AMD EPYC 9565 vs AMD EPYC 9684X Comparison

The AMD EPYC 9565 and AMD EPYC 9684X represent two distinct approaches to high-core-count server computing. The 9565, built on the newer Zen 5 architecture, wins the majority of benchmark comparisons, while the 9684X counters with a massive L3 cache advantage. The data shows a clear split: the 9565 dominates compute-heavy workloads and single-threaded tasks, whereas the 9684X leads in memory-bandwidth-sensitive and specialized workloads. The 9565 edges out the 9684X in average benchmark score (285,471 vs 266,914), a 7% advantage, but the 9684X retains a critical role for specific data-intensive applications.

The Verdict

The data indicates the AMD EPYC 9565 is the superior all-round performer for most server workloads. It wins 12 of 17 head-to-head comparisons, with a particularly decisive margin in single-threaded tasks (27.8% ahead in PassMark single-thread) and consistent 11.2% advantages across all Cinebench tests. The 9565 is the recommended choice for general-purpose compute, virtualized environments, and workloads that benefit from higher clock speeds and newer architecture efficiency.

The AMD EPYC 9684X is the specialist choice, not the generalist. Its 1152 MB of shared L3 cache — three times that of the 9565 — delivers wins in data encryption (20.5% ahead), random string sorting (16.4% ahead), and physics simulations (26.9% ahead). For workloads with large, frequently-accessed datasets that fit within cache, the 9684X's architecture is demonstrably superior. The 9684X also holds a 14.9% lead in PassMark integer math and a 4.4% lead in data compression.

The average benchmark score delta favors the 9565 (7% higher), and it also holds a 6.5% advantage in the 9684X's own nearest-rival data. The 9565's 99th percentile ranking among all CPUs is matched by the 9684X, but the 9565 achieves this with 24 fewer cores and 48 fewer threads. This indicates the 9565 is more efficient per core, making it the better choice for software licensing models based on core counts.

Architecture Differences

The two processors share the same AMD Socket SP5 platform, but their internal architectures diverge significantly. The 9565 uses the Zen 5 architecture on a 4 nm TSMC process, while the 9684X uses Zen 4 on a 5 nm process. This process node advantage contributes to the 9565's higher clock speeds: a 3.15 GHz base clock versus 2.55 GHz, and a 4.30 GHz boost versus 3.70 GHz.

Core counts favor the 9684X, which packs 96 cores and 192 threads, compared to the 9565's 72 cores and 144 threads. However, the 9565 compensates with architectural efficiency. The 9565 features 80 KB of L1 cache per core versus 64 KB on the 9684X, and identical 1 MB L2 cache per core. The L3 cache distinction is stark: the 9684X offers 1152 MB of shared L3 cache, while the 9565 provides 384 MB. This 3x difference is the defining architectural feature of the 9684X.

Transistor counts and die sizes differ substantially. The 9684X integrates 135,240 million transistors across 12x 72 mm² dies, while the 9565 uses 99,780 million transistors across 12x 70.6 mm² dies. The 9684X's larger transistor count reflects its massive L3 cache implementation, while the 9565's smaller die size per chiplet demonstrates the density improvement from the 4 nm process.

Both support DDR5 memory with twelve-channel buses, but memory bandwidth differs: the 9565 offers 576.0 GB/s versus 460.8 GB/s for the 9684X. Both provide PCIe Gen 5 with 128 CPU-only lanes and support ECC memory. The 9565 was released on 2024-10-09, while the 9684X debuted earlier on 2023-06-12.

FAQ

Q: Which processor has higher single-threaded performance?

A: The AMD EPYC 9565 is decisively ahead, winning PassMark single-thread by 27.8% (3696 vs 2891) and Cinebench R23 single-core by 11.2% (16226 vs 14591). Its higher boost clock of 4.30 GHz versus 3.70 GHz and newer Zen 5 architecture explain this gap.

Q: Does the 9684X's larger cache translate to wins in all cache-sensitive workloads?

