay tuned for the upcoming TOP500 and Green500 races, which are sure to have many NVIDIA GPU-accelerated systems on the list.

To learn more, connect with Dell Technologies at NVIDIA GTC, Dell Technologies World, and SC20.

Find out about the power of the Eni supercomputer HPC5, located inside Eni’s Green Data Center has a peak power of over 50 petaflops/s. It is the most powerful industrial supercomputer in the world. Together with its predecessor HPC4, the processing power of the two supercomputers exceeds 70 petaflops/s. It consists of 1,820 Dell EMC PowerEdge C4140 nodes, each with 2x Intel Gold 6252 24-core CPUs and four NVIDIA V100 Tensore Core GPUs accelerators. Watch the video and read the case study.

About the Author: Ravi Pendekanti

Ravi Pendekanti is Senior Vice President of Server Solutions Product Management at Dell, Inc. His organization is responsible for developing and bringing Dell’s flagship line of PowerEdge Servers and Converged Infrastructure systems to market, covering a broad spectrum of global customers from cloud service providers and small businesses to large enterprises and organizations, all running a wide range of workloads. Most recently, Ravi served as the Vice President of the Platform Business Group at Oracle where he looked after GTM activities including Product Marketing and Sales Enablement related to the platform business that included Engineered Systems (Exadata, Exalogic, OVCA and SuperCluster), Servers, Solaris and Networking. With over two decades of extensive global experience in the enterprise and SMB segment in both hardware (servers, storage and networking) and software, Ravi has also held leadership roles at Juniper Networks, Hewlett-Packard, Sun Microsystems and Silicon Graphics. He holds a MS in Computer Science and BS in Electrical Engineering.