ARM, the chip foundation of virtually all of the smartphones made today, has announced the next generation of its Cortex CPUs and Mali GPU, with significant performance improvements made to both. In fact, ARM claims that its new Cortex-A77 CPU should offer performance “comparable to today’s mainstream notebooks.”
Both the new Cortex-A77 CPU and Mali-G77 GPU should eventually end up in smartphones, though when ARM chips do appear in devices they’re typically renamed by Qualcomm and other licensees. (Those licensees may also have the ability to tweak Arm’s design and clock speed, based upon their contract.) ARM also announced a new Machine Learning processor, with 5 TOPS (trillion operations per second) per watt.
Though ARM didn’t release details of either architecture, the company said that the Cortex-A77 boasts 20 percent improvement in instructions per clock (IPC) compared to the Cortex-A76, the ARM processor featured in today’s devices. The Mali-G77 improves even further, providing more than 40 percent performance improvement over the Mali-G76 also used in today’s smartphones.
ARM is also well on its way toward improvements made in machine learning. Since announcing the Project Trillium machine-learning compute platform last year, ARM has made enhancements to its ML processor which include increasing energy efficiency by more than two times, as well as scaling to 8 cores for a total of 32 TOP/s, the company said.
Project Trillium, as well as the new Machine Learning processor based upon it, are being designed for the mobile market specifically, with an eye toward object detection. It’s that object detection, for example, which allows the software in a smartphone to identify an object in the foreground of pictures and blur the remaining background, creating a “portrait mode” created computationally, mimicking the bokeh effect created by optical lenses.