AMD EPYC 4564P vs AMD EPYC 4565P Comparison

FAQ

Q: Which processor has the higher average benchmark score?

A: The AMD EPYC 4565P leads with an average benchmark score of 95,764, while the AMD EPYC 4564P scores 95,183. The 4565P holds a 0.6% advantage over the 4564P in the nearestRivals comparison.

Q: How do the two CPUs compare in Cinebench multi-core tests?

A: The EPYC 4564P wins every Cinebench multi-core test. In Cinebench R23 multi-core, it scores 55,592 versus 54,405 for the 4565P, a 2.1% lead. The same 2.1% delta appears in R15 (5,603 vs 5,484) and R20 (23,348 vs 22,850).

Q: Which processor is better for single-threaded PassMark performance?

A: The EPYC 4565P wins PassMark single-thread by a wide margin, scoring 4,712 versus 4,292 for the 4564P — a 9.8% advantage. This is the largest single-thread win for the 4565P in the head-to-head data.

Q: Are there any tests where the 4564P loses by more than 10%?

A: Yes. The 4564P trails the 4565P by 9.6% in PassMark integer math (250,683 vs 228,761) and by 7.8% in floating-point math (152,003 vs 141,017). It also loses by 1.9% in extended instructions (64,345 vs 63,136).

Q: Does the 4565P win any benchmark where the 4564P is stronger in Cinebench?

A: The 4565P wins PassMark data compression (860,786 vs 858,105, +0.3%), extended instructions, floating-point math, integer math, and single-thread tests. The 4564P dominates Cinebench across all six tests and wins nine other PassMark tests.

Q: What is the release date difference between the two?

A: The EPYC 4564P launched on 2024-05-20, while the EPYC 4565P launched on 2025-05-12 — roughly one year later. Both are currently listed with Active production status.

Architecture Differences

The EPYC 4565P is built on Zen 5 architecture (codename Grado), while the EPYC 4564P uses Zen 4 (codename Raphael). This architectural gap is the primary driver of their different benchmark profiles. The 4565P is fabricated on a 4 nm process at TSMC, versus 5 nm for the 4564P, which explains the transistor count difference: 16,630 million for the 4565P versus 13,140 million for the 4564P. Die sizes are nearly identical at 2x 70.6 mm² for the 4565P and 2x 71 mm² for the 4564P.

Cache layouts show a meaningful divergence in L1. The 4565P provides 80 KB of L1 per core, while the 4564P offers 64 KB per core. Both have 1 MB L2 per core and 64 MB shared L3. That larger L1 cache on the 4565P likely contributes to its strong PassMark integer and floating-point performance, where it beats the 4564P by 9.6% and 7.8% respectively.

Memory bandwidth also differs: the 4565P delivers 89.6 GB/s versus 83.2 GB/s for the 4564P, though both support DDR5 with dual-channel memory and ECC. PCIe lane counts are another differentiator — the 4565P offers 24 Gen 5 lanes (CPU only), while the 4564P offers 28 Gen 5 lanes. Both CPUs include Radeon Graphics as integrated graphics, target the Server/Workstation market, and use the same AMD Socket AM5.

Clock speeds are nearly identical on boost (5.70 GHz for both), but the base clocks differ: the 4564P runs at 4.50 GHz base, 200 MHz higher than the 4565P's 4.30 GHz. Both have a 170 W TDP and locked multipliers. The 4565P belongs to the EPYC 4005 series, while the 4564P is in the EPYC 4004 series.

Head-to-Head Benchmarks

The benchmark data tells a clear story: the EPYC 4564P wins 11 of 17 head-to-head tests, while the 4565P wins 6. However, the magnitude of those wins matters more than the raw count. The 4564P's victories are largely in Cinebench, where it sweeps all six tests with a consistent 2.1–2.2% margin. In Cinebench R23 multi-core, the 4564P scores 55,592 versus 54,405; in R23 single-core it scores 7,848 versus 7,680. The same pattern holds in R15 (5,603 vs 5,484 multi; 791 vs 774 single) and R20 (23,348 vs 22,850 multi; 3,296 vs 3,225 single).

The 4564P also wins PassMark multithread (64,357 vs 63,474, +1.4%), physics (3,392 vs 2,976, +12.3%), data encryption (50,708 vs 48,268, +4.8%), find prime numbers (365 vs 294, +19.5%), and random string sorting (103,202 vs 81,318, +21.2%). The random string sorting result is the single largest margin in either direction — the 4564P is more than a fifth faster in that workload.

The 4565P counters with substantial wins in compute-heavy PassMark tests. Integer math favors the 4565P by 9.6% (250,683 vs 228,761), floating-point math by 7.8% (152,003 vs 141,017), and single-thread by 9.8% (4,712 vs 4,292). Extended instructions go to the 4565P by 1.9% (64,345 vs 63,136), and data compression is a narrow 0.3% win (860,786 vs 858,105).

