ARM's new DynamIQ tech puts cores at the service of AI

Although in existence for decades, artificial intelligence and machine learning have only recently hit mainstream, becoming the hottest thing in the tech industry. But for all the capabilities and ambitions of AI and ML, they are practically limited by the computing power that runs those algorithms and data processing tasks. In other words, advancement in AI requires not just more powerful but also more efficient computing processors. Which is why ARM Holdings is announcing the next evolution of its ARM chip design which it calls DynamIQ.

Just to be clear, ARM isn't announcing an actual chip that you'll see in a device soon. ARM actually doesn't make processors but only designs them for others, like Samsung, MediaTek, and others, to base their own chips on. That said, these designs are significant because they pretty much sets the baseline capabilities of each generation of ARM chips.

ARM compares this new DynamIQ technology to the revolutionary big.LITTLE technology it introduced back in 2011. ARM big.LITTLE basically enabled a processor to have two different sets of cores, one high-power and one energy-efficient, and which set goes to work depends on the task required. Those that required more power, like video decoding or gaming, would run on the more powerful, but also more power-hungry, cores while the slower cores would take over menial tasks.

DynamIQ promises to take this big.LITTLE technology to the next level, improving the heterogeneous processing capabilities of ARM processors. In a nutshell, this means that only the cores that are really needed will be activated, no matter the combination, be those be one, three, four, or seven. While currently processor configurations would have a more or less even number of cores (2+2, 2+4, 4+4, etc.), ARM DynamIQ would pave the way for 1+3 or 1+7 combinations.

Don't expect these ARM DynamIQ processors in smartphones yet, if at all. While AI is also making inroads in the mobile space, ARM is squarely targeting a different kind of market. Particularly workstations and servers dedicated to artificial intelligence and machine learning. In other words, Intel's lucrative market.

SOURCE: ARM