Think Silicon Compression Technology

  • Despite the limited power available in a battery operated hand-held device, a display must still have enough resolution and sufficient color depth to deliver the necessary information. Think Silicon proprietary framebuffer compression (TSFBc) efficiently reduces the power consumption of systems driving a display and thus distinctly extends battery life for hand-held applications. TSFBc is based on a proprietary lossy block-based algorithm and is performed in real-time. TSFBc can fit into systems with limited silicon area since the die area of the compressor is extremely small (below 20k gates). The decompression process is performed in real time using minimal hardware (below 2.5K gates). TSFBc has two variations, 4bits per pixel for systems with minimal available memory and 6 bits per pixel for devices that require exceptional image quality.

     

    4bpp TSFBc
    4bpp TSFBc is ideal for systems that have limited available memory. Visual artifacts are in the majority of use cases difficult to notice (as in the example below).

    image03
    original
    image03_text_tsc
    4bpp TSTBc

    6bpp TSFBc

    The 6bpp TSFBc has exceptional quality for any kind of images and is ideal for devices with high display resolution(4K). In the following example, all known block-based compression algorithms will result in poor performance while 6bpp TSFBc exhibits an exceptional quality.

    image01
    original
    image01_text_tsc
    6bpp TSTBc
  • Texture Compression is a very important aspect of computer graphics systems. Rendering and processing images requires significant memory bandwidth. Compressing images is of paramount importance as it reduces the required bandwidth and helps to utilize on-chip caches more efficiently. The result is higher speed and lower power consumption at the same time since the memory requests are significantly reduced. Think Silicon proprietary texture compression (TSTXc) is based on a lossy block-based algorithm and has four variations: 4bits per pixel with or without alpha channel, for systems with minimal available memory and 6 bits per pixel with or without alpha channel, for devices that require exceptional image quality. TSTXc compression process is performed offline in software while the decompression process is performed in 1-cycle using minimal hardware (below 3K gates).

     

    4bpp TSTXc
    4bpp TSTXc is ideal for systems that have limited available memory. Visual artifacts are in the majority of use cases difficult to notice (as in the example below).

    image03
    original
    image03_tsc
    4bpp TSTXc

    6bpp TSTXc

    6bpp TSTXc for RGB images ends-up with a significantly higher image quality and it does not suffer from pathological corner cases. In the example, all known block based texture compression algorithms will result in poor image quality. As can be seen in the example below, 6bpp TSTXc exhibits exceptional results.

    image01
    original
    image01_tsc
    6bpp TSTXc

    4bpp TSTXc with alpha

    4bpp TSTXc for RGBA images offers support for Alpha channel (A1 and A8) and is ideal for GUIs. 8-bits Alpha channel assumed in the example.

    icon037
    original
    icon037_tsc
    4bpp TSTXc with alpha

    6bpp TSTXc with alpha

    6bpp TSTXc for RGBA images offers support for Alpha channel (A1 and A8) with exceptional image quality for any kind of content. 8-bits Alpha channel assumed in the following example.

    icon025
    original
    icon025_tsc
    6bpp TSTXc with alpha
  • High quality compression is demonstrated in the following video.