The recent decision by the Compute Express Link (CXL) and Gen-Z consortiums to collaborate on their respective technologies is a positive development that is likely to accelerate the acceptance of both memory fabric standards. Although not strictly competing against one another, prior to the agreement there was no assurance that Gen-Z and CXL supported devices would be compatible on future hardware. With the Memorandum of Understanding (MOU) recently agreed to by the two organizations, a path forward for interoperability and other collaborations now exists.

This Hyperion Research study presents our updated five-year forecast (2019-2024) for HPC on-premise server systems. This forecast includes the potential impacts to the HPC market from the covid-19 virus' effect on the world’s economy. Overall, the forecast has been reduced dramatically for 2020, with a partial recovery starting in 2021 and impacts lasting over the entire period.

While SSDs were initially integrated almost exclusively into the largest and highest performing HPC systems, widespread HPC adoption of flash-based storage is now occurring due to the perfect storm evolution and intersection of requirements, technologies and economics.

Nvidia’s recently announced Ampere GPU for datacenters (A100) comes to the market at a time of increased market competition for HPC and AI silicon. However, rather than offering specialized datacenter products aimed at these two application categories, the new Ampere offering carries forward the dual HPC/AI approach previously introduced in Nvidia's Volta architecture. To realize this, the company has introduced a number of innovations that significantly boost performance in each area.

This Special Study looks at the evolution and changes in file systems used for technical computing. Although the file system sector may appear to be mature, significant changes are occurring across many fronts spanning business, technology, workload requirements, and competitive forces.

Researchers at the Tokyo Institute of Technological (Tokyo Tech) with assistance from colleagues at Hitachi, Hokkaido College, and the University of Tokyo recently announced a promising new processor architecture that could advance the performance and capability of quantum annealing systems. Known as STATICA (Stochastic Cellular Automata Annealer Architecture), the novel architecture seeks to enable more parallelism in both connectivity and computation compared with counterpart quantum annealing devices. The processor is designed to take on complex optimization problems typically found in important applications including portfolio, logistic, and traffic flow analysis workloads that often cannot be effectively computed on counterpart classical systems.

Hybrid quantum-classical algorithms show significant promise in effectively harnessing the computational capabilities of near-term quantum systems when combined with classical HPC systems on key applications. Two recent developments reveal a growing interest in advancing such algorithms: the first, a trio of European-based organizations partnering to build a complete quantum computing R&D ecosystem: the second, a commercial partnership between quantum annealing vendor D-Wave and the Japanese supercomputer vendor, NEC. Successful growth of hybrid capabilities will require a merging of quantum and related HPC skills and can only contribute to the overall development pace of the quantum computing sector writ large. Additional partnerships with similar intentions will likely emerge soon.

The US White House recently released its proposed FY 2021 budget to Congress that seeks to impose a number of budget cuts on federal agencies, some of which are the largest purchasers and users of HPCs in the world. Notable reductions are proposed for NOAA (-31%), DOE's Office of Science (-17%), and NASA (-11%).

• On the plus side, two new advanced computing technologies, artificial intelligence (AI) and quantum information science (QIS) are earmarked for near doublings in budget and are spread across a number of US government agencies.
• As has happened in the recent past, Congress will almost certainly increase funding levels for some critical R&D, but competing financial priorities driven by the covid-19 economic fallout could limit the ability of Congress to adjust funding levels for activities not directly linked to near-term covid-19 policies and programs.

The recent announcement of a novel neural network processor by Silicon Valley startup Groq reflects the diversification taking place in hardware accelerators being developed for the artificial intelligence (AI) market. Groq’s Tensor Streaming Processor (TSP) uses a simplified architecture to enforce deterministic behavior, which offers a unique attribute for machine learning inference in both datacenter applications and for those at the edge.

By most objective measures, the United States is the leader in AI capability, both in its AI technology and its application. Whether looking at R&D spending, research capability, ecosystem support, or commercial deployment, the US is at the top or near the top of its global competition. That strength is derived largely from successes in the US private sector. In particular, computer companies in Silicon Valley and elsewhere have been at the forefront in developing new AI hardware and software technologies. A receptive user environment in areas such as web services, financial services, healthcare, and scientific research, among others, has created a virtuous dynamic between supply and demand for AI technologies and products. As a result, the depth and breadth of AI capability in the US is such that the country is well positioned to maintain its lead position in the near-term.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place in Chicago, Illinois. This update summarizes a presentation from that meeting entitled, Blue Waters: Sustained Petascale Computing given by Dr. Brett Bode from the National Center for Supercomputing Applications at the University of Illinois.