The split is telling: the 4564P's wins are concentrated in Cinebench and memory/sorting-style workloads, while the 4565P excels at raw arithmetic and single-threaded PassMark tests. The 4565P's 4 nm process and larger L1 cache appear to boost its ALU/FPU performance, but the 4564P's higher base clock (4.50 vs 4.30 GHz) gives it the edge in latency-sensitive and multi-threaded Cinebench tasks.

Specification Differences

| Specification | AMD EPYC 4565P | AMD EPYC 4564P |

|---|---|---|

| Series | EPYC 4005 series | EPYC 4004 series |

| Architecture | Zen 5 | Zen 4 |

| Codename | Grado | Raphael |

| Process Node | 4 nm | 5 nm |

| Transistors | 16,630 million | 13,140 million |

| Die Size | 2x 70.6 mm² | 2x 71 mm² |

| L1 Cache | 80 KB (per core) | 64 KB (per core) |

| Base Clock | 4.30 GHz | 4.50 GHz |

| Memory Bandwidth | 89.6 GB/s | 83.2 GB/s |

| PCIe Lanes | 24 (Gen 5, CPU only) | 28 (Gen 5, CPU only) |

| Release Date | 2025-05-12 | 2024-05-20 |

| Launch MSRP | $589 | $699 |

The two CPUs share identical cores (16), threads (32), L2 cache (1 MB per core), L3 cache (64 MB), boost clock (5.70 GHz), TDP (170 W), socket (AM5), memory support (DDR5), memory bus (dual-channel), ECC support, integrated graphics (Radeon), market segment, and production status. The 4564P has a higher base clock by 200 MHz, more PCIe lanes (28 vs 24), and a slightly larger die (71 mm² vs 70.6 mm²). The 4565P counters with a smaller process node (4 nm vs 5 nm), more transistors (16,630M vs 13,140M), larger L1 cache (80 KB vs 64 KB), higher memory bandwidth (89.6 vs 83.2 GB/s), and a lower launch MSRP ($589 vs $699).

Where Each One Wins

AMD EPYC 4565P wins in scenarios that stress raw arithmetic throughput and single-threaded responsiveness. Its PassMark integer math score of 250,683 is 9.6% higher than the 4564P's 228,761, and floating-point math follows at 152,003 versus 141,017 (+7.8%). The 9.8% single-thread advantage (4,712 vs 4,292) makes it the better choice for lightly threaded workloads where per-core efficiency matters. Data compression (860,786 vs 858,105) and extended instructions (64,345 vs 63,136) add to its profile. The 4565P also offers more memory bandwidth (89.6 vs 83.2 GB/s), which can benefit memory-bound applications.

AMD EPYC 4564P wins in Cinebench rendering workloads across the board, with a consistent 2.1–2.2% margin in both single and multi-core tests. Its PassMark multithread score (64,357 vs 63,474) and physics score (3,392 vs 2,976, +12.3%) show strength in multi-threaded, physics-heavy simulations. The 4564P is dramatically faster at random string sorting (103,202 vs 81,318, +21.2%) and find prime numbers (365 vs 294, +19.5%), making it the pick for sorting, hashing, and prime-calculation workloads. Data encryption also favors the 4564P (50,708 vs 48,268, +4.8%). Its higher base clock (4.50 GHz) and additional PCIe lanes (28 vs 24) provide practical advantages for I/O-heavy server tasks.

The Verdict

The benchmark results indicate a clear trade-off between the two processors, and the choice depends entirely on workload mix. The EPYC 4564P is the overall benchmark winner with 11 head-to-head wins versus 6 for the 4565P. Its Cinebench dominance is comprehensive, and its massive leads in random string sorting (+21.2%) and prime numbers (+19.5%) are not marginal — they are decisive. For users running rendering, physics simulation, sorting, or encryption workloads, the 4564P is the data-backed pick.

The EPYC 4565P, however, is the stronger choice for arithmetic-heavy and single-threaded tasks. Its 9.6% integer math lead and 9.8% single-thread advantage suggest it extracts more per-core performance from its Zen 5 architecture and 4 nm process. The 4565P also has a lower launch MSRP ($589 vs $699), higher memory bandwidth, and more transistors, which could matter for future-proofing. But the benchmark data does not show the 4565P outperforming the 4564P in any Cinebench test — those all go to the 4564P.

The average benchmark scores put the 4565P ahead by 0.6% (95,764 vs 95,183), and both CPUs sit at the 96th percentile among all processors. The nearest rivals for the 4565P include the AMD EPYC 9175F (0.2% ahead), AMD Ryzen 9 PRO 9945 (0.3% behind), and AMD Ryzen 9 9955HX3D (1.7% behind). For the 4564P, the nearest rivals are the 9175F (0.5% ahead), the 4565P (0.6% ahead), and the Ryzen 9 PRO 9945 (0.9% ahead). The Intel Xeon 6520P trails the 4564P by 1.5%.

For most server and workstation buyers, the 4564P's broader benchmark wins and higher base clock make it the safer default. The 4565P is the pick for workloads that thrive on integer math, floating-point throughput, and single-threaded PassMark performance — and for buyers who value the newer 4 nm process and lower launch MSRP. The data does not support a blanket recommendation for either; it supports matching the CPU to the workload.