The WOLF VPX3U-XAVIER-CX6-HPC is a 3U VPX GPGPU HPC Module. Firstly, the NVIDIA XAVIER contains an embedded Volta GPU. This provides the CUDA, and Tensor Cores required for high end data processing functions. Such as deep learn inference, machine vision, audio processing and video encoding & decoding.
Secondly, the 512 CUDA Cores provides 1.4 TFLPS for processing. Whilst the 64 Tensor cores provide the inference engine. Enabling it to achieve up to 20 TOPS (INT8) of deep learning inference. The Xavier Industrial Model also has new safety, reliability, and security features.
The NVIDIA ConnectX-6 SmartNIC provides a secure, high speed network data transfer and configurable PCIe Gen4 switch. Ideally meeting the data transfer rates for tasks such a sensor data processing and other C4ISR Tasks. It also provides support for RDMA over Converged Ethernet (RoCE). As well as support for NVIDIA GPU direct. This allows the fastest method for transferring data across the network to the GPU.
Finally, WOLF implemented it’s advanced cooling technology on this module. Moving heat using a low weight, high efficiency path from the SoC die to the wedgelocks. For more information, contact us.
Features of the 3U VPX GPGPU HPC Module VPX3U-XAVIER-CX6-HPC include:
- Embedded Volta GPU with 512 CUDA cores (1.4 TFLOPS) & 64 Tensor cores (20 TOPS Int8)
- Embedded 8-core NVIDIA Carmel ARM64 CPU, 2.26GHz
- 32 GB LPDDR4 256-bit memory, 137 GB/s, ECC support
- ConnectX-6, up to 100GbE Ethernet, PCIe Gen4
- Module power: configurable from 40W – 70W
- 2x Deep Learning Accelerator (DLA) engines for inference operations, 10 TOPS (Int8)
- 2x Vision Accelerator engines for 7-way VLIW Vision Processor, 1.7 TOPS Int8
- Dedicated programmable audio processor
- HEVC (H.265) and AVC (H.264) Volta NVENC and NVDEC supports 4K and 8K video encode/decode
- CUDA® 10, OpenGL® 4.6, OpenGL ES 3.2, Vulkan™ 1.0
- Storage: 64 GB eMMC 5.1
- USB 2.0 (up to 480 Mb/s); optional front panel port