An interesting bit of info has surfaced on the Bilibili forums. According to a reputed tipster (金猪升级包), Intel’s 15th Gen Arrow Lake processors will make scant use of the advanced 20A node, relying on TSMC’s 3nm (N3P?) for most CPU compute tiles. He alleges that the 20A process will only be used for the midrange Arrow Lake-S CPU tile with 6+8 cores (6 “Lion Cove” P-cores and 8 “Skymont” E-cores). The higher-end 8P + 16E CPU tile will reportedly be fabbed on a TSMC process.
On the mobility side, all the CPU and iGPU tiles will be fabbed on a 3nm TSMC node. This isn’t certain, as Intel and AMD have continuously prioritized the notebook market. Not using your flagship manufacturing technology for your most important chips will be strange. The iGPU tile will be the same as the Core Ultra “Meteor Lake” CPUs. However, the Arrow Lake-S desktop platform will only get the 4 Xe Core variant (512 ALUs), leaving the 7 Xe Core iGPU (896 ALUs) for notebooks.
Arrow Lake will allegedly drop the LP E-cores (E cores on the SoC die), which makes sense as ultra-low-power efficiency isn’t a priority on desktop PCs. Only DDR5 will be supported, while the NPU will be retained (likely the same as Meteor Lake).
The 15th Gen Arrow Lake processors will likely be called the Core Ultra 2 series processors. However, until there’s an official announcement from Intel for better discoverability, we’ll continue to use the existing title.