The quantum information science and technology (QIST) industry is facing critical challenges to its continued advancement due to recent financial difficulties brought about by the on-going covid-19 crisis. A recent survey conducted by QED-C, a US government industry-driven consortium, found that almost one-third of responding US QIST companies could be forced to take drastic financial measures within three months if the economic impact of the pandemic continues at pace. This sector, which has yet to reach maturation as a self-staining revenue generator, currently relies on a wide range of funding from both government and commercial sources. Some respondents cited the need for timely and aggressive government support to keep the nascent QIST industry on its current development.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at Argonne National Laboratory in Illinois. This update summarizes a presentation from that meeting entitled, Argonne Training Program on Extreme - Scale Computing, given by Ray Loy, ATPESC 2020 Program Director, Argonne National Laboratory.

The US Department of Energy recently initiated a program to establish as many as ten Quantum Information Science (QIS) research centers over the next five years and is currently accepting applications for the first round of two to five centers. Each center will have a specific focus area among highlighted quantum technologies including computing, communication, and sensing.

Over the last three years, AMD has become a force to be reckoned with in the high performance computing market. The recent announcements by AMD that its future AMD EPYC™ CPUs and AMD Radeon Instinct™ GPUs will be used to power the El Capitan supercomputer at Lawrence Livermore National Laboratory (LLNL) in 2023 and the Frontier supercomputer at Oak Ridge National Laboratory (ORNL) in 2021, demonstrate that the chipmaker may be able to effectively compete against other vendors for inclusion in some of the most powerful HPC systems in the world.

A review of the number of quantum communications (QComm) research papers published over the last five years shows that QComm basic science as well as R&D efforts are centered within China, with the US a distant second and perhaps falling further behind. Most other nations are publishing a steady, but smaller stream of QComm-related publications well behind what China is producing.

Collaboration has long been a hallmark of the global HPC community. From joint development of new standards and architectures to advancing innovation in next generation technologies to investing in cutting edge research across academia, all facets of the HPC ecosystem have benefitted from some aspect of the HPC community banding together for the common good.

HPC in the cloud has undergone significant growth in the past few years, attracting new users and more fully supporting existing users. Cloud Service Providers (CSPs) are key to this growth as they are taking a stronger interest in the HPC market, focusing on delivering capabilities that more closely match on-premise systems in performance, security, and in some cases, price. This major shift in the sector has prompted Hyperion Research to revise its HPC cloud forecast to include an additional $1 billion in 2019, up from its original $3 billion.

China has emerged as the strongest regional contender to overtake the US in AI capability. The government has devised a strategy to attain global AI leadership by 2030, leveraging the country’s growing technological prowess, its enormous domestic market, and its position as the world’s second largest economy.

In February 2020, Intel announced that it will center its AI acceleration roadmap for the data center on Habana technology, abandoning the Nervana development that has driven Intel's AI accelerator strategy for the last four years. The decision was precipitated by Intel’s $2 billion acquisition of Habana Labs, an Israeli startup specializing in AI processors for the data center, in December 2019.

A review of quantum computing (QC) research papers published over the last five years shows that QC basic science as well as R&D efforts are spread around the world across a number of leading research centers and are primarily driven by a regionally diverse range of government-sponsored funding agencies.

The contours of Arm-powered high performance computing are still being shaped, but the upcoming upgrade of its Isambard infrastructure by the UK’s Great Western 4 alliance suggests that architectural diversity will be a feature of Arm-based HPC going forward. Hyperion Research believes such diversity will be an important aspect to Arm adoption in the HPC market in the coming years.

In 2018, the European Union (EU) officially entered the competition for global leadership in artificial intelligence, outlining its strategy under the Coordinated Plan on Artificial Intelligence. Like China and the United States, the EU sees the technology as a major driver for regional economic growth and productivity, as well as a way to solve the world’s most pressing environmental and social challenges.

