Intel Turns To AI For The Future: Deep Learning Chips Expected Next Year
Intel is on the verge of introducing a new development in personal computing with the 2017 rollout of an array of chips that will purportedly be dedicated to artificial intelligence. These will be released as co-processors, complementing the ones behind the computing process.
The Future Is AI
The AI chips are part of Intel's strategy designed to anticipate a deluge of applications and web services that will be using artificial intelligence. The dedicated chips will form part of the first-generation range but will give way to new chips that Intel aims to develop as mere modules within the main computing processors in the future.
Intel's aggressive push toward AI technology is largely driven by the waning PC market. The consensus is that the company is beating itself up for failing to spot the opportunity in the mobile market so it is now bent on rectifying that error in the emerging AI industry.
To this end, Intel's AI chips are not merely designed for computers but they are also being developed for other devices and applications such as smartphones, servers, drones, robots and autonomous vehicles, among others.
The Knights Crest
Out of the reported range of Intel AI chips, two have so far been officially identified. The first involves the Knights Mill, which is a chip designed for deep learning processes. It will be part of the stable of Xeon Phi chips and, according to Jason Waxman, corporate vice president of Intel's Data Center Solutions Group, it will perform four times faster than the existing Xeon Phi chip dubbed as Knights Landing. Both of these chips are part of a wider developmental initiative called Knights Crest.
Another potential AI chip will be related to the deep-learning technology associated with Nervana Systems. This company, which was acquired by Intel this year, focuses on building computer models for deep learning.
"We will apply Nervana's software expertise to further optimize the Intel Math Kernel Library and its integration into industry standard frameworks," Intel said in August. "Nervana's Engine and silicon expertise will advance Intel's AI portfolio and enhance the deep learning performance and TCO of our Intel Xeon and Intel Xeon Phi processors."
The Nervana Engine chip is being called Lake Crest and is expected to take the form of an add-on card. It is also going to be integrated in the Knights Crest project.
No Homegrown GPU
At this point, some observers believe that Intel still lacks complementary high-performance GPU technology. Previous attempts at developing its in-house GPU such as the 2009 project called Larrabee did not pan out.
An in-house GPU like the ones developed by AMD and Nvidia will help Intel and its chips to play a much larger part in emerging technologies today such as virtual reality and augmented reality.
Intel is addressing this issue by saying its chips will not rely on GPUs.
"Really, what most high-performance computers need — they don't need a GPU — they need parallel application performance, there are many ways to get that," Waxman told PCWorld.