Technology giant IBM has introduced its much anticipated Power 9 Chip, a POWER9-based server, and support from Google and Department of Energy CORAL. IBM.
Technology giant IBM has introduced its much-anticipated Power 9 Chip, a POWER9-based server, and support from Google and Department of Energy CORAL. IBM.
The company is intending to sell these chips to a third-party manufacturer, as well as cloud vendors like Google. In the meantime, IBM is releasing a brand new computer that is powered by this chip, the AC922, with the intention to offer these chips on a service situated on IBM cloud.
“We are generally taking our technology in order to market as a full solution”, Brad McCredie who is an IBM fellow & vice president of the cognitive systems, explained.
The company has already designed this new chip with a specific aim of improving performance on the common AI systems such as Caffe, Chainer, & TensorFlow which claims a rise of workloads that run on these systems by up to nearly four times.
In case it functions as described, it should provide the data scientists me building models & run them on any Power9-powered gadget with increased speed, that should permit them to run their jobs faster & complete the creation of models much quicker.
Patrick Moorhead who is a principal analyst at the Moor Insights & Strategy has a belief that IBM has actually excluded itself from any competition with this chip. The Power9 chip contains a new systems architecture which has been optimized for the accelerators utilized in this machine learning.
The IBM’s Power9 is generally the Swiss Army knife since it normally supports a good amount of IO & bandwidth, ten times of what is out there currently, as Moorhead said.
In case you think that Nvidia appears to have caught a perfect deal of the Artificial Intelligence or the machine learning workloads, it did not manage to escape the IBM’s notice either since they worked closely with the maker of the GPU chip. To be precise, McCredie states that IBM constructed a system bus which moves the workloads between the 2 chip types faster than any competitive systems.
Current workloads are now accelerated & the Nvidia GPU is just like any other accelerator.
“We have observed the trend unveiling. We established a strong relationship with them & a partnership among the Power system & the GPU. We now have a bus that is unique which runs across the processor & the GPU having 10x peak bandwidth over the competitor systems”, McCredie stated.
The current chips will power a supercomputer known as the Summit that is being constructed by Lawrence Livermore & Oakridge national laboratories. He states that the supercomputer is going to be constructed on top of various Power9 computers which will go at a price of $325 million.
Chirag Dekate, a research director at the HPC, machine learning & the rising computing technologies at Gartner, states that this release is just a continuation of the IBM’s vigorous approach of capturing high-developing market segments such as AI or artificial intelligence. Advances in sectors like AI allows IBM to compete well in the hyper-scale data center and the wider market data center forums. This has the potential to drive revenue for IBM, as well as to allow larger scale data center deployments, according to Dekate.
The chip is available as of December 5, 2017.