is there another data type other then double/single/uint with less memory requirements?

i have a problem with big matrics (20e3 by 20e3) currently they take too much ram, i guess that the problem has something to do with the floating point, for example uint takes only a fraction from a double data but the problem is that this data type round the number into an integer.

single is a half of double in memory consumption but still not good enaught.

is there another data type that meets my requirements? or shuld i just change the settings of the matlab floating point? if so, then how?

maybe you have another suggestion?

[Azzi Abdelmalek]

Duplicate

[Kobi]

it's not the same question. uint8 is the wrong choice for me

[Wolfgang Schwanghart]

As said by David Young, single is the smallest data type. Perhaps you could tell more about what kind of analysis you want to do. There are functions such as blockproc, for example, that let you work with images that don't fit into main memory.

Answer 1

[David Young]

If you need floating point, then single is the smallest data type.

The best solution may well be to process the large matrices in sections if the computation allows it.