SOSA, Turing and GPGPUs

Posted by Laura B
Jauary 25, 2022

Sources:
eizorugged.com
nvidia.com

Speak to any HPEC engineer and there is one thing they all want from their technology - enhanced processing capabilities allowing sensors to be processed at even faster rates. NVIDIA has stepped up to the challenge with their Turing GPGPU microarchitecture, named after British mathematician and cryptanalyst Alan Turing.

The Turing development platform, called RTX, has double the parallel processing capabilities from the previous generation. It offers increased performance in AI and deep learning, and is equally suitable for sensor-intensive applications such as radar, Electronic Warfare (EW), unmanned systems, Signal Intelligence (SIG-INT) and Digital Signal Processing (DSP).

EIZO Rugged Solutions has harnessed the RTX platform in a new family of GPU cards, offering low-latency GPGPU processing including AI, flexible support for external high-speed I/O, and customisations to support many different design integrations utilising the OpenVPX standard.

The Condor XR1S 6U VPX is a SOSA-aligned high-performance computing & GPGPU OpenVPX card that hosts dual GPUs based on the NVIDIA Quadro RTX architecture. The Condor XR1S 6U VPX is based on the NVIDIA® Turing™ architecture using either the NVIDIA RTX5000™ platform or NVIDIA RTX3000™ platform, offering exceptional graphics and GPGPU compute capability with CUDA® support, AI, and deep learning.

Using OpenVPX as its framework, SOSA (Sensor Open System Architecture) aims to ensure improved SWaP-C optimisation, extensive reconfiguration options and rapid technology upgrades within a wide range of applications. It provides a standard building block for a more flexible solution overall as connectivity is driven by functionality of the chosen slot profile rather than the limitations of a particular board or module.

The Condor XR1 offers High-Performance Embedded Computing Systems (HPEC) with a solution that delivers multi-card performance, resulting in reduced system SWaP and increased compute capabilities. When utilising the RTX5000 (TU104) GPU, the card offers up to 6150 CUDA cores, 768 Tensor cores, 96 RT cores, and provides up to 18.98 TFLOP floating-point 32 performance with CUDA™ and OpenCL™ support.

If you’re looking for a SOSA-aligned 6U VPX GPGPU output card capable of significant processing power we might just have what you need. Get in touch – info@sarsen.net