Snap-in Tegra K1 COM tackles vision processing, deep learning
Toradex’s SODIMM-style “Apalis TK1” module runs Linux on a 2.2GHz quad-core Tegra K1 SoC, and offers extensive I/O including SATA, A/V, USB 3.0, and PCIe.
The SODIMM-style, 82 x 45mm Apalis TK1 computer-on-module is pin-compatible with Toradex’s earlier Nvidia Tegra 3 based Apalis T30 and i.MX6-based Apalis iMX6 modules. The Apalis TK1 moves up to Nvidia’s Tegra K1 SoC, which here has four of its Cortex-A15 cores clocked to 2.2GHz. A fifth companion Cortex-A15 handles low-power duty. The real draw here, however, is the CUDA-enabled, 192-core Mobile Kepler GPU, which can be used for vision processing, GPU-accelerated parallel processing, and deep learning/machine learning in addition to good old fashioned graphics rendering.