IBM’s Power9 processor designed for the AI and ML era

Today, with the growing adoption of new age technologies, the processing power required to operate high-performance analytics leading to insights dwarfs in comparison to anything that existed earlier.

Infact, companies are forced to string together dozens of boxes using a traditional hardware architecture just to start a proof of concept project and they are realizing the pain of the elaborate set up and weeks spent waiting for an outcome. To address this issue, IBM unveiled its POWER9 processor. It is a new architecture that speeds AI applications and insights with groundbreaking new foundational capabilities.

Viswanath Ramaswamy, Director Systems Hardware, IBM India/South Asia talks to BW CIO about the relevance of Power9, the importance of servers in the age of AI and the biggest challenge while addressing it.

 1) Could you shed some light on the Launch of Power9 and its relevance in the market today?

The launch of Power9 system servers is a quantum leap forward towards the Cognitive Era. This is the first of its kind server built for Artificial Intelligence (AI) and Machine Learning (ML). Today, all forms of businesses are using AI and ML extensively and as it becomes more pervasive in the future, it will open up new opportunities and avenues. IDC estimates global spending on AI-related hardware and software will exceed $57B in 2021 from $12B this year.

Keeping this in mind, IBM had set out to design the POWER9 chip on a blank sheet to build a new architecture to manage free-flowing data, streaming sensors and algorithms for data-intensive AI workloads. The architecture aims to transform computing across every industry and profession, turning client data into faster insights. Data scientists will now be able to build applications faster, ranging from deep learning insights in scientific research and radiology imaging, to real-time fraud detection and credit risk analysis.

2) What according to you is the importance of servers in the age of Artificial Intelligence and Machine Learning?

The basic reason is that servers are the foundation blocks needed for running any AI and ML programs. Much has changed in the last four years but nothing has really challenged datacenters so clearly as the dawn of AI era. If we thought data was “big” before, the volume of data required to train a deep learning model is almost unfathomable.

The processing power required to operate the high-performance analytics that lead to insights dwarfs anything that has come before.

With Artificial Intelligence and Machine Learning creating new avenues for businesses, helping improve customer experience and spurring innovation, servers have a critical role to play. IBM PowerAI on Power servers with GPU accelerators provides at least thrice the performance of the x86 platform. We think that the combination of IBM Power and PowerAI is the best platform for AI developers in the market today. For AI, speed is everything — nothing else comes close. 

3) Power9 to support machine learning and AI and data-intensive, deep learning workloads in the enterprise. What sort of future do you envision in this regard few years from now?
The new Power Systems servers will deliver unprecedented scale and speed for deep learning and cognitive workloads. This means, it will improve the training times of deep learning frameworks by 3x, allowing enterprises to build more accurate, deep learning applications faster. It will also enable 10x faster performance bandwidth acceleration and 50% memory bandwidth. In the future, we expect smarter, faster and more intuitive systems leveraging the power of human imagination.
The PowerAI strategy is to create an enterprise software distribution of the open source machine learning / deep learning frameworks – and add value around them. We add ease-of-use tools on top of the frameworks and performance enhancements underneath them. We are also integrating with IBM Data Science Experience (DSX) – is a leading data scientist development tool.

4) What is the biggest challenge that is faced while addressing artificial intelligence?
The single biggest challenge that we heard from customers is that existing x86 systems were not delivering the performance improvements required by new analytics and AI workloads.

 Deep learning is a fast growing machine learning method that extracts information by crunching through millions of processes and data to detect and rank the most important aspects of the data. Four years in the making, we designed the Power9 chip on a blank sheet to address a major industry challenge given that existing x86 systems were designed for an old architecture defined by fixed code, not for free-flowing data, streaming sensors and algorithms.

  - Nivedhana Prabhu