Wednesday, May 31, 2017

New ARM Smartphone Processor Claimed Strong Features AI

ARM announced two new mobile processor design Cortex A75 and A55. Both are claimed to be able to support the features of machine learning and artificial intelligence (AI) in a mobile device.

Usually engine learning features and AI are handled through cloud computing (cloud). In this way, a device will not work too hard to handle those two features.

However, the company design processor design is used in almost all mobile devices are making the Cortex A75 and A55 to be able to work on two capabilities directly in the device thanks to DynamIQ technology, as reported by KompasTekno from GSM Arena on Tuesday (30/05/2017).

DynamIQ technology is designed entirely to boost machine learning performance and AI. Thus, both processing can be done directly in handheld devices.

In addition, DynamIQ technology is said to be more flexible and manageable than big.LITTLE used by previous generation ARM.

Because DynamIQ allows original equipment manufacture (OEM) to install a large processing core along with seven small processing cores within the same cluster.

Each of these clusters can be a maximum of eight processing cores. While one CPU can be filled up to a maximum of 32 clusters. In other words, a single CPU can be made to have 256 processing cores.

The effect of the new technology makes the ARM Cortex A75 and A55 CPUs have higher performance than its predecessor.

Cortex A75 showed a 48 percent better performance on the Octaned benchmark and 34 percent better on the Geekbench benchmark than its predecessor.

While the Cortex A55 showed a better 14 percent performance on the Octane benchmark and 20 percent better on the Geekbench benchmark. In addition, this middle-class CPU also has a single thread performance 18 percent better than A53.

ARM also announced the presence of Mali G72 graphics processor created using Bifrost architecture. When compared to its predecessor, Mali G72 is claimed to have 40 percent higher performance, 25 percent more power-efficient, and 17 percent more efficient in machine learning management.

No comments:

Post a Comment