Sign up for our newsletter
Technology / AI and automation

Intel “Still Playing Catch-Up” with AI Chip Offering

Intel’s tease of its pending high powered AI chip the Nervana Neural Net L-1000, or “Spring Crest” was broadly welcomed at its AI Developer Conference in San Francisco last week with the company saying “the first commercial NNP offering” is coming in 2019.

General Manager for Intel’s AI products, Naveen Rao said at the conference: “What we have created is the world’s first neural network processor (NNP)’’.

In an accompanying release, he added that chip will support bfloat16, a numerical format being adopted industry wide for neural networks.

“Over time, Intel will be extending bfloat16 support across our AI product lines, including Intel Xeon processors and Intel FPGAs. This is part of a cohesive and comprehensive strategy to bring leading AI training capabilities to our silicon portfolio.”

White papers from our partners

One close follower of the chip market, Scott Cadzow, told Computer Business Review: “What this means is that the neural processing capability of dedicated chips like the NNP-L1000 will be supported by mainline processors and thus the AI interface from neural networks to other processing paradigms can be incorporated.”

He added: “This will be important as I expect we’ll see chips like NNP-L1000 doing a similar role to crypto-coprocessors do today in taking load where their specialisation is recognised by the OS and development tools and brought into play as appropriate. This is a big advance – ultimately I can see AI support in chipsets being ubiquitous and Intel’s announcement appears to support this view.”

Intel Losing Ground?

Last month research house Gartner reported that South Korea’s Samsung Electronics has knocked Intel off its perch as the world’s biggest semiconductor company for the first time; a position that it had held unchallenged for 25 years.

And some have warned that as the AI and machine learning hardware race heats up —with Google announcing at I/O 2018 that is rolling out out its third generation of silicon, the Tensor Processor Unit (TPU) 3.0 — that Intel runs the risk of being left behind.

From Amazon to Facebook, via a growing swathe of start-ups, the battle is on to outmaneuver Nvidia on price or performance for machine learning tasks.

Gartner’s Mark Hung, Research VP for AI and IOT, told Computer Business Review today: “Intel is still playing catch-up in AI, especially when it comes to its chip offerings.  Its competitors, whether it’s Nvidia with its GPUs or Google with TPUs, have had multiple generations of chipsets deployed in production in a data center setting.”

He added: “Intel won’t be shipping its first generation product until the second half of next year. While Intel has constructed a strong overall portfolio with its acquisitions of Mobileye, Movidius, Altera, and Saffron, it needs to accelerate its efforts with the Nervana-based chipsets in order to not lose further ground to its main competitors.”

This article is from the CBROnline archive: some formatting and images may not be present.

CBR Staff Writer

CBR Online legacy content.