RAPIDS Momentum Builds with Analytics, Cloud Backing
Nvidia’s RAPIDS data science libraries picked up additional support this week from a roster of cloud, server and workstation vendors along with professional services giant Accenture.
Nvidia also announced that Apache Spark creator Databricks will integrate RAPIDS into its analytics platform. These and other integrations are aimed at data scientists using RAPIDS and other machine learning languages.
Databricks said RAPIDS open-source software packages can be installed on its runtime for Nvidia GPUs used to accelerate machine learning development.
The GPU leader (NASDAQ: NVDA) also announced during its technology conference that workstation makers Dell (NYSE: DELL), Hewlett Packard Enterprise (NYSE: HPE) and Lenovo (OTCMKTS: LNVGY) will integrate Nvidia’s Quadro RTX GPUs along with its CUDA-X AI data science software. HPE, IBM (NYSE: IBM) and Oracle (NYSE: ORCL) were early backers of RAPIDS.
The GPU-powered workstations are designed to allow data scientists to accelerate the processing of large data sets on full data pipelines. For instance, Lockheed Martin (NYSE: LMT) said it is using RAPIDS to push more data processing to local GPUs to reduce model development time.
Nvidia’s Quadro RTX platform is based on its latest Turing GPU architecture, delivering up to 260 teraflops of computing horsepower and 96 gigabytes of memory running on Nvidia’s NVLink interconnect technology.
Nvidia’s CUDA-X AI software is a collection of libraries that run computing applications on Nvidia’s GPU computing platform. Along with RAPIDS, the data science workstation also runs Caffe, PyTorch and TensorFlow machine learning libraries.
The workstations are available now from the hardware manufacturers along with more than a dozen system builders, Nvidia said.
Adding to the momentum, Accenture said it would integrate RAPIDS with its new business intelligence platform. The professional services company said the combination would be used as part of a joint venture with Japan’s Kansai Electric Power (TYO: 9503) to “transform business process and create new opportunities.”
Nvidia also announced RAPIDS integrations by Google Cloud and Microsoft Azure. Google (NASDAQ: GOOGL) will use the ML library to accelerate virtual machines and Kubeflow, the machine learning stack built for Kubernetes. Microsoft (NASDAQ: MSFT) will integrate RAPIDS into its machine learning service.
RAPIDS software also focuses on data preparation tasks for analytics and data science, Nvidia notes.
Nvidia previously said the RAPIDS platform uses the XGBoost machine learning algorithm for training on Nvidia DGX-2 supercomputers, yielding what it claims is a 50-fold performance boost over CPU-only systems.
The RAPIDS ecosystem includes Databricks’ web-based platform for big data processing in the cloud using Apache Spark and Anaconda, the open source distribution of the Python and R programming languages for data science and machine learning.
Editor’s note: This story has been updated.