For no other reason than because it's there, here is the result of asking Mathematica to factor RSA 100.
In 1991,
... the factorization took a few days using the multiple-polynomial quadratic sieve algorithm on a MasPar parallel computer.
and now,
It takes four hours to repeat this factorization using the program Msieve on a 2200 MHz Athlon 64 processor.
In comparison, Mathematica took little more than an hour. Unfortunately, it was not able to parallelize it (for a real comparison, the Msieve code would have to be ported over).
{% img center http://farm6.staticflickr.com/5510/12309071334575cda01f9z_d.jpg %}
Update: This has been much better documented on this StackExchange thread, which also links to notes on internal implementation, in case you're interested:
FactorInteger switches between trial division, Pollard , Pollard rho, elliptic curve, and quadratic sieve algorithms.
Finally, more on RSA Numbers, as well as a Mathematica Notebook, here