The Future of High-Performance Computing
The technological advances that pervade and enrich our daily lives continue to surprise and excite us year after year. From artificial intelligence sitting on our countertop (we’re looking at you, Alexa), to our touch screen controlled Tesla sedans, high-powered technology is proving that the future is now.
Also known as parallel computing and supercomputing, high-performance computing (HPC) aggregates data processing power to deliver efficient, reliable, and rapid results. In other words, high-performance computing takes problem-solving to the next level.
For example, a standard 3 GHz processor is capable of performing 3 billion calculations per second, while a high-performance computer can perform quadrillions of calculations per second.
Although this high-powered machinery may not be a necessity for everyday PC users, HPCs are the crucial game-changing tech that engineers, scientists, and mathematicians will carve the future with.
So what exactly is high-performance computing? Why is it important and what does the future of high-performance computing look like? Let’s dive in and find out.
What is high-performance computing?
High-performance computing is a form of digital computation that allows users to tackle and solve complex problems at a faster rate than standard PCs allow. Where a general-purpose PC may struggle to bring a large-scale simulation to life, a supercomputer delivers instant calculations accompanied by stunning visuals within moments. Engineered for top-of-the-line performance and mind-blowing speed, HPCs have become society’s answer for sophisticated computation.
Then and now in HPC
Dating back to the 1960s, high-performance computers were originally introduced to the market by Seymour Cray at Control Data Corporation. With each passing year, these supercomputers grew stronger and faster with the addition of extra core processors and increased parallelism. Since their inception, high-performance computers have come an incredibly long way. Modern HPCs are equipped with tens of thousands of processors, capable of massive parallel computing that standard laptops and desktops can’t compare to.
Take what you know about a modern computer setup, and replace it with a number of grouped racks stacked up with thousands of processor cores. IBM’s “Summit” [1] is the world’s fastest supercomputer. Stationed at Oak Ridge National Laboratory, this $200,000,000 HPC is comprised of 256 racks and takes up 5,600 square feet of floor space within the Tennessee-based lab. With 185 miles of cable wiring and the electricity consumption to power 8,000 homes, Summit displays the monumental progress HPCs manufacturers have made since the first supercomputer.
Back in 1964 when the CDC 6600 was the one and only leader in supercomputing, it was humbly equipped with a single processor capable of completing 3 million calculations per second. While this may sound impressive, the modern smartphone is tens of thousands of times faster. Even leading into the 1990s, high-performance computers dished out slower processing speeds than an iPhone X. Today, Summit houses 36,000 processors capable of completing 200 quadrillion calculations per second.
How are high-performance computers constructed?
The perplexing construction of a high-performance computer can be broken down simply. All supercomputers contain the many elements you associate with your general-use PC; processors, memory, and disk storage. HPCs simply have many more of each.
Supercomputers are constructed by the connection of multiprocessor computer nodes to a common memory. These many processors are linked together by a high-speed interconnection network, also called a switch.
Most supercomputers utilize operating system software stemming from UNIX. Every supercomputer is different in design and configuration, but all HPCs follow this general construction formula.
With leadership-class computational performance as the priority, modern supercomputers like the Summit boast hybrid configurations that combine features from shared memory machines and distributed memory clusters. This refined aggregation equips each of the Summit’s 4,608 nodes with multiple IBM Power9 CPUs and NVIDIA Volta GPUs effectively powering every node with over half a terabyte of high bandwidth memory plus DDR4 memory addressable by all CPUs and GPUs. Connected by NVIDIA’s high-speed NVLink, these powerful nodes give way to future innovation today.
The performance capability of a supercomputer is measured in teraFLOPS. Floating-point operations per second, or FLOPS, are most useful for scientific problem solving that requires floating-point calculations. Theoretical peak performance rates calculated by multiplying the number of processors by clock speed are typically expressed in FLOPS. To add perspective, the Summit’s theoretical peak speed is 200,000 teraFLOPS.
Just for comparison; what would take a standard PC 30 years to compute, the Summit can complete in one minute.
What is high-performance computing used for?
Given the nature of these extremely high-powered, high-performance computers, the unprecedented capacity for problem-solving lends itself to benefit society in a big way. Supercomputers provide the tools needed for a number of modern innovations stretching from smart appliances to nuclear test simulations.
Not only do supercomputers enable accelerated scientific research and computation, but they can also augment or replace experiment execution that may be too dangerous or impossible to complete in real life settings.
HPCs are engineered to take on the mighty tasks that your standard computer could never fathom. Scientists have used HPCs to generate climate models that provide visual insight into the evolution of climate over centuries. Capable of collapsing time and analyzing patterns, supercomputers have been an essential part of predicting future global climates and evaluating data supporting modern climate change.
Scientists have also taken the skies by storm, launching supercomputers into space for exploration and data collection. HPCs have thousands of uses that ultimately work to improve the technological scope of our world, changing lives and saving lives with incredible clock speed.
What is the future of high-performance computing?
The rapid development of supercomputers is understood when looking at the progression of speed and capacity in the past five years.
Supercomputer Titan debuted in 2012 boasting what was then an impressive 1.4 teraFLOP node performance with 18,688 nodes. 2018’s Summit delivered five times the computational power of the Titan with only 4.608 nodes.
The future of high-performance computers focuses on efficiency, making more with less. The future of HPCs lies in biosciences, climate modeling, geographical data collection, and many other disciplines. Supercomputers will be used to edit feature-length films and stream live even across the globe.
Research labs will be strengthened by HPCs that assist scientists in their quest to find renewable energy and analyze future and past universe evolutions. HPCs can even be used in medical fields to develop cures for diseases and deliver faster and more accurate diagnoses.
The future of high-performance computers is a bright one, and it’s unfolding right before our eyes.
[1] Oak Ridge National Library; Summit
