PXIe-GPU Compute & Graphics
PXIe-GPU Family of Products
RADX PXIe-GPUs bring the easy-to-program power of NVIDIA® Professional GPUs to PXIe for accelerated graphics, video, image, signal & data processing and ML / DL / LLM training & inference applications.
Easy-to-Program
LabVIEW, MATLAB, Python & C/C++ with CUDA®, OpenCL® & Tensor Core acceleration
Signal & Image Processing
4.0 to 91.1 FP32 TFLOPS — up to 100×–200× vs. Xilinx KU060 FPGAs and popular PXIe ECs
HW Video Encode / Decode
NVENC/NVDEC engines for real-time 2K/4K/8K video compression at up to 100× CPU CODEC performance
AI / ML / DL / LLM
Up to ~200× inference & training performance versus embedded controller CPUs
Wide Chassis Support
GPUs available for 50W, 58W, and 72W/slot PXIe chassis — eliminates the need for separate GPU servers
2D / 3D Graphics
Up to ~200× higher graphics performance vs. embedded controller GPUs
NVIDIA RTX A1000 · ADA RTX 2000/4000 SFF · BLACKWELL RTX PRO 2000/4000 SFF · L4 DC
Catalyst-GPU
Ampere → Ada → Blackwell GPU compute acceleration for all PXIe/CPCIe chassis
NVIDIA RTX A2000
Trifecta-GPU
High-performance GPU compute for demanding EW & T&M applications
COTS PXIE EGPU & PCIE-TO-PXIE SUBSYSTEM
Orca-C2X
Extreme performance FHFL GPUs in any PXIe system — no Thunderbolt required
Need help choosing the right GPU?
Contact RADX sales for application-specific guidance.