Exascale computer supercomputer

An exascale computer is a supercomputer that is capable of performing at least one exaflop (10^18 floating point operations per second). Such a machine would be significantly more powerful than any currently existing supercomputer and would be used for a wide range of scientific and engineering applications.

Designing an exascale computer would involve a number of technical challenges, including:

Power consumption: Exascale computers would require enormous amounts of power to operate, likely on the order of tens of megawatts. This presents a significant obstacle to building such a machine, as it would be difficult to provide the necessary power while also keeping the system cool enough to prevent overheating.

Memory and storage: Exascale computers would need to have very large amounts of memory and storage in order to store and process the massive amounts of data they would be working with. This would likely require the development of new memory technologies, such as phase-change memory or resistive RAM.

Networking: Exascale computers would need to have extremely high-speed interconnects in order to allow the various components of the system to communicate with one another. This would likely require the development of new networking technologies, such as photonics or silicon photonics.

Software and programming: Exascale computers would require new programming models and software to be able to make use of their immense computational power. This would likely involve the development of new parallel programming languages and frameworks, as well as new algorithms and libraries optimized for exascale computing.

One possible design for an exascale computer would be a distributed system made up of many thousands of individual nodes, each containing multiple processors. The nodes would be connected together by high-speed interconnects, such as InfiniBand or Intel's Omnipath, to allow for fast communication between the processors.

To minimize power consumption, the processors themselves would likely be based on a low-power architecture, such as ARM or RISC-V. They would also be designed to be highly energy efficient, possibly using techniques such as dynamic voltage and frequency scaling to adjust the power consumption of the processors in real-time based on the workload.

To provide the large amounts of memory and storage needed for exascale computing, the system would likely make use of non-volatile memory technologies, such as phase-change memory or resistive RAM. These technologies have the potential to provide much higher storage densities and faster access times than traditional DRAM or flash memory.

The system would also include a large amount of storage, likely in the form of solid-state drives or other high-performance storage technologies.

To handle the massive amount of data that an exascale computer would be working with, the system would require a powerful data management system. This could include a distributed file system, such as GPFS or Lustre, as well as a distributed database system, such as Hadoop or Apache Cassandra, to handle the massive amounts of data.

Finally, the system would require a powerful software stack, including parallel programming languages and frameworks, optimized algorithms, and libraries, to make use of the exascale computer's immense computational power.

Overall, an exascale computer would be an incredibly powerful and complex machine, requiring breakthroughs in a wide range of areas including hardware, power management, memory, storage, networking and software. Despite the challenges, the potential benefits of an exascale computer are huge, including the ability to solve some of the most complex and important scientific and engineering problems.

The total investment for exascale computers varies depending on the specific project and the country or organization developing it. However, it is expected to be in the billions of dollars range.

For example, the U.S. Department of Energy has announced a goal of developing an exascale computer by the end of the decade, and has allocated $258 million in funding for fiscal year 2021, with plans to invest $1.8 billion over five years for exascale computing research and development.

Similarly, the European Union has announced plans to develop an exascale computer by the end of the decade, and has allocated €1 billion ($1.2 billion) in funding for the project through its Horizon 2020 research and innovation program.

China has also announced plans to develop an exascale computer, and has allocated significant funding for the project. In 2020, China's first exascale supercomputer, the Sunway TaihuLight, was revealed to have costed $290 Million.

It's worth noting that these are just examples and the actual cost of building an exascale computer may be different and subject to change.