A closer speed estimate:
The Atari's chip took about 4 clock cycles per instruction.
The Intel chip does 4 instructions per clock cycle. (so multiply by 16)
The byte size is 8 bit vs 64 bits. (multiply again by 8)
The Intel has a floating point subsystem. (multiply by 10? if you are doing math calculations.
Probably more like 40 if you were to have each do 32bit floats.)
The Intel has 2 cores. (multiply by 2)
The Intel has L1 L2 RAM cache. (probably factored in to the 4 instructions per clock timing)
So the actual speedup is more like 1000 x 16 x 8 x 10 x 2 = 2,560,000 for floating point
and 1000 x 16 x 8 x 2 = 256,000 for data manipulation.
Multiply by another 10 for the newer faster 8 core desktop machines.