Following the new AI trend, AMD joins Nvidia and Intel in the race to put its chips into AI applications with the latest Radeon instinct line of graphic processors. The new line of processors is designed specifically targeting AI machine learning applications for everything from self-driving cars to art. The company is launching three new graphics cards under its latest Radeon Instinct brand from various families.
The MI6 will be based on the AMD’s Polaris architecture offering 5.7 teraflops of performance and 224GBps memory bandwidth consuming only up to 150 watts of power. Gazing at the MI8, this GPU is built around R9 Nano and offers 8.2 teraflops of performance and 512GBps of memory bandwidth, consuming only up to 175 watts of power. As the rate of RAM dependency of AI and deep learning are still not clear, AMD probably will hold the sales until it has a viable use-case.
The MI25 will be a Vega-derived chip possibly much faster than the other two versions of the new launches. Even though the company is still confidential about the technical specifications of MI25, they claimed that the performance of its top Radeon Instinct GPU beats Nvidia’s best offerings. As the MI6 is 5.7 teraflops and the MI8 is 8.2 teraflops, the MI25’s probably will be 25 teraflops beating Nvidia’s current Titan X Pascal card rated at 11 teraflops.
As machine learning is the current trend, AMD announced that it will release its ROCm deep-learning framework, as well as an open-source GPU-accelerated library called MIOpen by the first half of 2017.