Nvidia expands solutions for ‘Accelerated Computing’

Get a free Techzine subscription!

Nvidia will significantly expand its solutions for Accelerated Computing in the near future. More options via SDk’s, quantum computing and zero trust security are underway.

According to Nvidia, accelerated computing is a collective term for the use of specialized hardware applications to speed up processing power. This involves the use of parallel processing for frequent computational tasks. In particular, it must be able to offload computational workloads from CPUs and GPUs.

Data Processing Units (DPUs) are used for this purpose. In Nvidia’s case: BlueField DPUs. These processors can be used to handle computational tasks for highly data-intensive applications such as AI workloads. Resultingly, ‘normal’ CPUs and GPUs can handle their own computational and processing tasks more easily.

New SDKs

Recently, Nvidia announced a series of new capabilities that should extend accelerated computing. The most notable announcement is the release of 65 new SDKs that aim to provide more computing power options for researchers, data scientists and developers. The SDKs will be added to the existing SDK catalogue of 150 diverse solutions.

Among the new SDKs are solutions for real-time logistics tasks (so-called array computing) and AI applications such as graph neural networks. Also, there’s a specific SDK for quantum computing. In addition, several updates have been released for existing, popular SDKs for video analytics, deep neural networks and data center networking.

Quantum computing

In addition to the new SDKs, Nvidia announced the expansion of its initiatives in quantum computing. To this end, the chipset specialist is introducing the NVIDIA Quantum-2 platform. This Infiniband-based network platform for quantum computing should provide the basis for improved performance, broader accessibility to quantum computing technology and solid security, specifically aimed at cloud providers and supercomputing centers.

The NVIDIA Quantum-2 platform consists of the NVIDIA Quantum-2 switch, the ConnectX-7 network adapter and a BlueField-3 DPU. In addition, software to support the platform and the underlying architecture is included. According to Nvidia, the platform should eventually deliver a data throughput of 400 Gbps.

The Nvidia platform for quantum computing is now available on various infrastructures, including Atos, DataDirect Networks (DDN), Dell Technologies, Excelero, GIGABYTE, HPE, IBM, Inspur, Lenovo, NEC, Penguin Computing, QCT, Supermicro, VAST Data and WekaIO.

Zero trust security platform

Furthermore, to enhance its accelerated computing concept, Nvidia announced the arrival of a zero trust-based security platform. This platform combines the functionality of the Nvidia DOCA 1.1 zero trust solution, the BlueField-3 DPUs and the Nvidia Morpheus AI framework for cybersecurity. According to the company, this combined platform should bring greater security to data centers, including isolating applications from infrastructure and optimizing next-generation firewalls through accelerated computing and deep learning.

Image source: DANIEL CONSTANTE / Shutterstock.com