A successor to the groundbreaking interconnect is already on the table for development as the Graphics Processors with the Nvidia’s NVLink throughput started rolling out. Though IBM’s servers powered by Power9 chips are already incorporated with NVLink 2.0 and were expected to hit the market as early as next year but Nvidia haven’t yet announced any time frame for its GPUs integrated with the Power9 chip.
Despite IBM’s announcement of the chip release time frame the company hasn’t yet decided any particular date for it. ‘Volta’ codenamed GPU from Nvidia will be released next year mentioned an article on Nvidia’s website. According to IBM, the NVLink 2.0 Power9 chips will communicate at 25Gbps which is around 7 to 10 times the speed of currently available PCI-Express 3.0. This is because of the multiple communication links available for the NVLink 2.0 which when combined will provide a massive output.
NVLink 1.0 is used in GPUs like GeForce GTX 1080 and the recently available Tesla P100 with a Pascal architecture but the Volta chip will go beyond one step. No other components will work with the interconnect NVLink 2.0 I/O as it is designed for the Nvidia GPUs only as of now. Cognitive computing like GPU driven application will be hugely benefited with NVLink 2.0 technology.