51ÔÚÏß

New RADX Catalyst-X PXIe-GPUs Include NVIDIA Ada L4 Data Center, RTX 4000 SFF and RTX 2000 Professional GPUs for Compute-Intensive Signal, Image & Video Processing, ML / DL Training & Inference and LLM Applications

FORT WORTH, Texas, April 28, 2025 /PRNewswire/ -- RADX® Technologies, Inc. ("RADX"), today at NI Connect 2025, announced four new Catalyst "in-chassis" PXIe-GPU™ Modules that combine RADX' Patent-Pending Catalyst-X PXIe Carrier Technology with powerful and easy-to-program NVIDIA Professional GPUs. The new models expand the RADX PXIe-GPU offerings for with models that support up to 24GB of GDDR6 and 30.3 FP32 TFLOPS, providing unmatched compute performance and value for modular Test & Measurement (T&M), Automotive AD/ADAS, Electronic Warfare (EW) and other compute-intensive PXIe applications.

UNPRECEDENTED PXIe PERFORMANCE

The new models include a low-cost module based on the Ampere RTX A1000 GPU and three advanced modules based on Ada RTX 2000, Ada RTX 4000 SFF and Ada L4 Data Center GPUs with 16, 20 and 24 GB of GDDR6 and 12.0, 19.2 and 30.3 FP32 TFLOPS, respectively. These new GPUs represent the highest compute performance PXIe modules available today and are ideal for cost-sensitive, yet demanding signal/image/video processing and real-time Machine Learning / Deep Learning (ML / DL) inference and training applications, including Retrieval-Augmented Generation Large Language Models (RAG LLMs)– all of which require extreme memory and compute capabilities.

"With up to 30.3 FP32 TFLOPS, 24GB of GDDR6 and PCIe Gen 4 x8 support, these new Catalyst PXIe-GPUs are the most powerful compute accelerators available in PXIe today," said Ross Q. Smith, RADX CEO. "Now, for the first time ever, test system integrators and end-users can use popular and easy GPU software programming techniques to develop and deploy a wide variety of low-latency, wideband, signal, image and video processing applications, combined with ML/DL training and inference applications including 70 billion parameter RAG LLMs that use two of our Ada L4 DC PXIe-GPUs in parallel."

EASE-OF-PROGRAMMING VIA LABVIEW, PYTHON, C/C++ AND MATLAB

With their intrinsic ease-of-programming, NVIDIA GPUs have rapidly become the most popular compute accelerators in the world today, with literally millions of engineers, developers, computer scientists and others using NVIDIA GPUs to accelerate their applications. Catalyst PXIe-GPUs are essentially NVIDIA Pro GPUs repackaged for PXIe. So, they also feature outstanding ease-of-programming by directly supporting MATLAB™, LabVIEW™, Python and C/C++, accelerated via NVIDIA CUDA®, OpenCL® and TensorFlow®. These new PXIe-GPU models support popular computing frameworks including PyTorch™, Ansys™, FFmpeg, and others, which dramatically reduce development time and costs for Windows and Linux operating environments. For LabVIEW users, RADX PXIe-GPUs are supported by NGENE DeepLTK and CuLab () and G2CPU () Toolkits, which enable programmers to directly program PXIe-GPUs from within LabVIEW for Signal, Image and Video Processing, Data Science and ML/DL and LLM Applications under Windows and Linux and, with G2CPU, RT-Linux.

"G2CPU features highly optimized support for GPU-based signal, image, video and data processing, along with advanced multi-GPU scaling, async I/O and GPU-accelerated rendering under Windows, Linux and uniquely, RT-Linux, which is critical for low-latency T&M and Hardware-in-the-Loop projects," said Natan Biesmans, G2CPU CEO. "With these new Catalyst PXIe-GPUs in a PXIe-1095 (Gen 3) Chassis, low-latency LabVIEW / G2CPU applications can now sustain well over 1 GHz of Bandwidth per Channel with 16-bit I/Q data and 1M Point FFTs- with headroom for post-processing. This is a watershed level of performance for PXIe that can eliminate the need for GPU Servers in most cases."

NEW CATALYST-X PXIe-GPUS DETAILS, PRICING AND DELIVERY

Based on RADX' patent-pending Catalyst-X Carrier Card Technology that enables PCIe add-in-cards to be easily adapted to PXIe, these new models complement the existing lineup of RADX PXIe-GPUs based on NVIDIA T600, T1000, RTX A2000 and Ada RTX 2000 GPUs, as well as RADX Orca-C2Xâ„¢ PXIe-to-PCIe eGPUs. These new passively cooled, single-slot, PCIe Gen 4 x8 models, excluding the Ada L4 DC model, include 4 x miniDP outputs with support for up to 8k resolution. Pricing on these new models ranges from $5,999 to $15,999 USD, excluding US import tariffs on GPUs, with 30-to-60-day ARO delivery beginning in late Q225. For more info, please visit

DEMOS AT NI CONNECT 2025

RADX will be demonstrating Trifecta and Catalyst PXIe-GPU Modules, as well as several of the Company's high-performance, high-capacity fixed, removable and hot-swap PXIe Data Storage Solutions at NI Connect 2025, held 28-29APR2025 at the Fort Worth Convention Center (Texas). Stop by for a demo and say, "howdy."

ABOUT RADX

Founded in 2011, RADX Technologies, Inc., is a high-tech small business that develops COTS, HPC PXIe and PCIe hardware and software products including PXIe and PCIe GPUs, eGPUs, SSD / RAID Subsystems and Network Interface Cards that enable advanced signal/image/video/data processing and ML/DL/LLM training and inference TM applications within the EW, Semiconductor and Automotive markets. As an Emerson / NI Alliance Silver Partner, RADX focuses on products and solutions that complement the Emerson / NI PXIe and USRP product lines. RADX products are BAA / TAA compliant and are available on GSA from TestMart at . RADX is headquartered in California with development locations in Utah, New Mexico, and Taiwan. For more info on RADX, please visit , email [email protected] or call +1 (619) 677-1849 x1.

Media Contact

Ross Smith, RADX Technologies, Inc., 1 6196771849 700, [email protected],  

SOURCE RADX Technologies, Inc.