EIZO Rugged Solutions Inc. has introduced a new series of SOSA-aligned 6U OpenVPX processing cards featuring dual NVIDIA Blackwell embedded GPUs designed for high-throughput AI and sensor processing at the embedded edge.
The Condor XR2S 6U VPX Series functions as a high-performance computer and GPGPU processing platform tailored for advanced applications such as electro-optical/infrared (EO/IR) sensor processing, fire radar control, AI inference, signal processing, and real-time video analytics. The lineup includes two variants designed to maximize processing density within VPX-based systems: the Condor XR2S-B5000, which features dual NVIDIA RTX PRO 5000 Blackwell Embedded GPUs, and the Condor XR2S-B4000, built with dual NVIDIA RTX PRO 4000 Blackwell Embedded GPUs.
Integrating dual GPUs onto a single payload allows the modules to execute compute-intensive workloads in parallel. To facilitate a flexible, high-bandwidth, and low-latency interconnect, the architecture incorporates dual Microchip Switchtec PFX PCIe Gen4 switches between the GPUs and the VPX backplane fabric. This design minimizes host CPU overhead by optimizing PCIe lane distribution and supporting peer-to-peer GPU communication. Additionally, each PCIe switch enables mating I/O modules to communicate directly with the host CPU or the GPUs via NVIDIA RDMA, with individual configuration options for routing to various endpoint configurations.
Chris Fadeley, Chief Technical Officer (CTO) of EIZO Rugged Solutions, said, “As sensor resolutions and data throughput requirements continue to increase, customers need architectures capable of supporting significantly more fabric bandwidth and high-speed interconnect. The Condor XR2S 6U VPX series with dual GPUs and PCIe Gen4 switches enables greater PCIe and Ethernet scalability, allowing system designers to process larger datasets with lower latency and fewer infrastructure bottlenecks.”
The underlying NVIDIA Blackwell architecture supports capabilities for high-throughput high-performance embedded computing (HPEC) environments, including fifth-generation Tensor Cores, fourth-generation RT Cores, GDDR7 memory capacity, and GPUDirect RDMA. The XR2S-B5000 model provides 10,496 NVIDIA CUDA cores, 320 fifth-generation Tensor Cores, and 80 fourth-generation RT Cores alongside 24 GB of GDDR7 ECC memory per GPU. Meanwhile, the XR2S-B4000 variant delivers 7,680 CUDA cores, 240 fifth-generation Tensor Cores, 60 fourth-generation RT Cores, and 16 GB of GDDR7 ECC memory per GPU.
The architecture also leverages Multi-Instance GPU (MIG) technology. This allows a single NVIDIA RTX PRO Blackwell GPU to be securely and efficiently split into multiple fully isolated GPU instances, each with dedicated compute cores, cache, and high-bandwidth memory. As a result, system architects can allocate GPU resources dynamically across different applications, sensor pipelines, or AI models running in parallel.
Engineered for deployment in demanding environments, the Condor XR2S Series is designed to meet MIL-STD-810 environmental specifications for temperature, shock, and vibration. The hardware is fully aligned with the SOSA Technical Standard and supports 10.6.4 slot profiles, reinforcing its versatility and readiness for next-generation high-performance embedded deployments.
