At the moment, major semiconductor foundry both in Asia and America are boasting the 10nm process node. However, there are some reports that this technology is quite tricky to accomplish, especially on a complete die. This is why Intel came up with a hybrid technology that will feature multiple process nodes and dies on its next line of Kaby Lake-G processors.
According to a recent leak from Benchlife, the new heterogeneous or hybrid design will make use of multiple process nodes for various parts of the processor. For the main core and GPU, Intel will use its current 10nm process node. For other parts like IO and communication peripherals, it could use a much older technology which is the 14nm process. Finally, on other non-critical parts of the chip, Intel could use a process node that is a lot more older, for instance a 22nm process node.
In addition to multi-node design, Intel is also considering a multi-die design strategy. In this case, the CPU will be interconnected with other peripherals like HBM2 memory using the flip-chip technology. Multiple dies will be connected via a silicon bridge instead of a usual interposer commonly used in fanout chip design. This new strategy will ensure that good die-to-die interconnection is being maximized while maintaining a small die footprint and lower overall cost.
At the moment, apart from a few details, nothing much is known about the Intel Kaby Lake-G processors. However, consumers can expect chips made from this node will be used for embedded and laptop applications. According to the same reports from Benchlife, the Kaby Lake-G will come with two SKUs, a 65W and a 100W TDP chip. Furthermore, unlike the earlier Kaby Lake processors, this version will not have in-built PCH controller. Instead, system developers will need to implement their own external PCH bridge to use the Kaby Lake-G. For more updates on the Intel Kaby Lake-G, be sure to check us out at TheBitbag.