Operations per second is a key metric used to measure the computational performance of processors, systems, and devices, particularly in the fields of supercomputing, high-performance computing (HPC), and artificial intelligence (AI). It refers to the number of individual computations a system can perform in one second. Commonly expressed in terms like FLOPS (floating-point operations per second) or IOPS (input/output operations per second), this measure helps assess the speed and efficiency of hardware systems. For example, supercomputers like the Fugaku, operational since 2020, achieve performance in the exaFLOPS range, demonstrating the massive scale of modern computational capabilities. https://en.wikipedia.org/wiki/FLOPS

Different computational systems specialize in specific types of operations per second. GPUs often excel at floating-point operations, making them ideal for tasks like rendering graphics and training deep learning models, while CPUs are optimized for general-purpose tasks. Metrics like IOPS are more relevant for storage systems, reflecting the number of read/write operations a system can handle. These distinctions make operations per second a versatile benchmark across diverse technological applications, from scientific simulations to database management. https://www.intel.com/content/www/us/en/architecture-and-technology/floating-point-operations-explained.html

The evolution of operations per second metrics reflects the growing demands of computational workloads. With the advent of machine learning, big data analytics, and real-time applications, hardware manufacturers such as NVIDIA, Intel, and AMD have developed systems capable of trillions of operations per second. For instance, the NVIDIA DGX A100 platform, introduced in 2020, provides petaflop-scale performance for AI applications. This continuous scaling highlights the critical role of operations per second in shaping the future of computing technology. https://www.nvidia.com/en-us/data-center/dgx-a100/