Why is this Fortran code so much faster than its C++ counterpart?

No. If you look at the example in Compiler Explorer, you will see:

MAIN__:
        push    rbp
        mov     esi, 1
        movabs  rbp, 24000000000
        push    rbx
        mov     rdi, rbp
        sub     rsp, 552
        call    calloc                    #call to system calloc
        test    rax, rax
        ...

In this particular case, the system calloc implementation did better than C++'s operator new + memset. But as others have shown, you cannot generalize this to all platforms (OS’s, architectures).