National governments around the world are moving to enact new quantum computing research activities or expand the scope and ambitions of existing programs, driven by a range of both national security and industrial competitiveness imperatives. The bulk of current and planned programs center on funding basic and advanced research in quantum technologies, primarily at national laboratories or related academic facilities.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at Argonne National Laboratory in Illinois. This update summarizes a presentation from that meeting entitled Enabling Performance Portable Climate Simulations on Aurora and Frontier by Nichols A. Romero from Argonne National Laboratory.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at Argonne National Laboratory in Illinois. This update summarizes a presentation from that meeting entitled HPC in Industry, given by Brendan McGinty, NCSA Director.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at Argonne National Laboratory in Illinois. This update summarizes a presentation from that meeting entitled ECP Application Development, given by Andrew Siegel from Argonne National Laboratory.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at Argonne National Laboratory in Illinois. This update summarizes a presentation from that meeting entitled NAVGEM on the Cloud: Computational Evaluation of Commercial Cloud HPC with a Global Atmospheric Model, given by Daniel Arevalo from Devine Consulting.

An unusual event for a technical conference was held at the September Quantum Tech Congress in Boston: a panel discussion among various venture capitalist on their view of the industry, the investment outlook for the future, why they are interested in quantum computing, and advice for start-ups.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In October 2019, the 74th HPC User Forum took place in Edinburgh, Scotland. This update summarizes a presentation from that meeting, entitled A rm in the UK: Catalyst and Isambard, given by Simon Burbidge from the University of Bristol.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at the Argonne National Laboratory, In Lemont, Illinois. This update summarizes a presentation from that meeting entitled, HPC and Data Down Under: The Pawsey Supercomputing Centre, Square Kilometre Array and HPC in Australia, given by Mark Stickells, Executive Director of The Pawsey Supercomputing Centre.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at the Argonne National Laboratory, In Lemont, Illinois. This update summarizes a presentation from that meeting in the session entitled (Un)obscured by Clouds -- Applying Cloud Techniques to High Performance Systems Integration, by Arno Kolster from Providentia Worldwide.

The high performance computing (HPC) sector is in the early stages of a major reinvention, fueled by a confluence of innovation in HPC hardware and system architectures, changing algorithms and application bases, and new access models.

In 2018, Hyperion Research officially released its first market sizing and forecast for HPC spending in third-party clouds, projecting a spend of $7.4 billion in 2023 on HPC cloud resources. After a year of in- depth studies, interviews, and research, the forecast incorporates new findings that reflect a significant shift in the market, especially in the past twelve months, increasing our 2019 cloud spending forecast by $1 billion.

Plans to upgrade the MareNostrum 4 supercomputer housed in the Torre Girona chapel at the Barcelona Supercomputing Center (BCS) have been public since the middle of this year. This much had been disclosed about the planned MareNostrum 5 successor supercomputer:

At the September 2019 Quantum.Tech Congress in Boston, Dr. John Randall of Zyvex presented a contrarian view of how the industry can solve the quantum computing manufacturing challenge.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2019, the 73rd HPC User Forum took place at the Argonne National Laboratory, In Lemont, Illinois. This update summarizes a presentation from that meeting entitled, Cancer Computing, Computing for the Cure, given by Roy Chartier, Founder and CTO of Cancer Computer.

Throughout the year, Hyperion Research conducts surveys to ask buyers about their near-term expectations for spending on HPC. We ask whether in the next quarter the amount they expect to spend will increase, decrease or stay the same. If they expect spending to go up or down in the next quarter, we ask by what percentage.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In April 2019, the 72nd HPC User Forum took place in Santa Fe, New Mexico. This update summarizes a presentation from that meeting entitled, Safety, Reproducibility, Performance: Accelerating Cancer Drug Discovery with ML and HPCT technologies, given by Amanda J. Minnich, Staff Research Scientist, Lawrence Livermore National Laboratory.

The UK's National Quantum Technologies Programme recently entered its second phase of funding, part of which will be a £94 million investment in its four Quantum Technologies Research Hubs tasked with research on quantum sensors, computing, imaging, and communications. The Programme is an integral part of the UK's ultimate vision of creating a viable indigenous commercial capability in quantum-enabled products bound for the defense and security, oil and gas, finance, and healthcare sectors, to name a few. Close coordination and cooperation with key industrial partners will be critical to the overall success of this effort.

