The Evolution of the Mainframe Computer—to IBM’s Latest z16

April 13, 2022 by Arjun Nijhawan

With a history dating back to the 1940s, the mainframe computer has advanced significantly to today's mainframes, such as the IBM Z series.

It is no secret that computers have not always been as powerful, portable, and (relatively) inexpensive as they are today. Indeed, early computers often took up entire rooms and could cost millions of dollars. The 1940s electromechanical computer known as the Harvard Mark 1, for instance, was over 50-feet wide and 8-feet tall.


Harvard Mark 1

Harvard Mark 1, also called the IBM Automatic Sequence Controlled Calculator (ASCC), was the first mainframe computer ever invented. Image used courtesy of Britannica


Due to their vast sizes, such computers were historically referred to as mainframes because they were housed in large metal boxes or frames. Today, mainframes are a fraction of their historical size and can perform much more powerful computations—as IBM's latest z16 mainframe exemplifies. 


The Evolution of Mainframe Computers

The Mark 1 would represent a key moment in the history of modern technology. For the first time ever, scientists and engineers could perform basic numerical operations in a matter of seconds. A few years later, the ENIAC (Electronic Numerical Integrator and Computer) would herald a second breakthrough: the world’s first fully electronic and programmable mainframe computer.

While the Mark 1 was an electromechanical system, the ENIAC used vacuum-tube technology to store and transmit information. Vacuum tubes enabled engineers to create basic logic circuitry in the ENIAC. Programmability allowed the ENIAC to execute basic loop and branching conditions. 


Vacuum tubes used in early mainframe computers

Vacuum tubes were used in early mainframe computers. Image used courtesy of IBM


Still, vacuum tubes consumed a large amount of power. The invention of the transistor in the 1960s significantly reduced the form factor of the computer. Transistors enabled solid-state logic circuitry, and in 1981 IBM released what would become the world’s first commercially successful personal computer: the IBM PC. 

Even with the advent of the PC in the 1980s, mainframe computers have remained in use until the present day. Today, the word mainframe is used to distinguish the application of the computer as opposed to its physical form factor.

Today’s mainframes no longer occupy entire rooms or cost millions of dollars to acquire: the IBM z16 is about the size of a refrigerator. Instead, the word mainframe refers to those computers that typically have large amounts of memory and processors to handle millions of transactions but are also specifically designed with reliability, stability, and security in mind. 


The Mantra of Modern Mainframes: Reliability, Availability, and Serviceability

Two of the most important characteristics of computers today are processing power and available memory capacity. A laptop or desktop PC, for example, may be advertised as equipped with a 2.5 gigahertz quad-core processor with 16 gigabytes of RAM and 256 gigabytes of hard drive space. But mainframes are characterized by three additional traits: reliability, availability, and serviceability (RAS).


Transaction running through IBM Z and z/OS

Transaction running through IBM Z and z/OS. Image used courtesy of IBM 


While most consumers are familiar with the concept of reliability, modern mainframe computers must be able to continuously operate for years on end without needing repairs. Building upon this concept, a mainframe computer must also be available, meaning that it can perform automatic hardware and software recovery in case of failure without impacting the operation for users. Finally, a mainframe computer must be serviceable, indicating that it can determine exactly why a particular operational failure occurred.

Together, these three characteristics of mainframes make them an essential component of corporate and government backend infrastructure in our world today. For instance, due to their ability to handle massive transaction volumes with high availability and reliability, mainframes are a critical component of banking institutions


IBM z16 Targets AI and Transaction Workloads 

IBM, the predominant enterprise in the mainframe sector today, recently announced its latest z16 mainframe. The z16 itself has 200 cores based on the IBM Telum processor. A single IBM Telum chip has 8 processor cores (running at a 5-GHz clock frequency) that use a superscalar out-of-order pipelining strategy to maximize instruction throughput at the CPU level.

Importantly, the Telum chip also contains on-chip acceleration for AI processing via an “AI core.” The AI core is essentially a dedicated core on the Telum chip optimized for deep learning workloads. As the other CPU cores on Telum execute programs, the tightly-coupled AI core is able to run deep learning analysis on the data being sent throughout the system. In the case of banking applications, for example, IBM states that this could enable banks to detect fraudulent credit card transactions before they are even completed. 


IBM z16

The IBM z16 could help banks detect fraudulent transactions while they are ongoing. Image used courtesy of IBM 


Regardless of the application, it is clear that IBM mainframes will continue to target high-volume transactional workloads across industries. According to IBM, its z/OS operating system in conjunction with its mainframe hardware technology is optimized for handling large-scale transactional workloads using multitasking capabilities. In the years to come, this will likely play an even more critical role in mainframe technology.