Using words like speed and fast to describe contemporary computers, while accepted by precedent, is extreme oversimplification. There was a time when a programmer could read a program, calculate the sum of the instruction times, and confidently predict how long it would take the computer to run that program. Thousands of programmers and engineers have spent the last 30 years making such straightforward calculations impossible, or at least meaningless.
Today's computers are more powerful than their predecessors, not just because they use integrated circuits instead of vacuum tubes and have far shorter cycle times, but because of innumerable hardware and software architectural inventions. Each advance in integrated-circuit density brings an advance in computer performance, not just because it allows the same logic to work in a smaller space with a faster system clock, but because it gives engineers more space in which to implement ideas. In short, computers have gained capacity by becoming more complex as well as quicker.
The complexity of modern computers and their operating systems is matched by the complexity of the environment in which they operate. In addition to running individual programs, today's computer deals with varying numbers of unpredictably timed interrupts from I/O and communications devices. To the extent that the engineers' ideas were based on an assumption of a single program running in a standalone machine, they may be partly defeated by the randomness of the real world. To the extent that those ideas were intended to deal with randomness, they may win back some of the loss. The wins and losses change from program to program and from moment to moment.
The result of all these hardware and software wins and losses is the performance of the system. The speed of the system is the rate at which it can handle a specific sequence of demands. If the demands mesh well with the system's hardware and software architectures, we can say, "The system runs this workload fast." We cannot say, "The system is fast," or at least we should not.