With smaller caches and especially L1 you typically want higher associativity. With 32KB L1 at 64B line size, you can only cache 512 lines. Grouping them in la

With smaller caches and especially L1 you typically want higher associativity. With 32KB L1 at 64B line size, you can only cache 512 lines. Grouping them in larger buckets means less spillage as hot lines randomly end up in the same bucket.