A: No. The 9684X wins in data encryption, random string sorting, and physics, but the 9565 still wins PassMark extended instructions by 17.8% and floating-point math by 16.4%. The cache advantage helps specific patterns, not all memory-heavy tasks.

Q: How do the core counts affect multi-threaded performance?

A: Despite having 24 fewer cores and 48 fewer threads, the 9565 wins all multi-threaded Cinebench tests by 11.2%. Its higher per-core performance more than offsets the core-count deficit. The 9565's PassMark multithread score (135,221) also beats the 9684X (121,595).

Q: What is the memory bandwidth difference?

A: The 9565 provides 576.0 GB/s of memory bandwidth versus 460.8 GB/s for the 9684X, a 25% advantage. Both use DDR5 in a twelve-channel configuration, but the 9565's newer memory controller extracts higher throughput.

Q: Are these processors in the same market segment?

A: Yes, both target the Server/Workstation segment, use AMD Socket SP5, and carry a 400 W TDP. Both are currently in Active production status and rank in the 99th percentile among all CPUs.

Q: Which processor has a higher average benchmark score?

A: The 9565 averages 285,471 across all benchmarks, which is 7% higher than the 9684X's 266,914. This aggregate metric favors the 9565, though individual workload results vary widely.

Specification Differences

The following table lists only fields where the two processors differ:

| Specification | AMD EPYC 9565 | AMD EPYC 9684X |

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

| Cores | 72 | 96 |

| Threads | 144 | 192 |

| Base Clock | 3.15 GHz | 2.55 GHz |

| Boost Clock | 4.30 GHz | 3.70 GHz |

| Architecture | Zen 5 | Zen 4 |

| Codename | Turin | Genoa-X |

| Generation | EPYC (Zen 5 (Turin)) | EPYC (Zen 4 (Genoa)) |

| Process Node | 4 nm | 5 nm |

| Transistors | 99,780 million | 135,240 million |

| Die Size | 12x 70.6 mm² | 12x 72 mm² |

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

| L3 Cache | 384 MB (shared) | 1152 MB (shared) |

| Memory Bandwidth | 576.0 GB/s | 460.8 GB/s |

| Release Date | 2024-10-09 | 2023-06-12 |

| Launch MSRP | $10486 | $14756 |

| Part Number | 100-000001447 | 100-000100000892 |

Head-to-Head Benchmarks

The 9565 sweeps all six Cinebench tests with a uniform 11.2% margin. In Cinebench R15 multi-core, the 9565 scores 11585 against 10418; in single-core, 1635 vs 1470. The pattern repeats in R20 (48273 vs 43409 multi; 6814 vs 6128 single) and R23 (114937 vs 103355 multi; 16226 vs 14591 single). This consistency indicates a fundamental architectural advantage across both integer and floating-point rendering workloads.

PassMark results reveal a more nuanced picture. The 9565's biggest win is in single-thread performance, scoring 3696 versus 2891, a 27.8% lead. It also dominates extended instructions (209595 vs 177956, +17.8%), find prime numbers (2422 vs 2020, +19.9%), and floating-point math (549422 vs 472174, +16.4%). In multi-thread performance, the 9565 wins 135221 to 121595, an 11.2% margin matching the Cinebench results.

The 9684X's victories are concentrated in areas that exploit its 1152 MB L3 cache. The largest win is physics, where it scores 24686 against 18036, a 26.9% advantage. Data encryption shows a 20.5% lead (178453 vs 141936), and random string sorting follows with 16.4% (349126 vs 291941). The 9684X also wins integer math by 14.9% (844145 vs 717948) and data compression by 4.4% (2698807 vs 2579631).

The benchmark data shows the 9565 wins 12 of 17 tests, with the 9684X taking 5. The 9684X's wins are substantial where they occur, but they do not offset the 9565's broad dominance. The 9565's average benchmark score of 285,471 places it 7% above the 9684X, and it also leads in the nearest-rival comparison, being 2.1% behind the EPYC 9555P while the 9684X sits 8.5% behind the same rival. The 9684X's 6.5% deficit to the 9565 in its own rival list confirms the 9565's overall superiority in aggregate performance metrics.