Intel Corporation (NASDAQ:INTC) could dent NVIDIA Corporation (NASDAQ:NVDA)’s and Advanced Micro Devices, Inc. (NASDAQ:AMD)’s prospects in the Artificial Intelligence (AI) space, and drag NVDA stock and AMD stock.
Shares of NVIDIA Corporation (NASDAQ:NVDA) and Advanced Micro Devices, Inc. (NASDAQ:AMD) had a dream run in 2016. More recently though, Nvidia’s seemingly unstoppable rally was somewhat temporarily stalled, after Citron Research issued an extremely bearish call on the stock.
While some of the concerns raised by Citron are probably debatable, one threat in particular warrants a deeper look. And that’s the looming threat in the Artificial Intelligence (AI) space from Intel’s latest salvo, its third-generation Xeon Phi processor, dubbed Knights Mill. While Intel’s renewed aggression is also a threat to AMD’s fledgling foray in the deep learnings and AI segment, the threat is larger for Nvidia, given that a large part of the optimism around the stock is tied to the company’s prospects in the burgeoning field of AI.
Intel’s Xeon Phi Knights Mill – A Real Threat To Nvidia And AMD?
We all know how Intel Corporation (NASDAQ:INTC) all but owns the data centre processor market, with an estimated market share of 99%. Yet, the giant chipmaker hasn’t managed to gain much ground in futuristic markets like AI, where Nvidia reigns supreme. To take on Nvidia, in August 2016, Intel announced a new server processor, its third-generation Xeon Phi processor code-named Knights Mill, aimed specifically at artificial intelligence applications.
Also Read: Citron Says Nvidia (NVDA) Stock Belongs At $90, Does It Really?
Can CPUs take on GPUs, which are currently a clear favorite in this market? While some believe that Intel needs a GPU to take on the likes of AMD and Nvidia, the notion may not be entirely correct, going by this article on Extreme Tech (1):
“GPUs are good at these kinds of computing projects because the projects map well on to the hardware we use for gaming — not because there’s something magic about graphics processors that makes them uniquely and specifically suited to the tasks.”
The article also explains that Intel could get the job done by building “a GPU-style compute engine without any of the IP blocks or hardware that transform it into a graphics card.” While Intel’s Xeon Phi was introduced as a GPU, only to be “reinvented into a vector processor”, Joel Hruska of Extreme Tech opines that “There’s nothing to say Intel can’t bend it back a bit, possibly by building lower-precision registers or offering them as options on certain types of hardware.”
Follow Nvidia Corp (NASDAQ:NVDA)
Follow Nvidia Corp (NASDAQ:NVDA)
How Intel Claims To Have The Edge Over Nvidia
In fact, Intel Corporation (NASDAQ:INTC) claims (2) that GPUs aren’t even a scalable solution for the future. The primary reason for GPUs to be used in machine learnings applications is their ability to run multiple calculations simultaneously. However, Intel’s executive vice president and general manager of the Data Center Group, Diane Bryant, claims that “The market is still nascent, so the current implementations are small enough that they could use GPUs, but it won’t scale in the future.” Intel claims it’s Xeon Phi processor is way faster than Nvidia’s GPUs for machine learnings algorithms. To be specific, 2.3 times faster. Of course, Nvidia has refuted the claim (3), brushing aside Intel’s assertion with a counter-claim which suggests that the opposite is true:
“Nvidia claimed if Intel had used the latest technology, Nvidia would achieve 30% faster training machine learning models over Intel.”
As you’d expect, Intel responded by explaining how current systems use a Xeon processor in combination with GPU accelerators, which the company claims is “suboptimal implementation.” According the article on Forbes, Intel’s vice president and general manager of cloud, Jason Waxman said:
“the new Xeon Phi processor will eliminate the need to swap between a central processor and a GPU accelerator. With a chip like the upcoming Xeon Phi, all the processing required for machine learning tasks takes place on a single chip, thus eliminating the need to switch between a main processor and a GPU accelerator.”
Given Intel’s grip over the data center processor market, which translates to an estimated 99% market share, the option to partially or fully do away with GPU accelerators could be an enticing proposition for some, if not all of its customers.