IBM has announced the launch of its ambitious server S822LC. These servers are designed for high-performance computers and are likely to offer a fillip to machine learning. IBM used Nvidia's NVLink technology to develop the interconnect for the server. IBM is targeting advance analytics and artificial intelligence fields with the launch of these servers.

Like Us on Facebook

Apart from Nvidia's NVLink technology, the server makes use of Nvidia's Tesla P100 GPU accelerators. According to Tech Crunch, the servers are also equipped with IBM's POWER8 CPUs. With NVLink technology, the CPUs and GPUs can interact with each other at a much faster rate. IBM explained that this would allow the computers to process "much larger data sets than possible in x86 systems with GPUs on the PCI-E interface."

S822LC server sports the first version of the NVLink technology. However, PC World reported that the company is planning to release updated servers with NVLink 2.0 and Power9 chips next year. The second version of NVLink technology is expected to be significantly faster than the current version.

IBM is the first major hardware company to employ NVLink. However, Nvidia sells server DGX-1 which uses the technology. The server is priced at $129,000. The company also said that companies such as Cray, Dell, and Hewlett-Packard Enterprise are likely to ship products with the technology by next year.

IBM has revealed that the server is expected to provide 21 teraflops of half-precision floating point performance. It also claimed that the speed is twice as fast as the ones offered by Tesla K80 GPUs using PCI-E interconnect.

The company has not provided any pricing information. The servers are expected to start shipping this month.