Recent announcements highlight the growing interest in classical digital computers designed as quantum simulators to support near-term quantum algorithm and applications development. Currently true quantum computers are limited by issues of noisy circuits, limited qubit counts and interconnect schemes, and short circuit depth that can constrain the ability of researchers to develop and verify new quantum algorithms.

Propelled by a vision of HPC-cloud and HPC-enterprise convergence, Cray's Shasta architecture will underpin America's first three exascale supercomputers, including the just announced 1.5-exaflops El Capitan system headed for DOE NNSA's Lawrence Livermore National Laboratory in 2022.

The European Union Quantum Technologies Flagship, a ten-year 1 billion euro effort, tasked with fostering EU-wide capabilities in quantum technologies, recently released the first draft of its Strategic Research Agenda (SRA) that outlines the comprehensive set of technologies and applications the program hopes to address in the next decade. The SRA covers a broad range of quantum-related technologies but also addresses wider goals that include building viable quantum technologies infrastructures and supply chains, fostering a trained and educated quantum workforce and user base, and supporting the development of new quantum-based products and services.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In April 2019, the 72nd HPC User Forum took place in Santa Fe, New Mexico. This update summarizes a presentation from that meeting entitled Education & Workforce Development by Melyssa Fratkin from the Texas Advanced Computing Center (TACC).

Countries around the world are developing plans for the next generation of large supercomputers, with investments that exceed $300 million per system in many cases. This Quick Take provides Hyperion Research's estimate of schedules of installations and prices of accepted near-exascale and exascale supercomputers around the world.

This Hyperion Research study presents our latest five-year forecast (2018-2023) for HPC server systems. Worldwide revenue for the HPC technical server market grew 15.6% from 2017 to 2018 to a record $13.7 billion. Hyperion Research predicts the HPC technical server market will grow at a 7.8% CAGR between 2018 and 2023 to reach $19.9 billion by 2023.

Hyperion Research forecasts that the worldwide HPC server application workloads, representing the split-out of the HPC server market by application segments, will expand at a 7.8% CAGR to reach $19.9 billion in 2023, up from $13.7 billion in 2018 (see Figure 1). The categories in Figure 1 include high performance data analysis (HPDA) sub-segments: fraud and anomaly detection, affinity marketing, business intelligence, and precision medicine. Hyperion Research provides a separate detailed forecast on these sub-segments.

Hyperion Research forecasts that the worldwide HPC broader market (servers, storage, software, and service) will expand at a 7.1% CAGR to more than $39.2 billion in 2023, up from $27.8 billion in 2018.

Hyperion Research recently closed the books on the 2018 year for the high performance computing market in our market tracking database, the QView, as well as the Country Level Database, which tracks the HPC markets in 26 countries around the world as well as 13 different verticals. 2018 was another high growth year, with the server market reaching $13.7 billion, and the total market (servers, storage, software and technical support service) totaling more than $27.6 billion. The market has continued to grow at a steady rate of around 6% for the past 5 years.

Hyperion Research recently closed the books on the 2018 year for the high performance computing market in our market tracking database, the QView. 2018 was another high growth year, with the server market reaching $13.7 billion, and the total market (servers, storage, software and technical support service) totaling more than $27.6 billion. The market has continued to grow at a steady rate of around 6% for the past 5 years.

The Department of Energy at Oak Ridge National Laboratory (ORNL) recently announced plans for the development of a 1.5 exaflops system called Frontier to be delivered in 2021. US HPC maker Cray and chip maker AMD are the two key US commercial partners in this effort. Despite numerous press articles centered on the 1.5 exaflops peak performance of Frontier, ORNL's original RFP released in April of 2018 clearly called out the diverse workload requirements that Frontier would have to successfully handle that span the traditional modeling and simulation sector, big data analysis, and AI applications, while demonstrating a 50X improvement in solving key DOE science problems that today run at the 20 petaflops level. To meet those ambitious goals, strong support from DOE's companion $1.7 billion Exascale Computing Project (ECP) will be critical.

Hyperion Research recently closed the books on the 2018 year for the high performance computing market in our market tracking database, the QView. 2018 was another high growth year, with the server market reaching $13.7 billion, and the total market (servers, storage, software and technical support services) totaling more than $27.6 billion. The market has continued to grow at a steady rate of around 6% for the past 5 years.

