The 3300GTX provides CSPI‘s TeraXP™ Embedded Servers with a high performance coprocessor blade featuring many CUDA cores to accelerate parallel streaming and FFT based applications. Optimizing computational density in a 6U OpenVPX slot, the 3300GTX is configured with one NVIDIA MXM Graphics Processing Unit (GPU) and one FDR InfiniBand 56 Gbit/s Host Channel Adapter (HCA).
Application development for the GPGPU’s is enhanced by the use of the industry standard CUDA, OpenCL and OpenACC toolkits.
The 3300GTX board also features failover capabilities via the HCA supporting a 56 Gbps QSFP transceiver on the front panel and a VITA 66.1 optical interconnect on the backplane. The addition of this fiber-optic connectivity brings benefits in higher bandwidth and lower weight.
Features & Benefits
- One MXM site supporting a NVIDIA GPGPU for accelerated performance
- Implemented with the Mobile PCI Express Module (MXM) version 3.0 standard, capable of hosting newer GPGPUs as they become available
- Converged Fabric Technology for high bandwidth InfiniBand, Ethernet, and Fibre Channel connectivity
Technical Specifications
- 6U OpenVPX GPGPU Coprocessor Blade
- One MXM site supporting a NVIDIA GTX 560M GPGPU with Fermi Architecture
- Mellanox HCA with ConnectX®-3 Virtual Protocol Interconnect (VPI)
- Supports 56Gbps FDR InfiniBand to front panel transceiver
- VITA 66.1 (MT optical connector) interface on the backplane
- mDP Interface (one link to Front Panel, one link to backplane)
- Available in Air-Cooled or Conduction-Cooled versions
