Intel will once again be taking a leaf out of AMD’s rulebook for its upcoming 13th Gen Raptor Lake processors. Unlike Alder Lake, Raptor will be a refresh of sorts with beefed-up core counts, cache, and optimized latencies. The core architecture and the platform will, more or less, remain unchanged. It has already been confirmed that Raptor Lake will double the number of efficiency cores to 16 (24 overall) and increase the cache sizes across the board. A CPU-Z screenshot from @OneRaichu indicates that Intel will be going all-in with the cache, packing as much as 68MB for the Core i9-13900K. That’s nearly the same as the Ryzen 9 5950X (72MB).
This includes 36MB of L3 cache that is shared between the eight Raptor Cove (P) and sixteen Gracemont (E) cores. The remaining 32MB L2 is divided equally between the two, wherein the P cores get 2MB each, and the E cores share 4MB chunks per quad cluster. If reports about AMD’s Ryzen 7000 processors are to be believed, then Zen 4 will have 1MB of L2 cache per core, half as much as Raptor Cove, and twice as much as Zen 3. Therefore, the two chipmakers will be offering roughly the same amount of cache on their CPUs this coming generation. AMD will be concentrating most of it on L3 while Intel will be increasing both rather evenly. However, it’s important to note that neither is better as such. Different architectures have different cache requirements and bottlenecks but it’s always interesting to see an underdog influence a much larger opponent.
In related news, the same source states that Meteor Lake will essentially be a “tic” featuring a new die shrink and power optimizations with minimal changes to the core width. The scheduler, execution ports, and decoders will only get minor optimizations. The ROB size may change as is the tradition with most node shrinks but nothing too drastic.