Don Knuth, in its Volume 1 of The Art of Computer Programming argues against garbage collection by saying that early freeing can lead to a more efficient usage of the memory by the allocator.

In the following diagrams which represent a memory state, the heavy gray area is not used by the program but the garbage collector did not reclaim it yet, the black areas are in use, and the white areas are free to use. gc1 If the program tries to allocate one small block in the current state, the allocator will use some of the free space, gc2 Now the garbage collector reclaims the gray zone and we are left with two disjoint blocks of memory. gc3 On the other hand, if memory is managed manually the gray block has been freed eagerly, nogc1 Thus, when the application needs some extra memory it can be tightly packed and only leaves one large hole in the memory. nogc2

I guess this is all to be taken with a grain of salt since there are also easy arguments for garbage collection (for example fast allocation). All in all I feel that programmers using languages with GC will advocate for it, and the converse for the others. The reasonable choice to make is probably application specific and even in this case it is likely to be hard and controversial making the right decision at the planning stage.