This number still seems to be way bigger than actual possible number of seeds. Are you sure you are only checking the valid ones?
There is a 4 byte checksum in them and although the collision chance of 4 byte out of SHA256 hash is higher but I don't think it is high enough to produce that many variations. (I'll try to test it in the afternoon when I get back home).

Edit: I ended up testing it anyways. With 12 words it seems that about 5% of the combinations are correct. The rest should be rejected. So with your words there should only be around 40 million seeds to check not that huge number you posted.
Also:
Two questions, is it the number of seeds you check per second or is it number of keys that you derive from those seeds per second? If it is the first one then it is very slow, I am getting 16k/s with my c# managed code using only 1 CPU core (corei3)! You should be able to get around 100k at least if you run it on parallel and with a stronger CPU.