VPX3U-AD5000E-CX7 (WOLF-153L)
The VPX3U-AD5000E-CX7 HPEC module is equipped with an NVIDIA RTX 5000 Ada embedded GPU alongside a ConnectX smartnic. This powerful GPU is designed to deliver exceptional processing capabilities for high-performance embedded computing (HPEC) and artificial intelligence (AI) applications. The ConnectX-7 component ensures robust Ethernet and PCIe connectivity, facilitating the efficient transfer of large datasets.
The NVIDIA Ada architecture features CUDA cores tailored for HPEC, 4th generation Tensor cores optimized for AI and data science tasks, and 3rd generation Ray Tracing (RT) cores that enhance visual rendering accuracy. Utilizing TSMC’s 4N NVIDIA Custom Manufacturing Process, the Ada GPU achieves improved efficiency. The more compact design of the ada GPUs allows for a greater number of CUDA and Tensor cores to operate at elevated clock speeds while maintaining the same power consumption, resulting in a notable increase in performance per watt compared to WOLF’s earlier models. For instance, when comparing the 144L to the 153L, the latter can deliver two to three times the GFLOPS/W, depending on power and thermal conditions.
The NVIDIA ConnectX-7 SmartNIC enhances the module’s capabilities by providing PCIe and Ethernet connectivity. This component is especially suited for high-speed, secure data transfers essential for data-intensive operations, such as sensor data processing and various C5ISR applications. Additionally, the ConnectX-7 supports RDMA over Converged Ethernet (RoCE), which offers the fastest data transfer method between the network and the GPU.
Key Features
- NVIDIA RTX™ 5000 Ada GPU featuring 9728 CUDA Cores, 304 Tensor Cores, and 76 RT Cores
- NVIDIA® ConnectX®-7 enabling up to 100GbE network connectivity and a PCIe Gen5 switch
- 16 GB GDDR6 memory with a 256-bit interface, achieving speeds of up to 576 GB/s
- Configurable module power ranging from 100W to 140W
