From a full-fledged 3D-TV (with a three-dimensional picture on the screen without using stereo glasses), we are separated by one small problem - performance. It is still not possible to generate holograms in real time, because no processor can cope with rendering such a flow of information, even on the basis of a finished 3D model.
Japanese scientists published in the journal Optics Express a scientific work describing the architecture of a specialized holographic cluster HORN (HOlographic ReconstructioN). Installation works like this. After the calculation, on the cluster connected to the personal computer, the image arrives on the 0.7 ”HDTV microdisplay, and the hologram is taken from it.
')
The sixth cluster model (HORN-6) of 16 boards showed a performance of about 1 fps on a complex model of 1 million points, as well as a performance of 10 fps on a model of 100,000 points. According to the calculations of scientists, with this application, their installation is about 4600 times more productive than a PC based on a Pentium 4 with a clock frequency of 3.4 GHz.
Each HORN-6 board carries four Xilinx XC2VP70-5FF1517C FPGA processors and an interface to a 256 MB DDR-SDRAM memory module. On this architecture it is theoretically possible to make a ready-made holographic video system right now, if the cluster is scaled several times. Additionally: Three demo videos (MPG) and scientific work (PDF) in a zip-archive . The first video shows a dinosaur hologram generated using a computer model consisting of 11,646 points. On the second - a chessboard (44,647 points), on the third - an attraction (95,949 points).