Lemire breaks down Ryzen's 50% two-year Geekbench gains: microarchitecture is the driver
On Sept 18, Daniel Lemire posted a five-tweet thread breaking down how AMD Ryzen 7 improved over 50% in Geekbench overall performance from 2022 to 2024, concluding the gains come mainly from microarchitecture rather than frequency.
Confirmed
- Ryzen 7's Geekbench overall performance improved by more than 50% between 2022 and 2024.
- Frequency only rose about 15%, insufficient to explain the gains.
- Transistor count grew about 50%, with most of the budget going to CPU cores rather than cache.
- Microarchitectural changes: maximum instructions per cycle rose from 6 to 8, caches grew larger, and the reorder buffer (for better instruction scheduling) improved.
- In SIMD data-parallel workloads, Zen 5 processors clearly lead other CPUs.
Unconfirmed
- Lemire noted that next-gen Zen 6 details are not yet public, hinting it "could be wild".
Why it matters
- Lemire's analysis shows recent CPU gains come mainly from microarchitectural evolution (IPC, caches, scheduling, SIMD) rather than frequency scaling, offering a clear framework for understanding processor progress.
2026-09-18 ~ 2026-09-18 · 6 related posts
Primary sources
- Zen 5's SIMD data parallelism leads the pack, and Zen 6 'could be wild' — lemire · 2026-09-18
- [source] AMD Ryzen 7 CPUs gained over 50% Geekbench performance from 2022 to 2024 — lemire · 2026-09-18
- Lemire: 15% frequency gain can't explain Ryzen's 50% performance jump — lemire · 2026-09-18
- Ryzen transistor count up ~50%, mostly in cores not cache — lemire · 2026-09-18
- Zen's IPC jump: from 6 to 8 instructions per cycle, bigger caches and ROB — lemire · 2026-09-18
- Zen 5 leads the pack on SIMD data parallelism — lemire · 2026-09-18