Intel’s Kaby Lake family is set to expand again, with the Kaby Lake-G series making an appearance via leaked documentation. Kaby Lake-G will retain a 14nm manufacturing process and will reportedly have a package size of 58.5mm x 31mm.
Right off the bat, we can see that this is a marked increase in size compared to the desktop oriented Kaby Lake-S (37.5mm2) and the mobile Kaby Lake-H (42mm x 28mm). According to leaked slides, there will reportedly be 65W and 100W SKUs, each with a quad-core configuration and a separate dedicated GPU on the package that connects via PCI x8.
The TDP for both of the Kaby Lake-G processor SKUs are well above the 45W rating of Kaby Lake-H, which makes us wonder what’s in store for Intel’s heterogeneous “Mix and Match” design that it will employ. The heterogeneous design would play in Intel’s favor thanks to the use of an Embedded Multi-Die Interconnect Bridge (EMIB), which can enable high-bandwidth and high-density die-to-die interconnects.
EMIB allows Intel to efficiently and cost effectively integrate CPU or graphics components using different process nodes, produced in completely separates factories if need be. The company could then use its more [relatively speaking] exotic 10nm manufacturing process for its CPUs, while using more mature 14nm technology for graphics and other less mission-critical components.
According to BenchLife, the TDP increase and larger package size of Kaby Lake-G point to the inclusion of HBM2 memory and the use of AMD graphics. However, we have seen nothing to indicate that this is the case (either through the rumor mill or in the slides), although it’s quite possible that the publication is expecting some kind of hardware “fruit” from Intel’s licensing agreement with AMD.
And the fact that the leaked documents still clearly point to Intel’s GT2 graphics also throws some cold water on the Intel/AMD/HBM2 rumor.