64 seperate reverbs is gonna be hard for any cpu, regardless of which reverb algorithm is used.
may i suggest that an alternative is to send your grains through the same reverb, but to alter the amount of each grain that gets sent into the reverb. so, for example, grain 0 is sent 100% to the reverb, grain 1 is sent 90% to the reverb, and 10% is passed on dry, grain 3 is sent 10% to the reverb and 90% is passed on dry....etc
if think of reverb as the sound of a room, then the mix % is the position of the object in the room.