Intel's market position is far from ideal. Its 10nm problems have ended up causing major processor stock problems. The company to prosper and leave these problems behind is investing shovelfuls of money. Of particular note is the signing of Raja Koduri, who has the mission of developing a competent and dedicated GPU. by Jim Keller.
Jim Keller is the one who led the team that developed AMD's Zen architecture. Shortly after, he left the company for NVIDIA and finally went to Intel, we suppose, with an offering made up of many zeros. In an interview Keller has revealed that he wants to update the architecture and return to the 'tick-tock' system.
No products found.
Intel prepares an unprecedented revolution
Keller in the interview has given a lot of relevant data. While he hasn't said it, from what we've been hearing, Intel's monolithic processors have their days numbered. Keller wants to develop a new architecture every five years, instead of every 10 years, as is the case today. This will allow Intel to be much more competitive.
This process, as is logical, will not be from today to tomorrow. Intel currently offers Ice Lake processors based on the Sunny Cove architecture, while Tiger Lake is based on Willow Cove. The latter are characterized by a redesigned cache structure and will hit the market by the end of the year. Some processors that have not yet been designed by Keller.
Possibly the first architecture most worked by Keller is Golden Cove that would arrive next year. Although some experts suggest that an architecture entirely developed by Keller may take a long time to arrive. According to speculation, it would be in 2023, five years after Keller signed for Intel, when an architecture fully developed by him arrives.
Radical change that comes relatively late
For an hour and a half Keller talks about the elements needed to develop a new architecture for Intel. He emphasizes that a new architecture should be presented every three years, but if it is from scratch, the ideal is every five years. Keller indicates that today this happens every ten years and stresses that he is ready to change this line.
But like everything in this life, it involves "many difficulties to be solved." There are technical aspects, corporate strategies or the direction of the company, they must be in favor of change. It also indicates that renovating the architecture every 10 years is a significant cost saving. This is because renewing it every 5 years implies a lot of investment and financially it may not be good, being aware that administrators must also think about the interests of investors.
Intel is undoubtedly on a pilgrimage through the desert, as it has already happened to AMD. The Bulldozer architecture FX processors were a fiasco, unable to compete with the Core. The company passed Bulldozer by a steamroller and destroyed it, stopping the machines to launch something new, as was Zen. Note that Zen came to the market very green and that it has been polishing and improving, but AMD's margins of maneuver were minimal.
Intel is not AMD by chance. The company will not stop short to launch a new architecture, it will be making small changes. Currently we see that there are only refried, but there are small adjustments and changes that will end up leading to a new architecture. Intel is clearly going through a bad time, but despite everything, it is in a better situation than AMD between the Bulldozers and Zen.
Source: MD
