According to a report from The Information, Apple plans to overtake Intel in terms of sheer compute density in the coming years. Leveraging TSMC’s N3 (3nm) process, the Cupertino giant is looking to increase the core count of its flagship products to as much as 40 compute cores.
Apple first plans to release a dual-chiplet SKU with a total of 20 cores (10 per die) for its second-generation processors. These will be fabbed on TSMC’s 4nm-class nodes, and act as a testbed for an MCM architecture. Both the M1, M1 Max, and M1 Pro are based on a monolithic design. The 2nd Gen SKUs will power the M2 Macs and iPads in 2022 and 2023.
Finally, we have the third-generation lineup, codenamed Ibiza, Lobos, and Palma. These will be fabbed on TSMC’s 3nm process and leverage four dies/chiplets. Basically, we’re looking at up to 40 cores, though it’s unclear whether Apple will enable all of them in one of the SKUs. These monster processors are slated to land with the M3 Macs and iPads in 2024, giving AMD and Intel ample time to catch up both in terms of IPC and compute density. From what I can tell, the x86 leaders will still have an advantage in terms of sheer IPC and compute, but Apple’s chips are likely to be more power-efficient.
Source: The Information