The new SOSA-aligned VPX6-731 rugged HPEC dual GPU Module joins the company’s NVIDIA-accelerated GPU family, bringing with it NVIDIA Blackwell architecture and enabling integrators to accelerate AI/ML and sensor processing at the edge with high-speed connectivity and advanced thermal management in a COTS solution.
The device delivers up to 100 TFLOPS of peak compute performance and 48 GB of GDDR7 memory by integrating one or two NVIDIA RTX PRO 5000 Blackwell GPUs in a single-slot 6U VPX module. Curtiss-Wright offers conduction-cooled, liquid flow-through (LFT), and air flow-through (AFT) configurations to support higher power density and energy efficiency. This powerful solution is designed for SWaP-constrained aerospace and edge computing applications requiring advanced display, encoding, decoding, sensor processing and deep learning inference for AI workloads.
Key Features include:
- Support for one or two NVIDIA RTX PRO 5000 Blackwell GPUs in a 6U OpenVPX board;
- Available with conduction, LFT or AFT cooling for energy efficiency and higher power density;
- Up to 48 GB of GDDR7 memory providing a large local buffer and faster throughput compared to prior generation GPU boards;
- NVIDIA Blackwell architecture supporting multiple data types for improved performance, efficiency, and programming flexibility;
- Part of Curtiss-Wright’s Fabric100 portfolio for proven interoperability in advanced HPEC systems
Image of the VPX6-731 courtesy Curtiss-Wright
