This a little of topic but still closely relates to the tiling problem.
When writing GeoTiff files, by default the tile size is number of samples x number of lines. I.e., the whole image is written as one whole block.
This is neither tile-based nor line-based and thus extremely slow to read by other software (I tested ArcMap and Envi).
In a classical line-based approach the block size is number of samples x 1.
Converting the image via e.g. gdal_translate speeds up reading operations significantly.
this is how it looks like in gdalinfo:
tile-based:
Band 1 Block=256x256 Type=Float32, ColorInterp=Gray
line-based:
Band 1 Block=16635x1 Type=Float32, ColorInterp=Gray
snap GeoTiff:
Band 1 Block=16635x10662 Type=Float32, ColorInterp=Gray
Is there a specific reason for doing it like this?