Supercomputers and AI Join Forces to Tackle the World's Toughest Challenges
In this article
Global ecosystems are rapidly changing before our eyes. Disruptions in water, food, energy and critical mineral systems are all reaching new levels in recorded history. Moreover, the rate of this change is increasingly outpacing our collective ability to interpret, act on and meaningfully temper these disruptions, causing people and businesses to become overloaded, stressed and imbalanced.
On a positive note, technology is booming to keep up with our world's evolving demands. For instance, internet users will grow to 5.3 billion by the end of 2023; 90 percent of big data has been generated within the last two years; data center spending on artificial intelligence (AI) technologies is expected to increase by 4x by 2025; the overall market for AI is predicted to reach $89.8 billion by 2025; and PwC deems AI to be a "$15.7 trillion game changer" based on predictions of its total economic impact on the global economy through 2030.
Enterprise spending on these technologies is ramping up as well, with 97.2 percent of companies committing to further invest in their AI and big data capabilities.
Supercomputing is a type of high-performance computing (HPC) that leverages a powerful computer to reduce the overall time to solution. HPC and AI are a force multiplier. Combining HPC with AI allows you to accomplish multitudes of tasks simultaneously. Instead of taking a million computing cores and doing one thing, HPC carves it into 10,000 cores here and 10,000 cores there. It's like multi-tasking on steroids.
Companies that must differentiate (i.e., all companies today) can use HPC and AI to drive competitive advantage. For example, companies can leverage HPC and AI to add disruptive capabilities and capabilities that promote the societal good. The value these technologies can bring to enterprises and humankind is as astounding as it is exponential. Their applications can help us better understand how the universe works, unveil discoveries in gene therapy, show us how to build more effective cancer drugs, reveal a roadmap to thwart future pandemics, etc.
Large language models (LLMs), generative AI, molecular modeling, physical simulations, aerodynamics, nuclear fusion and cryptanalysis, to name a few, all require the capability of HPC arranged to function — a marriage we are calling high-performance architecture (HPA). Applications of HPC and AI include critical support for climate modeling, advances in healthcare and life sciences, financial services, manufacturing, transportation, space travel and more.
In both the private and public sectors, today's business leaders are asking critical questions like: How do I run my business for optimized outcomes? How do we keep our borders secure? How can we improve our fuel economy? How many molecules or viral transports can I model in a day? How do I know the value of my oilfield? While the language is all business, it is IT's job — with the help of a third-party advisor like WWT — to translate these business objectives into the language of technology (i.e., HPC, AI and HPA).
AI speaks to business needs more effectively because the data it generates and collects helps solve real business challenges. Business people don't necessarily care about IT infrastructure's speeds and feeds. They want to know: Is our water or air cleaner? Do we have safer transportation? Did our time to market decrease significantly? As IT leaders, the more you can align your recommendations for big systems to the needs, goals and desired outcomes of the business, the better off you will be.
For example, let's consider the case of a major oil company that wants to apply supercomputing and AI to its business. The company wanted to predict with high confidence when and where a hurricane would hit the Gulf of Mexico. Why? Because its biggest oil refinery was sitting in the potential path of that hurricane.
According to the best estimates of current HPC models, using information gleaned from some of the planet's biggest supercomputers, the hurricane would impact a 40-mile swath. The problem was that this oil company needed to shut down its assets based on those predictions.
Typically, it only takes three days for a refinery shutdown. However, it can take 30 to 45 days to restart, meaning oil companies suffer a significant financial hit during downtime — not to mention the negative impact on market prices. If the oil company assumed a conservative approach, it would stop production. But if the hurricane was likely to miss the area, that would mean an unnecessary revenue decline. In this scenario, the value of supercomputing and AI is clear to the company's bottom line.
By tightly coupling HPC and AI, you can better ask and answer these types of questions.
Scientists typically measure computer performance in floating point operations per second (FLOPS). Since computers are quicker than people, their performance in FLOPS has so many zeros that researchers use prefixes. For example, "giga" is a number with nine zeros. A personal computer operates in the gigaflop range, at about 150,000,000,000 FLOPS or 150 gigaFLOPS.
"Tera" has 12 zeros. In 1996, computers hit the terascale milestone with the Department of Energy's Intel ASCI Red supercomputer. ASCI Red's peak performance was 1,340,000,000,000 FLOPS or 1.34 teraFLOPS.
Exascale computing is even faster. This type of ultra-powerful supercomputing performs billions of computations per second utilizing an infrastructure of HPA CPUs and GPUs to process and analyze data. "Exa" signifies 18 zeros, which means an exascale computer can perform more than 1,000,000,000,000,000,000 FLOPS or one exaFLOP — more than one million times faster than ASCI Red's peak performance. This is about 1,000 times faster than the current most powerful supercomputers in the world.
Exascale computing represents a significant achievement in computer engineering, as it enables more accurate and efficient simulations and analyses of complex phenomena in science, engineering, medicine and other domains. This area is a national priority for many countries, including the US, as it can enhance a nation's competitiveness and leadership in science and technology.
Once you've answered significant questions regarding current capabilities and goal-setting — where you want to be in two, five, 10 years and beyond — you will want to determine your metrics. Commonly reported metrics are processor speeds and FLOPS. You can use these numbers to estimate how fast your applications will run on different machines. (Of course, faster is always better if you get the right results!)
More sophisticated users realize that manufacturer specifications (such as theoretical peak floating-point issue rates) are rarely achieved in practice. Instead, they may use simple benchmarks to predict relative application performance on different machines. Others might feel that because their data or algorithm is so proprietary they prefer to keep it in-house to make their supercomputer.
Many organizations hire a third-party expert to guide their strategy development and to build their reference architecture. Once you have a reference architecture, you should consider manufacturers who can help you reach your desired future state as efficiently as possible. IT professionals will focus on network switches, servers, terabytes, etc. Business stakeholders will set their sights on making products or services more relevant. Their decisions must be aligned and driven from the top down to be effective. Successful leaders will also bring in key stakeholders early enough to support the success of their supercomputing/AI project. When done right, there is no end to what these powerful technologies can help you accomplish.
HPE helps empower world-changing innovation and discovery in the exascale era with faster time-to-results and accelerated AI. HPE can help organizations achieve monumental performance and unlimited transformation with the HPE Cray product line, which is purpose-built for supercomputing and AI workloads and powered by the latest generation processors from AMD and Intel — designed to provide unprecedented performance, high-bandwidth memory and built-in acceleration for the fastest-growing workloads. HPE's density-optimized systems offer a complete, scalable solution with power and cooling options that deliver superior performance at a lower total cost of ownership.
We will help you solve complex problems quickly and efficiently with high-performance architecture (HPA). As organizations like yours continue to explore and leverage AI/ML applications, the opportunities for HPC to transform and create business opportunities are limitless. From energy exploration to data analytics and health sciences, to self-driving automobiles and voice recognition, HPC is everywhere. And it is helping companies and products remain competitive in today's real-time, data-driven world.
At WWT, we work with the latest technologies and partners to better enable customers to leverage HPC solutions quickly. We can show you the art of possible in our Advanced Technology Center (ATC), where you can evaluate, design and realize the benefits of HPC solutions.