G2CPU and RADX

Enable LabVIEW-based Real-Time, High-Performance Computing in PXIe Systems in record time for LabVIEW with efficient on-GPU interfaces to Python, MATLAB and Open Architecture AI Apps.

G2CPU RTSA Demo: PXIe Ultra-Wideband RF Record, Playback and Analysis

See this LIVE at GDEVCON EU

1,25 GHz from NI VST to RADX GPU through G2CPU Zero Copy.

1M+ Point FFT on RADX GPU with G2CPU MPP for Gapless Real-Time Analysis and AI.

Lossless data storage and compression to RADX SSDs up-to 7.1GB/s 61TB per PXIe slot using G2CPU Test and Measurement - ZIP.

RADX Technologies builds COTS PXIe modules that put NVIDIA-based GPU compute, high-speed data storage and high-performance networking inside the PXIe chassis: the only PXIe-GPUs on the market, the fastest, highest-capacity PXIe-SSDs available and the only 100GbE RDMA NICs in the market today.

G²CPU is the COTS toolkit that puts RADX hardware to work directly within LabVIEW, without writing a line of CUDA. Available on the NI Tools Network, G²CPU uniquely unlocks full GPU performance to enable high-performance and analysis in LabVIEW with support for on-GPU co-processing in MATLAB, Python, PyTorch, TensorFlow, LLMs and other open-architecture computing frameworks.

Separately, G2CPU and RADX remove key barriers to PXIe performance. Together we close the loop: acquire, process on the GPU, compress, and stream to SSD or LAN at full PCIe bandwidth, all inside the NI ecosystem, all based on COTS products you can order today.

RADX TECHNOLOGIES

The highest-performance COTS PXIe Compute, Data Storage and Networking modules available today.

Visit RADX

G²CPU for LabVIEW

One stop shop for GPU and high-performance, software-defined compute and I/O intensive applications across the NI ecosystem.

Visit G2CPU

RADX Technologies

High-performance COTS PXIe modules for defense, aerospace and T&M.

PXIe-SSDs: unmatched capacity and performance

Venturi patent-pending PXIe-SSDs with 4 to 61.44 TB per slot of high-speed, data storage with up to 7.1 GB/Sec sequential read & write.

PXIe-SSDs: Unprecedented features

Trifecta PXIe-SSDs with removable 1, 2, 4 and 8TB data cartridges for rapid system software changes and secure data handling in sensitive environments.

The only PXIe-GPUs available today

RADX patented Catalyst PXIe-GPUs bring easy-to-program NVIDIA Pro GPUs to PXIe, with 1 to 30+ FP32 TFLOPS and up to 24GB GDDR for unprecedented performance in HPC applications

Full, unshared PXIe bandwidth

Every module gets the whole slot’s bandwidth, not a shared fraction. With Orca-C2X, add virtually any PCIe card to virtually any PXIe system, with no Thunderbolt required.

Host adapters for every fabric

RADX patented Catalyst PXIe-HBAs cost-effective RDMA PXIe-GbE with 4×10, 2×25 and 2×100 GbE and 4x20Gbps PXIe-USBC host adapters.

COTS, compliant, available now

In production since 2019, RADX PXIe Modules are BAA/TAA compliant, GSA-available, and programmable from LabVIEW, MATLAB, Python and C/C++ with ~30- day lead times and discounts for VARs, SIs, OEMs and Educational Users.

Visit RADX

G²CPU the GPU Toolkit for LabVIEW

GPU acceleration, native to LabVIEW and the NI ecosystem.

100x to 500x performance boost

CUDA in LabVIEW RT

Up-to 50GB/s per GPU lossless compression and decompression

Write once, run anywhere

Your code, your LabVIEW, now at GPU speed. Decide at runtime whether it executes on CPU, CUDA, OpenCL or Intel oneAPI. No recompiling, no retooling. Multiple GPUs, mixed vendors, same application.

Massive parallel processing

Batch your operations, saturate your GPU, allowing you to do over a billion FFTs per second. Any N-dimensional operation batches across any array of N dimensions or more.

Unmatched I/O

Zero-copy streaming from instrument to GPU to disk at full PCIe bandwidth. Asynchronous I/O for raw throughput, low-latency synchronous I/O where determinism matters.

TM Zip compression

GPU-accelerated compression built for test and measurement data. Compress inside the pipeline, so your storage and your network carry more of what you actually captured.

Debug like it’s still LabVIEW

LabVIEW probes work on data that lives on the GPU, on any device, from any vendor. Because debugging shouldn’t stop where acceleration starts.

Built for the AI workflow

Share a live GPU buffer with any CUDA-capable environment through standard CUDA IPC. Prototype and train in Python, deploy in LabVIEW, and let the data stay exactly where it is.