Hyperion Research recently closed the books on the 2018 year for the high performance computing market in our market tracking database, the QView. 2018 was another high growth year, with the server market reaching $13.7 billion, and the total market (servers, storage, software and technical support service) totaling more than $27.6 billion. The market has continued to grow at a steady rate of around 6% for the past 5 years.

This Quick Take presents the new Hyperion Research forecast compared to the previous (May 2018) five-year HPC forecast. The HPC market grew faster than expected in 2018 at 15.6% growth over 2017, resulting in sales of $14.5 billion for the year.

Hyperion Research recently closed the books on the 2018 year for the high performance computing market in our market tracking database, the QView. 2018 was another high growth year, with the server market reaching $13.7 billion, and the total market (servers, storage, software and technical support services) totaling more than $27.6 billion. The market has continued to grow at a steady rate of around 6% for the past 5 years.

Hyperion Research recently closed the books on the 2018 year for the high performance computing market in our market tracking database, the QView, as well as the Country Level Database, which tracks the HPC markets in 26 countries around the world as well as 13 different verticals. 2018 was another high growth year, with the server market reaching $13.7 billion, and the total market (servers, storage, software and technical support service) totaling more than $27.6 billion. The market has continued to grow at a steady rate of around 6% for the past 5 years.

Worldwide factory revenue for the high-performance computing (HPC) technical server market jumped for the full-year 2018 by 15.6% over the exceptionally strong 2017 calendar year period to $13.7 billion. The 2018 result was 6.2% higher than Hyperion Research's forecast $12.9 billion figure for the year.

The Gen-Z Consortium, launched in October 2016, brought together multiple technology vendors to create a new open system interconnect standard to work with the growing number of new processors and memory options available on the market now, as well as in the future:

IBM's recent announcement of their new Q System One universal quantum computer is yet another milestone in the firm's long-term commitment to transitioning quantum computing (QC) hardware from one-off research status into a capable commercial offering. Although the new Q System One, as announced, does not demonstrate any significant advances in current quantum computing capability, as measured by the number of qubits per system, it does show that IBM can design and manufacture a system that, at the right price, could be attractive to a wide range of users looking to integrate quantum computing into their overall R&D process.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users.In September 2018, Hyperion Research hosted its 70thUser Forum in Dearborn,Michigan. This update summarizes a presentation from that meeting in the session, Leadership Computing Initiatives, entitled The European Processor Initiative, given by Jean-Marc Denis, the Chair of the Board, European Processor Initiative.

IBM's decision to acquire Red Hat not only boosts the combined firm's potential to exploit the overall cloud market, but it will likely have a significant impact on the global technical computing market as well. This is based on the rapidly growing number of enterprise IT data centers that are integrating HPC into their AI and other data-intensive workloads that their enterprise servers cannot effectively handle alone.

This inaugural, two-day AI Hardware summit held in Mountain View, California, at the Computer History Museum, brought together researchers, vendors, and users to explore the development of the AI ecosystem from a hardware perspective. Large companies, startups, and analysts joined to hear over 30 speakers and roundtables. Although the overall theme was all AI hardware, many of the presentations focused on AI processors and the work that is being done to design hardware to further the development of the evolving and growing field of AI, machine learning, and deep learning.

The HPC User Forum was established in 1999 to promote the health of the global HPC industry and address issues of common concern to users. In September 2018, Hyperion Research hosted its 70th User Forum in Dearborn, Michigan. This update summarizes a presentation from that meeting given in the session, Leadership Computing Initiatives, entitled EuroHPC and the European HPC Strategy presented by Leonardo Flores of the European Commission.

Over the past decade, the relative positions of HPC vendors in terms of revenue have shifted and evolved due to acquisitions, changes in product lines, and advances in technology. Hyperion Research sees this pattern continuing as we project out over the coming years as new areas of technology emerge and current technology evolves.

Hyperion Research forecasts that the worldwide HPC server-based AI market will expand at a robust 27.1% CAGR from 2017-2022 to reach $1.6 billion in 2022, up more than four-and-a-half-fold from $346 million in 2016, the first year in which we tracked this figure. We define the HPC AI market as a subset of the high-performance data analysis (HPDA) market that includes machine learning (ML), deep learning (DL) and other AI workloads running on HPC servers.

