To extend DStu's idea into hyperdrive, let's do this.

First, to define a variable: The number of cards in each stack is N.

Start of turn, you have 2x*Village, Watchtower, 2xCouncil Room. On top of your deck, lots of Highways, and the other cards you need.

After playing Vil/Vil/CR/CR, you play 11 total Highways, and you have 11 cards in hand:

X, Watchtower, TR, TR, KC, KC, IW, IW, Workshop, CR, Goons. The X is completely irrelevant.

TR|TR|KC|KC allows you to triple a total of 6 action cards.

Tripling IW/IW/WS allows you to gain 9 cards (the 9 engine cards in the above hand) and gives you 6 bonus actions, of which you'll need just one.

Watchtower allows you to put all 9 gains on the deck.

Triple CR draws all 9 and gives 3 Buys

Triple Goons gives 3 Buys.

Your sixth tripling goes unused; this engine won't support it.

You'll be able to do this N/2 - 1 times before the KC, TR, and IW piles run out. The -1 is effectively irrelevant here, so I'll drop it in the following equations. After all this is done, you have available:

More actions than you'll ever use.

3N Buys (6 * N/2).

N/2 copies of Goons in front of you.

Those 3N buys can empty any three Green piles you want other than Vineyard. Your final deck will contain 4.5N (plus an insignificant constant) action cards and 3N Green cards. Gardens and Silk Road will both be worth (3N/4) points each, so take your pick. Since Duke increases as the square, that's an obvious one. So:

Empty Duchy pile for 3N points.

Empty Duke pile for N^2 points.

Empty Gardens or SR pile for (3/4)N^2 points.

Each buy earns N/2 points from Goons, for a total of (3/2)N^2 points.

Total: (13/4)N^2 + 3N points. For N = 1E42, the second term isn't even within significance. Your total score is within <0.0000001% of 3.25E84.

TR, KC, IW, CR, Duchy, Duke, and Gardens/SR are all empty: game over, one turn after getting it all set up... though that part will take a significant amount of time.

Edit: 42 x 2 is 84. Not 48. Blah.