Yes there is a faster way, you can spare 1 cycle in a loop:

clr a
mov dpl,a
mov r2,a
mov r0,a


L1:
movx a,@dptr (2 cycles)
movx @r0,a (2 cycles)
inc r0 (1 cycle)
inc dpl (1 cycle)
djnz r2,L1 (2 cycles)

ret

your loop takes 8 cycles, but you can do it in another way:

L1:
movx a,@dptr (2 cycles)
movx @r0,a (2 cycles)
dec dlp (1 cycle)
djnz r0,L1 (2 cycles)

ret

You can test in, works perfectly OK.
This solution gives you 12.5% time save.
If you need a faster solution, take a device with dual data pointers and fast data move option (like some DALLAS devices have).

Franc