SAP Upgrades AI Platform With Volta GPUs
Building on collaboration announced earlier this year, SAP said this week it would incorporate Nvidia's AI platform based on the graphics processing specialist's Volta architecture into its machine-learning portfolio.
The upgrade to SAP's Leonardo machine learning suite includes the addition of Nvidia's Tesla V100 GPUs. The new Nvidia DGX-1 systems will be used for AI development projects as part of a hardware platform for SAP's machine learning foundation. SAP claims the combination represents the first enterprise platform using the Volta architecture.
SAP noted in a blog post that it has been using Nvidia's processors to train algorithms and data sets as part of its machine learning application development. Nvidia asserts that its Volta-based DGX-1 system provides a three-fold increase in training and a four-fold boost in inference. The combination has been used to train machine learning-based applications such as image classification and feature extraction along with product image classification.
SAP began installing DGX-1 systems at development centers in 2016, including Nvidia's Tesla P100 GPUs in production datacenters in Germany, Singapore and Palo Alto, Calif. The facility in Germany has now been upgraded with V100 processors, the partners announced at a conference in Munich on Tuesday (Oct. 10).
Each DGX-1 includes eight P100 GPU accelerators, or the equivalent of 250 two-socket servers, used to train deep learning models that parse unstructured data and spot patterns used to support decisions.
This week's announcement boosts the processing power behind SAP's machine learning platform. Hailing Volta as "the new driving force behind artificial intelligence," Juergen Mueller, SAP’s chief innovation officer, said the company would leverage the new Tesla processors to embed machine intelligence into "a wide spectrum of enterprise scenarios,"
Hence, SAP said it would incorporate its Leonardo machine learning capabilities into next generation S/4HANA enterprise resource planning suite with an eye toward automating repetitive tasks while delivering real-time analytical results to customers.
Among the machine learning applications trained using Nvidia (NASDAQ: NVDA) GPUs are a cloud-based tool for analyzing video to detect brand logos. The tool also measures other brand metrics with the goal of helping marketers target advertising budgets.
Another leverages deep neural network technology to prioritize customer service requests based on analysis on unstructured data like text messages. Recommendations are based on previous responses to service requests, the company said.
The machine learning push comes as enterprise AI adoption continues to expand. According to a survey released this week by business analytics vendor Teradata (NYSE: TDC), 80 percent of enterprises have some form of machine or deep learning in production. While many expect to double the return on their AI investments over the next five years, lack of IT infrastructure and in-house expertise are seen as the key stumbling blocks to enterprise deployment, the survey found.
In the meantime, major players like SAP (NYSE: SAP) are betting that GPU processing power will help accelerate model training and inference capabilities as AI pushes deeper into enterprises. SAP and Nvidia announced their AI partnership in May.
