Accelerating HPC and AI: From the Desktop to the Data Center
- March 19, 2019
Coming to NVIDIA GTC
High-performance computing and artificial intelligence might conjure up visions of futuristic data centers. However, HPC and AI workloads — including research, CAD, and data analysis — typically begin on a laptop or desktop workstation. From there, teams can tap into data center resources.
Diagram from TACC user guide for Stampede2.
Dell Precision Workstations with NVIDIA GPUs can accelerate anything from gaming to data analytics, design to visualization. On my personal wish list for design, is the Dell Precision 7730 mobile workstation with up to 8.9TFLOPS with the NVIDIA Quadro P5200 GPU, and 8TB of storage! With up to 64GB of memory, it easily tackles virtual reality, analytics and graphics/design. If you don’t have enough room for the 17” display in your motorcycle backpack, check out the sleek Precision 5530 with Ubuntu® Linux. It has memory expansion capabilities up to 64GB for a good balance of price and performance in a mobile workstation.
In cool new tech, the NVIDIA Quadro RTX 8000 can render complex models and scenes with physically accurate shadows, reflections, and refractions, to empower users with instant insight. And it will be available in Dell Precision 5820, 7820 and 7920 towers.
Extending the gains with GPU virtualization
The NVIDIA virtualization platform now extends the power of GPUs to support up to 32 virtual desktops running performance-hungry workloads. Often these solutions leverage NVIDIA virtual GPUs (vGPUs) and familiar VMware technologies used in data centers around the world, with everything integrated into VMware Horizon desktop and application environments.
Plenty of Vroom with GPUs
Dell EMC PowerEdge servers including the R640, R740/xd and R7425, can get a boost from NVIDI T4 Tensor Core GPU accelerators. NVIDIA T4, with 320 Tensor Cores and 70W low-profile design, can support a wide range of workloads from machine learning to virtual desktops. From the PowerEdge T640 tower servers to R940xa rack servers, the NVIDIA V100, equipped with 640 Tensor Cores, delivers 125 teraFLOPS of deep learning performance. In the PowerEdge C4140 server, with NVIDIA NVLink interconnect technology, V100 accelerators can be interconnected at up to 300GB/s to unleash even more application performance. And there’s more good news: GPUs can accelerate more than 580 HPC applications.
Take it with you
Via the NVIDIA GPU Cloud (NGC), you can bring your deep learning containers with you wherever you’re working, while reducing complexity. For example, NVIDIA RAPIDS open-source software is available in containers accessible via the NGC container registry. This suite of open source software libraries is designed to give people the freedom to execute data science and analytics pipelines on GPUs.
The fastest-adopted GPU now in the world’s best-selling PowerEdge Server
The new NVIDIA T4 accelerator for distributed computing environments can accelerate training and inference, video transcoding, and virtual desktops. It can also accelerate mainstream applications in enterprise environments, enabling companies to analyze massive amounts of data and make accurate business predictions at unprecedented speed.
NVIDIA reports that this next-level GPU accelerator has become the fastest-adopted server GPU. And, it is now available in one of the world’s best-selling servers, the Dell EMC PowerEdge R740. The T4 GPU is also available in PowerEdge R740xd, R640 and R7425 servers.
To learn more
If you’re attending the NVIDIA GPU Technology Conference (March 17–21) in San Jose, be sure to stop into the Dell Technologies booth #1311 for a first-hand look at systems that accelerate HPC and AI from the desktop to the data center. In the meantime, explore Dell EMC solutions for high-performance computing and artificial intelligence.