Intel New Knights Landing Chip Set To Compete With NVIDIA
This new chip offered by Intel is the first full implementation of Intel’s Scalable System Framework and differs from its predecessors in three critical ways. First the chip is a bootable CPU and not a PCIe accelerator. Second the chip has an on-die Omni-Path interconnect for high bandwidth and low latency scaling between nodes. Finally the chip offers high bandwidth stacked die memory.
KNL was unveiled at the ISC High Performance event which was originally the international supercomputing event designed to target High Performance Computing (HPC) and for training Deep Neural Networks (DNNs). DNNs are used to power autonomous cars, natural language processing, image recognitions and the ads you see on Google.
Intel started the show off with a bang by boasting they can beat NVIDIA in Deep Learning by two times and NVIDIA countered by saying their new accelerators can beat Intel’s by double as well. Deep Learning has been touted by every tech company as the Next Big Thing. Intel is late to the party but has been expending a great deal of time, money and energy to capitalize on the potential of Deep Learning.
NVIDIA has seen their Datacenter business skyrocket by more than 63% to a mind-blowing $142 million dollars in their latest quarter with AI applications being a huge net boost to NVIDIA’s Telsa product sales.
Intel sees the pot of gold at the end of the rainbow and is now offering a new more attractive approach to this market by offering t=e GPU-class performance on a host CPU that is potentially simplifying programming, compatible with the massive base of Xeon applications and reducing data movement costs. The issue however is that NVIDIA already has momentum from their GPUs therefore Intel will have to make sure the product they are offering is at the very least comparable to those already out on the market.
Consumers will be expectedly cautious and will take their time validating Intel’s GPU performance claims before making the transition. Given that NVIDIA currently dominates the Deep Learning application market, Intel will have a lot of ground to make up.