Let’s break it down.
• The 12900K has 8 big cores with hyperthreading plus 8 little cores without hyperthreading.
• The single-threaded performance of the big Golden Cove cores in Cinebench will likely be around 30% higher than the Sunny Cove cores found in 11900K.
• It is possible the multithreaded performance of the big cores when working together to be further improved due to improvements in hyperthreading, due to larger shared cache sizes, due to DDR5, etc. So far, when the all-core and single-core frequency is the same, the multithreaded score with hyperthreading is equal to the single threaded score times the number of physical cores times 1.25. It is possible that Intel manages to increase the latter factor from 1.25x to 1.3x yielding a further 4% IPC improvement in multithreaded workloads.
• So best case scenario for the big cores is to have a combined multithreaded IPC improvement of 1.3*1.04=1.35x. That’s 35% IPC uplift (in MT Cinebench).
• The 11900K with adaptive boost does around 6200 in CB R20. Assuming same clock speeds from the 8 big cores of the 12900K we get from them 8370 points.
• The 8 little cores should be behaving like Skylake cores. Since they don’t have hyperthreading it will be around what an i7 9700K can do when clocked at 3.7GHz on all cores. Scaling based on the 9700K gives that the 8 little cores should contribute around 2870 points.
• Adding the two scores together gives us 11240 points, which is right on with the 11300 estimate that has surfaced recently in Chinese forums.