Nvidia Adds Scaling Software Stack to AI Cloud Computing

Nvidia doesn’t have all the internal pieces to build its massive AI computing empire, so it’s bringing in software and hardware partners to scale its so-called cloud-based AI factories.

The chipmaker’s latest partnership is with Rescale, which provides the middleware to orchestrate high-performance computing workloads across public and hybrid clouds.

Rescale adopts Nvidia’s artificial intelligence technology for what the middleware vendor has called the world’s first compute recommendation engine, which automates the selection of available compute resources in the cloud for high-performance applications and simulation. said Edward Hsu, chief product officer at Rescale.

“We have over 1,000 applications and releases from major simulation software vendors, we work with all major cloud vendors and specialty architecture vendors to truly optimize the performance of computational science and engineering workloads “, said Hsu.

Rescale will also offer customers Nvidia’s AI Enterprise software platform, which includes pre-programmed AI models such as Modulus, which is used for scientific applications and simulation. Rescale can integrate AI Enterprise with scientific or engineering applications from companies such as ANSYS, Siemens and others for cloud deployment.

Companies are beginning to use more compute-intensive approaches to science and engineering discovery, and the compute recommendation engine helps customers decide what kind of computing to use, Hsu said.

Some considerations for customer computing requirements include architecture type, geographies, data sovereignty, compliance, or regions where more GPU resources are available for applications such as video and artificial intelligence.

“It gets quite complex and what we’ve done is take all of our years extracting that information from the metadata generated by running compute jobs and applying a combination of proprietary software Rescale, benchmarks [and] TensorFlow. Running on an Nvidia GPU really expands that compute recommendation engine,” Hsu said.

Major cloud providers already offer virtualization and orchestration technologies. Rescale already partners with all major cloud providers and additionally adds the compute recommendation engine for customers to optimize the performance of their workloads on these architectures.

More and more companies are adopting a hybrid cloud approach to meet their growing computing needs. Rescale uses Nvidia’s AI Enterprise middleware to match engineering applications to hardware resources across the clouds.

Hsu described the many elements of Nvidia’s partnership as a complete package. For example, if a customer is building a turbocharger for cars, the Rescale platform can provide recommendations on what parameters to change and what compute resources to use to get the best, fastest results.

“The way we’ve integrated Rescale is that they can access all of our software solutions and then run on one of several different Nvidia GPU instances in the cloud,” said Dion Harris, product marketing manager of data center at Nvidia, during a press conference. Report.

Nvidia already dominates the scientific computing space, but Rescale offers HPC customers an indirect way to adopt its AI Enterprise software to train or use AI models.

Nvidia envisions a future where it ties its hardware and software together in a black box. Nvidia is modeling its AI software to run more efficiently on its GPUs. Nvidia’s software stack is based on the CUDA parallel programming framework, and the company is accelerating the deployment of its latest Hopper-based GPUs in the cloud.

Last month, the GPU maker partnered with Oracle to bring AI Enterprise software and thousands of GPUs to Oracle’s cloud infrastructure. The company’s GPUs are also widely adopted by Meta to run AI applications.

Nvidia’s hardware also works with open source parallel programming frameworks such as OpenCL, which is maintained by The Khronos Group. Intel’s OpenAPI parallel programming framework has a tool called SYCL that strips out CUDA-specific code to run on accelerators from Intel and other chipmakers.

Related