S3TC

S3TC, or S3 Texture Compression, is a lossy texture compression method used in real-time 3D graphics. Developed by S3 Graphics, it became widely adopted and is commonly referred to by its DirectX-era name DXTn, covering DXT1, DXT3, and DXT5 variants. The technique is designed to reduce texture memory usage and bandwidth while preserving as much visual quality as possible for real-time rendering.

The algorithm operates on fixed-size blocks of texels, specifically 4x4 blocks. Each block encodes two 16-bit color endpoints in RGB565 format and a small palette of colors derived by interpolation between those endpoints. Each texel in the block selects one of the palette entries via a few bits of index data. DXT1 stores only color data for the block, while DXT3 and DXT5 add alpha information: DXT3 contains explicit per-texel alpha data, and DXT5 uses an alpha pair plus interpolated indices to represent transparency.

Block sizes differ by variant: DXT1 uses 8 bytes per 4x4 block, while DXT3 and DXT5 use 16 bytes per block. This results in substantial memory savings—commonly on the order of 4:1 to 8:1 depending on content and variant—at the cost of some loss of detail and the potential for block-based artifacts, especially in gradients or sharp edges.

S3TC/DXTn is supported by major graphics APIs and hardware, notably OpenGL (GL_EXT_texture_compression_s3tc) and Direct3D. It has been a de facto standard for texture compression in many years and influenced subsequent BCn-based formats used in modern GPUs.