Intel admitted that the rejection of monolithic crystals will reduce the cost of production
Intel CEO Robert Swan represented the company at Credit Suisse’s virtual conference this week, and his talk could be given a little more attention. Speaking about layout solutions, the head of Intel admits that the use of chiplets can reduce the cost of production.
For rival company AMD, this has become an undeniable thesis since the creation of the first EPYC and Ryzen processors, the company even conducted a comparative analysis of the potential costs in options with a monolithic processor crystal and using chiplets. In the case of the 32-core EPYC savings were obtained twice, and a 64-core processor with the monolithic chip AMD would have even failed to create.
Robert Swan began his reasoning about the choice of processor layout by stating that customers, no matter how Intel products are made, are more important for them to rhythmically improve performance as new generations of processors are released. Intel begins to denote a multi-chip layout everywhere with the characteristic phrase “disaggregated design”, which roughly means “separate layout”.
It not only gives you the flexibility to design a wide range of products, Swan says, but also allows you to achieve high production volumes with lower costs. It is enough to recall Intel’s recent history of under capacity to understand why the company believes in the potential of multichip products. As soon as it became clear that the 10nm process was delayed, Intel’s pipeline began to fill up with 14nm processors with an ever-increasing die area. This not only increased the unit cost but also reduced the total production of processors. Small crystals can be produced in large quantities, allowing new products to be brought to market faster.
Intel will be more active in engaging contractors in the production of some of the crystals of such composite products. At the same time, the company’s profit margin may decrease slightly, since paying a contractor is not as profitable as making a product on its own. For this reason, Intel will carefully analyze the economic impact of each decision to place orders on the assembly line of other companies. If we talk about the range of products preparing for release in 2023, Intel plans to decide on the location of their production in January next year. In this context, we mean mainly 7nm products, some of which will be produced by contractors.
