In the hit film “Chappie” about a futuristic robot, audiences are taken to a world where technology is used in exciting and unexpected ways. In a scenario that put a smile on the face of AMDers worldwide, in one scene a number of Sony PlayStation 4 consoles, each with AMD Radeon™ graphics and AMD CPU processing inside, are connected together, effectively combining into a supercomputer cluster. While “Chappie” is a Hollywood creation, the idea behind connecting game consoles to create a supercomputer is not as science fiction as we may think. Clustering allows almost any device to become part of a supercomputer by being part of a pool of resources.
The world’s first supercomputers used expensive, specialized processors that put high performance computing out of reach most organizations, but today, many of the most powerful supercomputers make use of off-the-shelf processors to produce “nodes.” These nodes are then connected together to form clusters, similar to what we see with an array of AMD-powered PS4s in “Chappie”.
Supercomputer clusters typically make use of server-orientated processors such as the AMD Opteron™ family of processors, but occasionally academics and engineers want to try something different and use less specialized computing resources. Because combining nodes to form a cluster is done using widely available equipment such as Ethernet, it is possible to make a cluster from a number of other devices, such as the Gizmo 2 or game consoles, for example.
To create the supercomputer cluster in “Chappie”, a number of PlayStation 4 consoles, each packing 1.84 teraflops of AMD computing power inside are connected together. That processing power is delivered in a System-on-Chip design with eight AMD CPU cores and 18 AMD Radeon Graphics Core Next (GCN) compute units, typically used to process all the software, games, videos, and motion sensor capabilities that enable players to interact with others through online services. Even more impressive, the same compute architectures found in the PlayStation 4 really are powering some of today’s supercomputer clusters.
In recent years, high performance compute clusters achieved their immense compute capabilities through a mix of processor technologies. The CPU, which has been at the heart of supercomputers for decades has been supplemented with accelerators such as AMD FirePro GPUs, which provide immense compute capability and do so in an energy efficient manner.
Energy efficiency is a key to making a powerful high performance compute cluster given cooling thousands of processors is both a budgetary and engineering challenge. That is why power efficient CPUs and GPUs make ideal components to build compute clusters and have been a significant contributing factor in the massive growth in the amount of compute power in supercomputer clusters. Such is the efficiency of GPUs, the world’s most energy efficient cluster makes use of AMD FirePro accelerators.
By pooling compute and storage resources, academics, scientists, corporations and governments are able to tackle some of the world’s most challenging questions, from medical research to mineral exploration. In many cases, supercomputer clusters are made up of more than 10,000 nodes, with more than 100,000 compute cores, each core working towards producing a final result. And if you think that only a scientist gets to feel the effects of a supercomputer cluster you may be surprised. When you glance at the weather forecast in the morning you may be surprised to find out that it is in part calculated by running immensely complex forecasting models on clusters. Cool, right? Now think about this; supercomputers do even more than just science and predicting the weather. The car you drive has most likely been partly designed and tested using a supercomputer that is analyzing computational fluid dynamics to increase fuel efficiency.
One might say that supercomputer clusters have played a significant part in the world around us and with the growth in machine learning and big data analysis; it is set to play an even bigger role in our daily lives in the future. Perhaps the most amazing part in all this is that supercomputer clusters are powered by the same type of technology that is found under your desk or television and, you probably didn’t even realize how powerful these common devices actually can be.
John Taylor, is the Corporate Vice President of Marketing at AMD. His/her postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites, and references to third party trademarks, are provided for convenience and illustrative purposes only. There is no endorsement, express or implied, of this blog by the movie “Chappie”. Unless explicitly stated, AMD is not responsible for the contents of such links, and no third party endorsement of AMD or any of its products is implied.
*Originally Posted by llatif in AMD Business on Apr 28, 2015 1:56:37 PM