New Catalyst-X™ PXIe-GPUs Employ NVIDIA® Blackwell RTX Pro SFF PCIe GPUs to Deliver Watershed Performance for Cost-Effective, Software-Defined, Compute-Intensive Signal, Image & Video Processing, ML / DL Training & Inference and Air-Gapped RAG / LLM Applications
FORT WORTH, Texas, May 12, 2026 /PRNewswire-PRWeb/ -- RADX® Technologies, Inc. ("RADX", www.radxtech.com), a leader in COTS, high-performance, computing technologies and products for modular Test & Measurement (T&M), today at NI Connect 2026, announced two new Catalyst-X™ PXIe-GPU™ Modules based on NVIDIA's latest Professional Small Form Factor (SFF) PCIe GPUs: the Blackwell RTX Pro 2000 (https://tinyurl.com/5acu2svt) and the Blackwell RTX Pro 4000 SFF (https://tinyurl.com/3e4r7brp). The new PXIe-GPU-BWRTXPro2k-16GB and PXIe-GPU-BWRTXPro4kSFF-24GB PXIe-GPUs employ RADX' Patented (US Patent No. 12284547) Catalyst-X PXIe Carrier Card Technology to transform the NVIDIA SFF PCIe GPUs into single-slot PXIe modules.
The PXIe-GPU-BWRTXPro2k-16GB features 16GB of GDDR7 that supports 288 GB/Sec of memory bandwidth and delivers up to 17.0 FP32 / 110.0 TF32 TFLOPS. The PXIe-GPU-BWRTXPro4kSFF-24GB features 24GB of GDDR7 that supports 432 GB/Sec of dedicated memory bandwidth and delivers up to 24.0 FP32 / 155.0 TF32 TFLOPS. The new PXIe-GPUs feature exceptional memory bandwidth and compute performance, with support for GDDR7, DisplayPort 2.1b, and enhanced nvCOMP, NVENC and NVDEC compression technologies. As a result, these new PXIe-GPUs set new standards for price/performance as cost-effective, easy-to-program, COTS accelerators for software-defined, compute-intensive PXIe T&M applications in Aerospace, Defense, Automotive, Manufacturing, Computer Vision, Semiconductor and other important segments.
"Our new Blackwell-based products help fill out the Catalyst PXIe-GPU lineup by closing the performance gap between our Ada-based RTX 2000 and RTX 4000 SFF based PXIe-GPUs and our flagship L4 Data Center PXIe-GPU," said Ross Q. Smith, RADX Co-Founder and CEO. "Plus, the new features like enhanced compression and decompression enable new applications for PXIe systems, such as advanced, air-gapped LLMs like Gemma 4, for the first time."
"The capabilities of the new Blackwell-based RTX Pro 4000 SFF PXIe-GPU is rivaled only by RADX' L4 and Orca eGPUs," said Natan Biesmans, CEO of G2CPU. "The RTX Pro 4000 SFF sustains over 7.0 GB/Sec or 1.77 GSPS of 32-bit I/Q from the PXIe-8881 to the PXIe-GPU in LabVIEW / G2CPU with about 20% more headroom than the Ada-based RTX 4000 SFF, which is quite remarkable in a Gen 3 PXIe System."
EASE-OF-PROGRAMMING VIA LABVIEW, PYTHON, C/C++ AND MATLAB - With their intrinsic ease-of-programming, NVIDIA GPUs are the most popular compute accelerators in the world, with millions of developers using NVIDIA GPUs to accelerate their applications. Catalyst PXIe-GPUs offer the same ease-of-programming, since they are essentially NVIDIA Pro SFF GPUs repackaged for PXIe using RADX' patented Catalyst-X technology (US Patent No. 12284547). RADX PXIe-GPUs employ NVIDIA CUDA®, OpenCL®. TensorFlow® and Vulcan® for compute acceleration of Python and C/C++, MATLAB™, LabVIEW™, PyTorch™, Ansys™, FFmpeg, and other frameworks, as shown in the attached graphics, all of which can dramatically reduce development time and costs for use under Windows and Linux.
For LabVIEW, RADX PXIe-GPUs are supported by toolkits from G2CPU (www.g2cpu.com), NGENE (http://www.ngene.co) and Graiphic (https://graiphic.io/), which enable direct programming from within LabVIEW for signal, image & video processing, data science and AI under Windows and Linux.
For real-time, wideband signal and image processing applications in PXIe Systems under Windows, Linux and Real-Time Linux, G2CPU uniquely provides optimized I/O support for popular NI PXIe digitizers, VSTs and video capture modules using Zero Copy, Async I/O and other advanced techniques that are essential for high-performance applications.
PRICING AND DELIVERY - Both of these new Blackwell-based PXIe-GPUs are single-slot, passively cooled, PCIe Gen 4 x8 modules with 4 x miniDP 2.1b 8k outputs and feature a US COO, BAA/TAA Compliance and an ECCN of 4A994.L. The Q226 MSRP for the PXIe-GPU-BWRtxPro2k-16GB is $9,999 USD and the Q226 MSRP for the PXIe-GPU-BWRTXPro4k-24GB is $12,999 USD. Delivery for both products is ~30 to ~45 Days ARO. Additional details are shown in the attached Table of Key Specifications. For more info, visit www.radxtech.com.
DEMOS AT NI CONNECT 2026 – RADX will demo the new Blackwell-based PXIe-GPUs along other RADX PXIe Modules in Booth #K139 at NI Connect in Fort Worth, TX, 12-14MAY2026. At NI Connect, RADX will show the G2CPU (https://www.g2cpu.com/) RTSA Toolkit featuring PXIe-GPU accelerated, LabVIEW-based, real-time signal analysis, record and playback with lossless real-time compression. In addition, RADX will demo PXIe-GPU accelerated Computer Vision by Acquired Data Solutions (https://acquiredata.com/), with Local, Air-Gapped PXIe-based LLM-based control by Extreme Scale Solutions (https://www.extreme-scale.com/).
ABOUT RADX – Founded in 2011, RADX Technologies (www.radxtech.com) is a US-based, high-tech SMB that develops and manufactures advanced, COTS, High Performance Computing (HPC) technologies and products for Modular Test and Measurement (T&M) applications. RADX is an NI Alliance Partner and RADX products are designed and assembled in the USA, are BAA / TAA compliant and are available from RADX directly, from TestMart (www.testmart.com) for USG / GSA purchases, and from qualified PXIe system integrators and value-added resellers worldwide.
Media Contact
Ross Q. Smith, RADX Technologies, Inc., 1 6196771849 700, [email protected], www.radxtech.com
SOURCE RADX Technologies, Inc.
Share this article