A number of embedded applications leverage the brute-force serial processing of x86 CPUs. Higher speed CPUs lead to higher performance embedded computers. This model works in general computing applications, but parallel computing solutions offer another option for high efficiency and performance.
Graphics Processing Units (GPUs) differ from CPUs in basic architecture. GPUs contains a larger number of smaller and efficient cores designed for parallel processing. GPUs effectively handle math-intensive computations focused on image processing, video encoding/decoding and other similar functions.
New GPU Development Options
In recent years, a number of GPU-based embedded solutions have come to the market. They provide R+D departments, start-ups, makers and hobbyists easier access to GPU-based embedded application solutions. The Nvidia Jetson TX1 is just such a solution.
Embedded GPU SOMs are driving alternative server topologies as well. NVIDIA partners Connect Tech and USES Integrated Solutions jointly developed a scalable array server using up to 24 TX1 modules. This compact solution targets markets such as micro data centers and machine learning which require the scalability, high performance and low power of GPU-based systems.
SEARAY Provides Electrical and Mechanical Flexibility
The TX1 uses Samtec’s SEARAY™ High-Density Open Pin Field Arrays for routing power, I/O and system control signals from the SoM to a carrier card. SEARAY’s 0.050″ x 0.050″ pitch grid provides the grounding and routing flexibility needed for solutions like the TX1. SEARAY’s various stack heights (ranging from 7 mm to 40 mm) offer mechanical flexibility in several embedded applications.
Samtec collaborated with Connect Tech on the development of the GPU scalable array server. In addition to identifying the SEARAY mated connector, Samtec provided engineering and product support to optimize the performance of the entire system.
For more information on Samtec’s SEARAY™ High-Density Open Pin Field Arrays, please download the following resources: