Such a shame the SGI engineers didn't try allocating more silicon to that tiny texture memory, this is perhaps (not an expert so don't quote me) the biggest bot

Such a shame the SGI engineers didn't try allocating more silicon to that tiny texture memory, this is perhaps (not an expert so don't quote me) the biggest bottleneck and annoying constraint in the system. It gets even crazier when you compare how much space the VI filter/blur takes on the silicon, removing it could easily allow for double the cache memory.