The best known error term in PNT is to my knowledge Big-Oh of
$$x\exp \left(-C \frac{ (\log x)^{\frac35}}{(\log \log x)^{\frac15}} \right),$$ and for $C$ one can take $-0.2098$, given in:

K. Ford. Vinogradov’s integral and bounds for the Riemann zeta
function. Proc. London Math. Soc., 85(3):565–633, 2002.

Also see "Updating the error term in the prime number theorem" by
Tim Trudgian http://arxiv.org/abs/1401.2689, especially note the footnote on page 4.

Thus there was some progress yet the factor $\frac35$, already known from Vinogradov-Korobov, was not improved up to now.

On the elementary proof: "Substantially different" is always hard to answer, but meanwhile there are elementary proofs that give an error term of basically the same quality as Vinogradov-Korobov, which was not at all the case for the original elementary proof. Thus, there was quite some progress there.

For an overview of elementary proofs you could consult this earlier MO question: Prime Number Theorem w/o Complex Analysis

In particular note that not all proofs are based on Selberg's formula. For example as mentioned in a comment by Voloch the one by Daboussi is not, as for example explained in its MR review by Diamond: "This paper gives an elementary proof of the PNT that is remarkable in that it makes no use of Selberg's now famous formula."