AI Supercomputer TX-GAIA

Think of the internet as a network that connects people through the web pages or chats. Presently, over 5 billion people will be connected to the internet and 2020 the numbers are expected to reach 25 billion with the global annual traffic expected to exceed the equivalent of 500 billion DVDs. Only powerful super computers are able to support massive rapid computations can cope with the ever-increasing amount of data.

So as to power the AI applications and research across science, engineering and medicine the Massachusetts institute of Technology (MIT) Lincoln laboratory supercomputing center has installed a new GPU-accelerated supercomputer that is powered by 896 NVIDIA Tensor Core V100 GPUs. It is ranked as the most powerful super computer in the world.

The introduction of Artificial Intelligence into the work place has brought diversity. The new super computer has a peak performance of 100 AI peta FLOPs as measured by the computing speed which is required to perform mixed precision floating point operations commonly known as deep neural networks.

The system features a measured performance of around 5 petaFLOPs and is based on the HPE Apollo 2000 system which is specifically designed for the HPC and optimized AI. Deep neural networks continue to grow in size and complexity with time.

The new TX-GAIA computing system at the Lincoln laboratory has been ranked as one of the most powerful artificial intelligence supercomputers in any university. The system which was built by Hewlett Packard Enterprise combines traditional high-performance computing hardware with almost 900 intel processors and hardware that is optimized for AI applications in addition to the use of Nvidia graphics processing applications.

Machine-learning supercomputer

The new TX-GAIA supercomputer is housed within the EcoPOD modular data center and was first revealed to the world in 2011. The system joins other machines in the same location including the TX-E1 which supports collaboration with MIT campus and other institutions. Researchers at the institution are thrilled to have the opportunity to achieve incredible scientific and engineering breakthroughs.

Top 500 ranking

Top 500 ranking is based on LINPACK Benchmark which is basically a measure of a system’s floating-point computing power or how fast a computer solves a dense system of linear equations. The TX-GAIA’s performance is 3.9 quadrillion floating-point operations per second. Or rather petaflops. It has a peak performance of 100 petaflops which makes top any other in any university in the world. A flop is basically a measure of how fast a computer can perform deep neutral network (DNN) operations. DNNs basically refer to a class of algorithms that learn to recognize patterns in huge amounts of data.

Artificial intelligence basically has given rise to various types of miracles in the world which include speech recognition and computer vision. It is this kind of technology that allows Amazon’s Alexa to understand the questions and self-driving cars to recognize objects in their surroundings. As the complexity of the DNNs grow so is the time it takes for them to process massive amounts of datasets. Nvidia GPU accelerators that are installed in TX-GAIA’s are specifically designed for performing these DNN operations quickly.


TX-GIAA is housed in a modular data center called an EcoPOD at the LLSC’s green, hydroelectrically powered site in Holyoke Massachusetts. It joins the ranks of some of the most powerful systems at the LLSC such as the TX-E1 which supports a collaboration with MIT campus and other users.

TX-GAIA will be tapped for training machine learning algorithms which include those that use DNNs. This implies that it will more likely crunch through terabytes of data for instance hundreds of thousands of images or years’ worth of speech. The systems computation power will be able to expedite simulations and data analysis and these capabilities will be able to support projects across R&D areas. This include improving weather forecasting, building autonomous system, accelerating medical analysis, designing synthetic DNA as well as in the development of new materials and devices.

Why supercomputing?

High-performance computing plays a very important role in promoting the scientific discovery and addressing of grand challenges as well as in the promotion of social and economic development. Over the past few decades, several developed countries have invested heavily into a series of key projects and development programs. The development of supercomputing systems has advanced parallel applications in various fields along with related software and technology.

Significance of super computing

A super computer is a high-performance computing which does not necessarily refer to a very large or powerful computer. A super computer comprises thousands of processors working together in parallel and it responds to the ever increasing need to process zillions of data in real time with quality and accuracy. HPC allows people to design and simulate effects of new drugs, provide faster diagnosis, better treatments and control epidemics as well as support in the decision-making process in areas such as water distribution, urban planning and electricity.

A supercomputer is of great benefit in a competitive industry as it helps in the digitization process. It also helps to direct benefits to our health in that super computers are able to detect genetic changes and it also comes in handy during weather forecasting.

The next wave of AI

The adoption of artificial intelligence has exploded in the last few years with virtually every kind of enterprise being on the rush to integrate and deploy AI methodologies in their core business practice. The first wave of artificial intelligence was characterized by small scale proof of concepts and deep learning implementations. In the next wave we will be able to see large scale deployments which are more evolved and concerted effort to apply to AI techniques in production to solve real world problems and drive business decisions.

Artificial Intelligence basically is a supercomputing problem and is expected to double in size within the next few years. AI thrives on massive data sets and there is a great convergence that occurs between AI and simulation. Most of the organizations that are performing simulation are increasingly adding machine and deep learning into their simulation.