Core banking systems lie in the heart of any bank today as they embrace all services that are provided by banks to carry out their day-to-day business. Examples of these services can be customer management, transaction processing, payment processing, auditing, and statistical reporting. Software infrastructure in such core banking systems was built based on the monolith system architecture, which was a common practice in building distributed software systems earlier before the advent of cloud computing. Monolith architecture incorporates all core banking services into a single, fully integrated piece of software where the codebase can become overwhelmingly complex and incomprehensible over time. As a consequence, it becomes rather challenging to make bug fixes and feature additions in monolith core banking systems and serious maintenance costs can arise to adapt to new technologies. Microservices Architecture (MSA) has emerged in line with the developments in virtualization technologies and cloud computing (e.g., Kubernetes, Containerization) as a software engineering practice to overcome the problems that can be encountered in monolith software. In this paper, a system architecture for an MSA-based core banking system was proposed by particularly taking performance efficiency into account. The justification of this architecture was made through a real-world case study in a commercial bank having a large-scale legacy core banking monolith. Load tests were applied to both this legacy core banking monolith and the new MSA-based banking system that reflects our architectural proposal and average request throughputs and response times were analyzed in both systems to demonstrate how better the proposed architecture performs in comparison to its monolith counterpart concerning performance-efficiency.