In April 2018, researchers from Goethe University Frankfurt and the Frankfurt Institute for Advanced Studies (FIAS) reported that they had accomplished the first successful three- dimensional simulations of the effects of colliding black holes carried out by using a first-order formulation of the Einstein equations.

This Quick Take looks at the growing availability of commercial cloud-based quantum computing (QC), supported through either direct access to true QC systems or QC simulators based on traditional digital hardware. In addition to QC hardware, most cloud-based QC providers are also rolling out their own software development environments to help existing and new QC software developers more effectively explore QC programming. The choice of options for potential QC users is growing, and each of the QC providers offers a distinct take on QC architecture and programming.

In August 2017, Hewlett Packard Enterprise (HPE) and NASA embarked on a daring experiment –- blast a supercomputer into space, place it aboard the International Space Station, and see how it functioned amid shifting radiation levels and violent G-forces. Since that time, the project known as Spaceborne has consistently delivered one trillion calculations per second, revolved around the earth at least 4,275 times, and has been described by HPE as a great success that has “exceeded expectations.”

Hyperion Research forecasts that the worldwide HPC broader market (servers, storage, software, and service) will expand at a 9.6% CAGR to more than $38.3 billion in 2022, up from $24.2 billion in 2017.

This report looks at the European approach to developing both exascale computers and an indigenous, self-sufficient HPC supply chain in Europe. In recent years the EU and its Member States, along with countries including China, Japan, and the United States, have increasingly recognized the strategic importance of investing more heavily in supercomputing to accelerate scientific and industrial innovation.

First quarter 2018 HPC server revenues were strong, with notable high quarter over quarter growth from the first quarter of last year. Hyperion Research has been analyzing the data to ascertain what these numbers mean and what they may indicate about full-year 2018 in HPC. These assessments incorporate Hyperion Research discussions with a global base of people in the HPC sector as well as an analysis of the Hyperion Research QView database, a long-standing collection of quarterly revenues and projections for the major HPC vendors.

Hyperion Research forecasts that the worldwide HPC server market will grow at a 9.7% CAGR to reach more than $19.5 billion in 2022, up from $12.2 billion in 2017 (see Figure 1). In 2022, the North American HPC server market will be, as it is today, the largest single market with 40% of global share, but the over 15% CAGR between 2017 and 2022 for the Asia/Pacific region and Japan will be the largest for all regions tracked.

At SC17 in Denver, the Japanese RIKEN K computer emerged for the third straight time as the world's most powerful supercomputer based on the High Performance Conjugate Gradient (HPCG) benchmark list. Although China's Tianhe-2 supercomputer has been widely seen as number one in the world based on its LINPAC rating, the HPCG test that the K computer excelled on may be more representative of the range of real-world HPC problems encounter by users. Riken's K computer has been either number one or two since the HPCG list came out in 2014.

This Hyperion Research study presents our latest five-year forecast (2017-2022) for HPC server systems. Worldwide revenue for the HPC technical server market grew 5.8% from 2016 to 2017 to a record $12.3 billion. Hyperion Research predicts the HPC technical server market will grow at a 9.8% CAGR between 2017 and 2022 to reach $19.5 billion by 2022.

DOE's recent announcement of the operational Summit supercomputer at Oak Ridge National Laboratory (ORNL) marks a shift in the global lead for supercomputer performance from China back to the United States.

Understanding physical behaviors in harshenvironments is extremely hard, and at times,critically important. That is the situation GeneralElectric (GE) engineers faced when designing theirnew heavy-duty gas turbine generator. Gaining even 1% greater efficiency could save their electric utility customers millions of dollars andimprove GE’s world-wide competitiveness. Thekey issues involved a better understanding of the fluid dynamics and reactive flow behaviors in the turbine’s 1,500 degree Celsius combustion chambers and the interactions between the 2 to 16 individual flames in the chambers. GE Power engineers had reached the limits of theory and experiments, motivating them to apply advanced modeling and simulation.

Hyperion Research forecasts that the worldwide HPC server application workloads, representing the split-out of the HPC server market by application segments, will expand at a 9.8% CAGR to more than $19.5 billion in 2022, up from $12.2 billion in 2017 (see Figure 1). The categories in Figure 1 include high performance data analysis (HPDA) sub-segments: fraud and anomaly detection, affinity marketing, business intelligence, and precision medicine. Hyperion Research provides a separate detailed forecast on these sub-segments.

