"Why trade 120megs of mesh for 80pixels of JPG.?" Exactly. What's the point of using a 70,000 vertices figure when it's covering only a visible area of 1000 pixels? You could use a 10,000 vertices figure instead, saving lots of resources, without any visible difference at all. That's what the LoRes figures were made for - with polygon smoothing enabled and a good bump or displacement map, who can see the difference?