Deep learning uses powerful GPUs to train deep neural networks using massive data sets. Until now, hardware hasn’t kept up with the power of GPUs. For those who want to harness the power of the GPUs and other accelerators to enable computers to learn faster than ever, Cubix built C-Link – a PCIe Gen 3 bus with an effective peer-to-peer (P2P) transfer rate of 12.5Gb per second, simultaneously in both directions among all 8 GPUs in C-Link.
C-Link Systems are especially designed to support data analytics and machine learning, with lab results surpassing NVIDIA CUDA test expectations. Each C-Link System is tested for performance and stability before shipping, and Cubix will help test your application to optimize GPUs. Get results with a high degree of accuracy with C-Link’s highest possible P2P transfer speeds.