Virtual memory is

Concept. Virtual memory is a memory-management technique in which the operating system, assisted by the CPU's memory-management unit (MMU), gives each process a logical address space whose size is independent of the physical RAM installed. The pages a process is actively touching occupy physical frames, the remainder of its address space is kept on secondary storage, and the translation between logical and physical addresses is invisible to the running program.

Applying it here. Trace what the technique actually delivers to a running program:

1. The program issues addresses in its own logical address space; on every reference the MMU consults that process’s page table and translates the logical page number into a physical frame number.

2. If the referenced page is not resident, the hardware raises a page fault; the operating system fetches that page from the swap area into a free frame, evicting a victim page when no frame is free, and then restarts the faulting instruction.

3. At any instant only the working set therefore occupies RAM, while the address space the program may legally use is bounded by the address width and the size of the backing store, not by the DRAM fitted in the machine.

4. None of this bookkeeping is visible to the program, so the program behaves as though main memory were far larger than the machine physically has — an appearance produced by address translation and demand paging, not by extra hardware.

Cross-check — what each offered description names.

Hence virtual memory is an illusion of an extremely large main memory: the size a program perceives comes from address translation and demand paging, not from the DRAM actually installed. It adds no physical main memory, it is not itself a form of secondary memory, and it is a standard feature of ordinary operating systems rather than a special supercomputer technology.