tl;dr: AMD's Zen 6 builds on Zen 5 with a faster branch predictor, dual-decoder utilization by a single thread, larger OoO structures, doubled VNNI throughput, and 4x 256-bit vector loads per cycle, delivering roughly 18-21% single-threaded SPEC 2026 gains over Zen 5. Notably, Zen 6 and Zen 6c now diverge architecturally for the first time: Zen 6 is tuned for high clocks (12-core CCDs, 48MB L3), while Zen 6c has larger OoO buffers to hide latency (32-core CCDs, 128MB L3), signaling AMD's shift toward workload-specific core specialization.