NVLink was introduced with the Pascal generation graphics processors in 2016. NVLink was developed to address the bandwidth limitations which the PCI-Express data path imposed as demand for high data throughput continues to accelerate. Since its inception, NVIDIA improved the NVLink implementation and increased its bandwidth capability from a theoretical maximum of 80 GB/s bidirectional data throughput to 100 GB/s found on NVIDIA Quadro RTX 8000 and NVIDIA Quadro RTX 6000 graphics boards. With the latest Ampere architecture-based NVIDIA RTX A6000 professional graphics card, NVLink now provides up to 112GB/s of bidirectional data transfer between two connected NVIDIA RTX A6000 graphics boards.
