Swedish flag with blue background and yellow cross, symbolizing Sweden's national identity and heritage.

Vretenborgsvägen 28, floor 6
SE – 126 30 Hägersten
Phone: + 46 8 683 03 00

Norwegian flag featuring a blue cross with white borders on a red background, symbolizing Norway.

Midstranda 51
NO – 2321 Hamar
Phone: + 47 62 54 02 91

Flag of Denmark, national flag with red background and white cross, symbolizing Danish heritage and pride, used in official and cultural contexts in Denmark.

Hassellunden 14
DK – 2765 Smørum
Phone + 45 70 300 310

Get connected

Linkedin
Contact us

SP GR5/Bs1-yzRCS3

Bragi – SOSA Aligned 3U OpenVPX GPGPU Card with NVIDIA RTX PRO Blackwell by Concurrent

The Concurrent Bragi is a rugged SOSA aligned 3U OpenVPX GPGPU processing card powered by the NVIDIA RTX PRO 4000 or 5000 Blackwell GPU, designed for high-performance GPU computing in defence and mission-critical embedded systems. With up to 10,496 CUDA cores, 24GB ECC GDDR7 memory, up to 1,824 FP4 AI TOPS, NVIDIA GPUDirect RDMA, and conduction cooling rated to -40°C at card edge, it delivers next-generation AI inference and GPU processing capability in a rugged VPX form factor. Available from Recab, your Nordic partner for Concurrent advanced modular computing solutions.
contct-now-icon (2) Get datasheet
Description
The Concurrent Bragi is a high-performance GPGPU processing card built for defence, radar, intelligence, and mission-critical signal processing applications that demand GPU-accelerated computing within rugged 3U OpenVPX chassis systems. Powered by the NVIDIA RTX PRO Blackwell generation GPU in either the RTX PRO 4000 with 7,680 CUDA cores and 16GB ECC GDDR7, or the RTX PRO 5000 with 10,496 CUDA cores and 24GB ECC GDDR7, it delivers exceptional parallel processing capability for AI inference, computer vision, electronic warfare signal processing, and real-time video analytics workloads. Both variants are built on the Blackwell GPU architecture supporting DirectX 12, OpenGL 4.6, Vulkan 1.3, CUDA, CUDA-X, and OpenCL, with FP4 AI performance reaching up to 1,334 TOPS on the RTX PRO 4000 and up to 1,824 TOPS on the RTX PRO 5000, making them capable platforms for on-platform AI inference without cloud dependency. NVIDIA GPUDirect RDMA enables direct data capture from network interfaces or FPGAs into GPU memory, bypassing the host CPU to minimise latency in high-throughput signal processing pipelines. Hardware video codec support covers H.265 HEVC, H.264 MPEG4/AVC, and AV1 with 9th Generation NVENC encoders supporting 4:2:2 chroma and 6th Generation NVDEC decoders for efficient video processing. The board is SOSA aligned per the relevant 3U VPX slot profiles and supports both 1 inch pitch conduction cooling per VITA 48.2 and 1.2 and 1.5 inch pitch air flow through per VITA 48.8. Environmental ratings cover CC4 at -40°C to 85°C at card edge, CCW4 at -40°C to 71°C at cold wall, and AFT2 at -40°C to 55°C, with 40g shock and 0.1g2/Hz random vibration resistance. Windows and Linux on x86 are supported. Available from Recab, your Nordic partner for Concurrent rugged embedded computing and OpenVPX solutions.
Features
  • NVIDIA RTX PRO Blackwell GPU with up to 10,496 CUDA cores, 320 Tensor Cores and up to 1,824 FP4 AI TOPS performance
  • Up to 24GB ECC GDDR7 graphics memory with 256-bit memory interface for high-bandwidth GPU workloads
  • NVIDIA GPUDirect RDMA for direct data capture into GPU memory bypassing host CPU for ultra-low latency pipelines
  • SOSA aligned 3U OpenVPX form factor with conduction cooling rated CC4 at -40 degrees C to 85 degrees C at card edge
  • H.265, H.264 and AV1 hardware codec with 9th Gen NVENC encoders and 6th Gen NVDEC decoders for video processing
  • 40g shock and 0.1g2/Hz random vibration resistance for deployment in rugged airborne and ground vehicle platforms
Specifications
GPU NVIDIA RTX PRO 4000/5000
Memory 16/24 GB GDDR7
Cores 7,680/10,496 CUDA Cores
AI TOPS Up to 1,824 FP4 AI TOPS
Operating Temp -40C to 85C
Power 90W-150W
Form Factor 3U VPX
Shock 40 g
Vibration 0.1 g²/Hz
Humidity 95% Without Condensation
  • Defence and intelligence GPU-accelerated signal processing in rugged 3U OpenVPX embedded systems
  • AI inference and machine learning at the edge in SOSA aligned mission computing platforms
  • Electronic warfare and radar systems requiring high-throughput GPGPU processing with low latency RDMA
  • Real-time video analytics and hardware video encoding and decoding in airborne and ground vehicle platforms

Get datasheet