On x86, a double is only 8 bytes. How good of precision are you getting on z80 with less?How did you determine that he's using 64-bit doubles instead of 32-bit floats, or even 24-bit floats? (although those would have horrible precision).
Just a guess. On x86 I would use 64 bit floats. Perhaps AHelper can clarify, is it just that the makeup of the tree (not necessarily the numbers) that is smaller?