A reliable workflow
Bump derives slope at render time
A bump shader derives local slope from a scalar map at render time; a normal map stores the direction directly.
Displacement can alter the silhouette
Displacement consumes a height field to move vertices or samples and can change silhouettes when sufficient geometry is available.
Choose by evidence, cost, and distance
Choose normal maps for efficient directional detail, height for reusable scalar relief, and displacement when geometric depth is worth the cost.
Compare the data each map actually stores
The normal map vs bump map vs height map distinction begins with representation. A height map stores one scalar per texel: relative high and low. A bump shader samples that scalar and derives a slope during rendering. A normal map stores the already-derived three-component direction, usually in tangent space. Displacement also consumes height-like data, but uses it to move vertices or surface samples when the renderer and mesh support that operation. These maps may begin from related information, yet they are not different file names for the same calculation.
Compare lighting, silhouette, and cost
In a normal map vs bump map vs height map comparison, normal mapping gives direct control over fine directional lighting and is efficient for real-time materials. Bump retains an editable height source but spends shader work deriving local gradients and varies with implementation. Height alone does nothing until a shader interprets it. Displacement can change silhouettes, intersections, and true surface depth, but requires subdivision, tessellation, or another sampling technique and can cost more memory and processing. None of these methods guarantees convincing results if the source contains baked shadows or an incompatible UV scale.
Choose by the visible requirement
Use the normal map vs bump map vs height map question to identify what viewers must see. For pores, cloth weave, and small scratches that affect highlights without changing the outline, a normal map is often appropriate. For an authorable scalar source that may feed several effects, preserve height. For broad stones, deep tracks, or terrain that must alter a grazing silhouette, consider displacement or geometry. Assets can combine them: displacement for large form and a normal map for microdetail. Avoid duplicating the same strength at both stages, which can exaggerate slopes.
Keep conversions honest
A height map can be converted to a normal map because local differences provide slope, but the reverse loses information: integrating normals into a globally consistent height field is not generally exact. Likewise, a color photograph is neither height nor bump data unless its brightness reliably follows relief. The most useful normal map vs bump map vs height map workflow preserves the original scalar data, documents scale and edge behavior, exports vector maps losslessly, and validates every interpretation in the target renderer. Choose the map from available evidence rather than from the most dramatic preview.