Drawing on the power of their close relationship, Dell Technologies and NVIDIA are streamlining the path to GPU computing.
Thousands of people will be diving into some of the latest and greatest technologies for artificial intelligence and deep learning during NVIDIA’s GTC Digital conference. The online event provides developers, engineers, researchers and other participants with training, insights and direct access to experts in all things related to GPU computing.
The virtual crowd at GTC Digital will include many from Dell Technologies, including data scientists, software developers, solution architects and other experts in the application of the technologies for AI, data analytics and high performance computing.
At Dell Technologies, we’re investing heavily in servers and solutions that incorporate leading-edge GPUs and software from NVIDIA. In this post, I will offer glimpses of some of the exciting things that the Dell Technologies team is working on with our colleagues at NVIDIA.
NVIDIA EGX servers
Dell Technologies was among the first server companies to work with NVIDIA to certify systems for the NVIDIA EGX platform for edge computing. This cloud-native software allows organizations to harness streaming data from factory floors, manufacturing inspection lines, city streets and more to securely deliver next-generation AI, IoT and 5G-based services at scale and with low latency.
Early adopters of EGX edge — which combines NVIDIA CUDA-X software with NVIDIA-certified GPU servers and devices — include such global powerhouses as Walmart, BMW, Procter & Gamble, Samsung Electronics and NTT East, as well as the cities of San Francisco and Las Vegas.
NGC Container Registry
Providing fast access to performance-optimized software, NGC is NVIDIA’s hub for GPU-accelerated AI, ML, and HPC applications containers, software development kits and tools. NGC hosts containers for top AI and data science software, HPC applications and data analytics applications. These containers make it easier to take advantage of NVIDIA GPUs on-premises and in the cloud. Each is fully optimized and works across a wide variety of Dell Technologies solutions.
NGC also hosts pre-trained models to help data scientists build high-accuracy models faster, and offers industry-specific software development kits that simplify developing end-to-end AI solutions. By taking care of the plumbing, NGC enables people to focus on building lean models, producing optimal solutions and gathering faster insights.
Dell Technologies offers NGC-Ready systems for data centers and edge deployments. These systems have passed an extensive suite of tests that validate their ability to deliver high performance when running containers from NGC.
The bar here is pretty high. NGC-Ready system validation includes tests of:
- Single- and multi-GPU deep learning training using TensorFlow, PyTorch and NVIDIA DeepStream Transfer Learning Toolkit
- High-volume, low-latency inference using NVIDIA TensorRT, TensorRT Inference Server and DeepStream
- Data science using RAPIDS
- Application development using the NVIDIA CUDA Toolkit
Along with passing the NGC-Ready tests, these Dell EMC servers have also demonstrated the ability to support the NVIDIA EGX software platform that uses the industry standards of Trusted Platform Module (TPM) for hardware-based key management and Intelligent Platform Management Interface (IPMI) for remote systems management. NGC-Ready systems aim to create the best experience when it comes to developing and deploying AI software from NGC.
Support for vComputeServer
NVIDIA Virtual Compute Server (vComputeServer) with NGC containers brings GPU virtualization to AI, deep learning and data science for improved security, utilization and manageability. And even better, the software is supported on major hypervisor virtualization platforms, including VMware vSphere. This means your IT team can now use the same management tools across the rest of your data center.
Today, vComputeServer is available in Dell EMC PowerEdge servers. And if you’re not sure which PowerEdge server best fits your accelerated workload needs, we can offer some help in the form of an eBook, “Turbocharge Your Applications.” It includes good/better/best options to help you find your ideal solution for various workloads.
Dell Technologies offers a variety of NVIDIA GPUs in the Dell EMC PowerEdge server family. The accelerator optimized Dell EMC PowerEdge C4140 server, for example, offers a choice of up to eight NVIDIA GPUs in configurations with up to four V100 GPUs, or eight V100 GPUs using NVIDIA’s MaxQ setting (150W each). And this is just one of many PowerEdge servers available with NVIDIA GPU accelerators. For a broader and more detailed overview, check out the “Server accelerators” brochure.
Dell EMC Isilon with NVIDIA DGX reference architecture
In another collaboration, NVIDIA and Dell Technologies have partnered to deliver Dell EMC Isilon with NVIDIA DGX reference architectures. These powerful turnkey solutions make it easy to design, deploy and support AI at scale.
Together, Dell EMC Isilon F800 All-Flash Scale-Out Network-Attached Storage (NAS) and NVIDIA DGX systems address deep learning workloads, effectively reducing the time needed for training multi-petabyte datasets and testing analytical models on AI platforms. This integrated approach to infrastructure simplifies and accelerates the deployment of enterprise-grade AI initiatives within the data center.
Dell Technologies is collaborating with NVIDIA, along with Mellanox and Bright Computing, on the Rattler supercomputer in our HPC & AI Innovation Lab in Austin, Texas. The Rattler cluster is designed to showcase extreme scalability by leveraging GPUs with NVIDIA NVLink high-speed interconnect technology. Rattler not only accelerates traffic between GPUs inside servers, but also between servers with Mellanox interconnects. Teams use this system for application‑specific benchmarking and workload characterizations.
The bottom line
NVIDIA and Dell Technologies are delivering some great hardware and software to enable and accelerate AI and deep learning workloads. We’re working closely to help our customers capitalize on all the good things that we are doing together.
- Dell Precision Data Science Workstation for the New Era of AI Productivity [S22535]
- Tuning GPU Server for DL Performance [S21501]
- Quantify the Impact of Virtual GPUs Using NVIDIA nVector [S21988]
- Weakly Supervised Training to Achieve 99% Accuracy for Retail Asset Protection [S21427]
- How NVIDIA Quadro Virtual Workstations, Virtual PCs and Virtual Apps are Transforming Industries [S21670]