At its Data-Centric Innovation Summit, Intel presented its roadmap for future generations of Xeon processors . The Xeon-SP CPUs in the Cascade Lake series, which should be finished by the end of 2018, will gain an expansion of their AVX command set for AI calculations known as Deep Learning Boost (DLBoost for short). Compared to the current Xeon processors, which entered the market last year, the Cascade Lake chips are expected to be about 11 times faster at recognizing images, for example. Nonetheless, the new processors will still be manufactured using 14-nanometer construction; this also applies to their successor, Cooper Lake, which will be launched in 2019. Intel is planning additional improvements for Cooper Lake with regard to DLBoost’s performance and increased power in general for AI applications. 10-nanometer CPUs are not expected to be available until 2020.
Last year, Intel made approximately one billion (USD) in revenue off of CPUs for artificial intelligence workloads, according to its own statements. According to Navin Shenoy, the head of the corporation’s Data Center division, many customers prefer Xeon CPUs to specialized AI processors due to their greater versatility.