A few days ago, we reported on Intel 's decision to discontinue its Nervana chips for Artificial Intelligence . Apparently, Nervana underperformed, and continuing with this architecture after the acquisition of Habana Labs made no sense. To avoid erroneous speculation, the company wanted to provide all the details regarding this decision . Well, their version.
Artificial intelligence is a multi-billion dollar potential market expected to grow year after year. All the major chip manufacturers are developing their own solutions focused on this field. Technically, NVIDIA seems to have the biggest lead , with powerful solutions already on the market and the most innovative in this area.
[amazon box="B07S9M32FG"]Intel gives its version of why it has stopped working at Nervana
Naveen Rao, Managing Director of Intel's Artificial Intelligence Platforms Groups, gave an interview on Forbes. It has stressed that the factor of change in development priorities has nothing to do with performance. Despite the fact that they have stopped working at Nervana, Intel promises "future improvements" in the NNP architecture of Havana.
Customer input helped the company make the decision to build its Havana-based AI roadmap, bringing future enhancements to the PNN Nervana architecture and software to create what Intel believes will be leadership products.
Naveen Rao, Managing Director of Intel Artificial Intelligence Platforms Groups
It appears that much of the weight of the platform change would have been customer demand. Customers would have shown great interest in the Goya and Gaudí chips from Habana Labs for offering more performance at a lower cost. These chips are based on a converged architecture, something that offers even more power.
Gaudí, for example, takes advantage of a great variety of future products such as accelerators, processors or network equipment. Intel did not want to continue developing two separate lines, assuming unnecessary expense.
Source: TH
