IBM unveiled Power9 processor, and implemented it to build next-generation Power Systems Servers, focused on artificial intelligence workloads.
IBM claimed that new power systems (called AC922 Power Systems), can improve performance by four times across common AI frameworks including Chainer, TensorFlow, Caffe, and Kinetica. It will help enterprises to build more accurate AI applications in less time.
In addition to performance improvement, AC922 feature the latest version of Peripheral Component Interconnect Express (PCIe Gen4) expansion bus, as well as NVIDIA NVLink and OpenCAPI, the combination of which speed up the movement of data by up to 9.5 times over x86.
If AC922 works as described by IBM, it can benefit data scientists with optimized and simplified deployment of deep learning frameworks and libraries, so that they can easily perform their tasks like accessing deep leaning insights in scientific research and detect real-time fraud.
Power9 and AC922 will be implemented in Summit and Sierra supercomputers, which are tapped by Google and US Department of Energy, and are expected to be the most powerful data-intensive supercomputers of tomorrow.
“We’ve built a game-changing powerhouse for AI and cognitive workloads,” said Bob Picciano, SVP of IBM Cognitive Systems. “In addition to arming the world’s most powerful supercomputers, IBM POWER9 Systems is designed to enable enterprises around the world to scale unprecedented insights, driving scientific discovery enabling transformational business outcomes across every industry.”
“Google is excited about IBM’s progress in the development of the latest POWER technology,” said Bart Sano, VP of Google Platforms. “The POWER9 OpenCAPI Bus and large memory capabilities allow for further opportunities for innovation in Google data centers.”
Also read: IBM adds new features to Watson to advance AI applications
IBM said that it has worked for four years to design the Power9 processor, aiming to build an architecture that can manage free-flowing data, streaming sensors and algorithms for AI and deep learning workloads on Linux.
The Power9 chip is generally available now.