China is a leading contender to stand up the first exascale systems in the world with a planned availability in the 2020-2021 time frame. To meet that aggressive goal, Chinese HPC developers are focusing on the use of indigenous processors to power these systems. There are a number of indigenous processor developments in China destined for the HPC sector that span a wide range of processor types: some are wholly unique to a single Chinese developer, while others are based on existing schemes that include ARM and even an x86 variant.

This Quick Take looks at the recent announcement by the Department of Energy (DOE) of the release of a Request for Proposal (RFP) for exascale systems to be installed at Oak Ridge National Laboratory (ORNL) and Lawrence Livermore National Laboratory (LLNL) in the 2021-2022 time frame. The RFP also says that a third system (different from the ORNL computer) may be installed at Argonne National Laboratory (ANL) in 2022.

Throughout the year, Hyperion Research uses our surveys to ask buyers about near-term expectations for their spending on HPC. We ask whether, in the next quarter, they expect the amount they spend to increase, decrease or stay the same. If they expect spending to go up or down in the next quarter, we ask by what percentage. The questions and resulting data are not limited to servers but also include anticipated spending on cloud usage and on desktop computers. We then use this data collected over multiple quarters, along with historical revenue patterns and other sources, to help us generate our HPC market forecasts.

On April 16, 2018, Hewlett Packard Enterprise (HPE) in conjunction with SUSE and Arm teamed up with three leading UK universities to form the Catalyst UK program with the objective of accelerating the adoption of HPC by industry in the UK. The three-year program will center on the installation of an Arm processor-based supercomputer deployment spread across the University of Edinburgh, the University of Bristol, and the University of Leicester, due to be completed in summer 2018. In addition, the program is looking to reach out to UK industry to jointly develop Arm-based applications and workflows. Hyperion Research believes this effort is an important indicator of the UK’s commitment to building an indigenous HPC ecosystem centered on Arm processors, and one that could have wider implications for bolstering Arm-based HPC adoption in future European HPC designs.

The x86-64 processor base has solidified its overwhelming lead as the processor of choice for the HPC sites surveyed. ▪Within the last two years. the number of RISC processors identified at respondent sites has declined by over 40%, while the reported use of vector processors has essentially disappeared. ▪It is worth noting the appearance of a new albeit small presence of custom processors in the survey, reflecting some of the new custom chips being used in experimental HPCs around the world, such as those in leading-edge Chinese-built systems.

Hyperion Research covers the worldwide market for deep learning (DL) and machine learning (ML) that is challenging enough to require the use of HPC server systems and other HPC resources (we do not cover the mainstream DL/ML market that does not use HPC, e.g. social media and large internet providers).

This document highlights findings from surveys taken by HPC users across industry, academia, and government institutions. The questions utilized for this report center on market growth rates for three areas: HPC servers, technical computing desktops, and cloud usage for technical computing applications. This information is aggregated and used in our market projections, and it helps us better see where the market is shifting. The time period for this set of surveys is from April 2015 to June 2017.

This Quick Take looks at Amazon's recent announcement of a new EC2 Bare Metal offering that provides users with direct, non-virtualized access to a processor, memory, storage and related networking instance. Amazon uses custom hardware in an effort to wring out the highest possible performance from the basic hardware set up while still offering a full complement of cloud-based software support. This development is a positive one for HPC users looking to migrate workloads to public clouds as Amazon is addressing one of the most vexinghurdles of HPC in the cloud: the performance overhead of running applicationsin a virtualized, performance limiting environment.

Taiwan-based TSMC, the world's largest chip foundry, recently announced (https://www.eetimes.com/document.asp?doc_id=1332869) that high performance computing (HPC) has supplanted smart phones as the most important driver of its business, although presumably not yet the largest financially.

Major government buyers we interviewed consistently named Cray as one of the handful of global supercomputer makers capable of true innovation at the system architecture level. These vendors maintain their ability to stay at the forefront of the HPC market by continually reinventing themselves and their products. Cray's capture of the $600 million CORAL 2 contract to deliver the 1.5-exaflops Frontier supercomputer to ORNL in 2021 is the company's latest demonstration of its ability to mutate in response to the evolving HPC market environment.