First, the problem is somewhat related to the encoding of your tiff file. The data values are encoded as single-precision numbers, and the background points are encoded as 0xff7fffff in hexadecimal (equal to -3.40282346639e+38, minimum possible number in IEEE754 single-precision floating-point format).
Second, the tiff file only has 1 channel, and have no color information, so you need to create your colormap.
Both problems can be solved by converting the loaded matrix from single to uint16. Note that we are converting the values, not typecasting. Converting to uint16 will convert the background pixels from -3.40282346639e+38 to 0. The elevation values lie in the range [0 - 10000], uint16 is sufficient to hold these values. The integer values will make it an indexed image that can take value from a colormap. Try the following code
[A,R] = readgeoraster('DEM_30s.tif', 'OutputType', 'double');
low_single = typecast(uint8([255 255 127 255]), 'single');
mask = low_single==A;
A = A - min(A(~mask), , 'all');
brown_color = [0.8 0.2 0.1];
light_brown = [0.95 0.9 0.8];
t = linspace(0, 1, max(A,,'all'))';
cmap = t.*brown_color + (1-t).*white;
gs = geoshow(A, cmap, R);
gs.CData(repmat(mask,1,1,3)) = 255;