Condor NVA2104AxX

High Performance GPU

Chip-down NVIDIA® Ampere A2000 with up to 8.25 TFLOPS FP32 Single Floating Point Performance

I/O Intensive

Four 3G-SDI & two CVBS (NTSC/PAL/SECAM) video inputs with four 3G-SDI video outputs

All-In-One Solution

Simultaneously capture, process, display, encode, decode, and stream video/audio data, including metadata

Condor NVA2104AxX

XMC Video Capture & GPGPU Card Supporting Four Analog/Digital Inputs and Outputs

The Condor NVA2104AxX is a rugged high-performance XMC graphics card that captures both analog & digital raw frame-by-frame audio/video and metadata with exceptionally low latency. This card is designed with the NVIDIA® Ampere A2000 GPU supporting 8 GB GDDR6 graphics memory with Error Correction Code (ECC), and features 2,560 CUDA Cores, 80 Tensor Cores, and 20 RT Cores for accelerated performance in neural network training and inferencing functions. This all-in-one solution provides the ability to simultaneously capture, process, display, encode, decode, and stream video data while supporting CUDA and OpenCL based GPGPU computing, AI processing, deep learning and H.265/H.264 encoding & decoding.

XMC Video Capture & AI-Enabled GPGPU Processing Solution

The RTX A2000 supports PCI Express Gen 4, which provides double the bandwidth of PCIe Gen 3, improving data-transfer speeds from CPU memory for data-intensive tasks like AI and data science. The Condor NVA2104AxX card is very I/O intensive as it supports four 3G-SDI inputs, two CVBS (NTSC/PAL/SECAM), and two stereo line in audio inputs, as well as four 3G-SDI outputs. SDI VANC KLV metadata insertion and extraction is also supported. This innovative video capture and GPGPU card is designed for applications that combine legacy video with newer digital video formats in high-end surveillance applications. The video capture (frame grabber) auto detects the input resolution and then transfers raw video frames directly to GPU memory or host memory using NVIDIA GPUDirect™ RDMA. In GPU memory, the applications can do processing such as image analysis, image enhancement, 360 degree video stitching, sensor fusion and target detection, using GPGPU (CUDA / OpenCL), all with very low latency.

Condor NVA2104AxX Block Diagram
Condor NVA2104AxX Block Diagram

Product Specifications

Graphics ProcessorNVIDIA RTX A2000 GPU (Ampere Architecture)
Supporting DirectX 12, OpenGL 4.5, and Vulkan 1.2
InterfaceXMC 1.0 or XMC 2.0
8 Lane PCIe 4.0
Graphics Memory8 GB GDDR6 with ECC memory
128-bit Memory Interface
up to 192 GB/s Memory Bandwidth
Video InputsFour 3G-SDI and Two CVBS (NTSC/PAL)
Rear Pn6 XMC I/O. VITA 46.9 x12d+x8d+24s
Video OutputsFour 3G-SDI
Rear Pn6 XMC I/O. VITA 46.9 x12d+x8d+24s
Audio InputsTwo Stereo Line In
GPGPU Capabilities2560 CUDA Cores. 80 Tensor Cores. 20 RT Cores.
Up to 8.25 TFLOPS FP32 Single Floating Point Performance
Supports CUDA 11 (Compute Capability 8.6) and CUDA-X
OpenCL 3.0 and Shader Model 5.1
H.265 (HEVC) / H.264 (MPEG4/AVC) Hardware Encode & Decode
NVIDIA GPUDirect® RDMA, NVENC, NVDEC
Power Consumption25 - 50 W
Operating Temperature (MIL-STD-810)-40°C to 70°C (Rugged Air Cooled)
-40°C to 85°C (Rugged Conduction Cooled)
Please refer to the Hardware User Guide for details on
temperature/performance characterization.
Vibration (MIL-STD-810)0.1 g²/Hz
Shock (MIL-STD-810)40 g
Humidity (MIL-STD-810)95% Without Condensation
Software & Platform SupportWindows or Linux on x86
VPX & PCIe

Ruggedized Products for Air, Land, Sea, and Space

Explore the industries we proudly serve!