AMD EPYC 9684X vs Intel Xeon 6747P Comparison

The AMD EPYC 9684X and Intel Xeon 6747P represent two distinct approaches to high-core-count server processing, and the benchmark data shows a decisive split between multi-threaded dominance and single-thread efficiency. Across the head-to-head benchmark suite, the AMD EPYC 9684X wins 12 of 14 tests, while the Intel Xeon 6747P takes the remaining 2, which are both single-thread Passmark results. The average benchmark scores reflect this gap: the AMD EPYC 9684X posts an average score of 266,914 against 238,263 for the Intel Xeon 6747P, placing both in the 99th percentile of all CPUs.

Head-to-Head Benchmarks

The most striking pattern in the head-to-head data is the consistency of the AMD EPYC 9684X’s multi-threaded victories. In Cinebench R15, R20, and R23 multicore tests, the AMD part scores 10,418, 43,409, and 103,355 respectively, compared to 8,712, 36,301, and 86,432 for the Intel Xeon 6747P. Each of these represents a 19.6% advantage for AMD, a uniform margin that suggests a fundamental throughput advantage rather than a workload-specific quirk. The Passmark multithread test tells the same story, with the AMD EPYC 9684X scoring 121,595 versus 101,685, again a 19.6% lead.

The gap widens considerably in certain integer and encryption workloads. The AMD EPYC 9684X leads by 80.2% in Passmark integer math (844,145 vs 468,518) and by 75.5% in find prime numbers (2,020 vs 1,151). Data encryption shows a 96.6% advantage for AMD (178,453 vs 90,789), while random string sorting sees the largest relative gap at 93.5% (349,126 vs 180,382). These are not marginal wins; they indicate that the AMD processor has more than double the throughput in some operations. Floating-point math also favors AMD by 29% (472,174 vs 365,904), and physics simulation by 84.3% (24,686 vs 13,398). Data compression shows a 47.2% lead (2,698,807 vs 1,833,378), and extended instructions favor AMD by 24.8% (177,956 vs 142,557).

The Intel Xeon 6747P’s only victories come in the Passmark single-thread tests, where it scores 3,236 against AMD’s 2,891. That 10.7% advantage is meaningful for lightly threaded tasks, but it is the sole area where Intel wins. Notably, the head-to-head data does not include Cinebench single-core results for the Intel part, so the comparison rests entirely on those two Passmark single-thread entries. The overall picture is one of AMD dominating every multi-threaded and most single-threaded workloads by wide margins, with Intel taking a narrower single-thread win.

Architecture Differences

The architectural divide between these two processors is substantial. The AMD EPYC 9684X is built on the Zen 4 architecture with the Genoa-X codename, part of the EPYC 9004 series, and fabricated on a 5 nm process by TSMC. It uses 12 chiplets, each 72 mm², totaling 135,240 million transistors. The Intel Xeon 6747P uses the Granite Rapids architecture, also on a 5 nm process but fabricated by Intel, with a die size of 2x 598 mm² and no transistor count listed in the data.

Cache configurations differ sharply. The AMD EPYC 9684X has 64 KB of L1 and 1 MB of L2 per core, with an enormous 1152 MB of shared L3 cache. The Intel Xeon 6747P has 112 KB of L1 and 2 MB of L2 per core, but only 288 MB of shared L3. That 4x difference in L3 capacity is a key factor in the AMD part’s performance in cache-sensitive workloads like data compression and encryption. The AMD processor also supports twelve-channel DDR5 memory with a bandwidth of 460.8 GB/s, while Intel uses eight-channel DDR5 at 409.6 GB/s. Both support ECC memory.

PCIe connectivity favors AMD as well: the EPYC 9684X provides 128 Gen 5 lanes, while the Xeon 6747P provides 88 Gen 5 lanes. The AMD part uses Socket SP5, the Intel part Socket 4710. Core counts diverge sharply, with AMD at 96 cores and 192 threads versus Intel at 48 cores and 96 threads. Base and boost clocks are higher on Intel (2.70 GHz / 3.90 GHz) than AMD (2.55 GHz / 3.70 GHz), which explains the single-thread win. The Intel part has no integrated graphics listed as "N/A", while AMD’s is null. The AMD EPYC 9684X launched on 2023-06-12 with a launch MSRP of $14756, while the Intel Xeon 6747P launched on 2025-02-23 with a launch MSRP of $6497. Both are active in production and have locked multipliers.

FAQ

Q: Which processor wins the majority of benchmark tests?

A: The AMD EPYC 9684X wins 12 of 14 head-to-head tests, including all Cinebench multicore and Passmark multithread benchmarks. The Intel Xeon 6747P wins only 2 tests, both Passmark single-thread.

Q: How large is the multi-threaded performance gap?

A: In Cinebench R15, R20, and R23 multicore, the AMD EPYC 9684X leads by exactly 19.6% in each test. The Passmark multithread score shows the same 19.6% margin.

Q: Does the Intel Xeon 6747P have any meaningful advantage?

A: Yes, in Passmark single-thread tests, the Intel Xeon 6747P scores 3,236 versus 2,891 for AMD, a 10.7% advantage. This is the only category where Intel wins.

Q: What explains the AMD part’s dominance in integer and encryption workloads?

A: The AMD EPYC 9684X has 1152 MB of shared L3 cache versus 288 MB on Intel, and 96 cores versus 48. This combination likely drives the 80.2% lead in integer math and 96.6% lead in data encryption.

Q: How do memory systems compare between the two?

A: AMD uses twelve-channel DDR5 with 460.8 GB/s bandwidth, while Intel uses eight-channel DDR5 with 409.6 GB/s. Both support ECC memory.

Q: Are there differences in PCIe lane availability?

A: Yes, the AMD EPYC 9684X provides 128 Gen 5 lanes, while the Intel Xeon 6747P provides 88 Gen 5 lanes, both CPU-only.

The Verdict

The data points to a clear verdict for workloads that scale with core count and cache. The AMD EPYC 9684X offers roughly 20% higher multi-threaded performance across Cinebench and Passmark multithread, with even larger margins in specific workloads like integer math (80.2%), physics (84.3%), and random string sorting (93.5%). The AMD processor’s 96 cores, 192 threads, and 1152 MB of L3 cache make it the superior choice for database workloads, scientific computing, and any application that benefits from massive parallel throughput. The 12 wins out of 14 tests are decisive.

The Intel Xeon 6747P is the better option only when single-thread performance is the primary constraint. Its 10.7% lead in Passmark single-thread scores, combined with higher base and boost clocks (2.70 GHz and 3.90 GHz versus 2.55 GHz and 3.70 GHz), makes it suitable for lightly threaded applications or latency-sensitive tasks where a single core’s speed matters more than aggregate throughput. However, the Intel part’s 48 cores and 288 MB of L3 cache place it at a structural disadvantage in any multi-threaded scenario.

From the data, the choice hinges on workload characteristics: pick the AMD EPYC 9684X for almost everything except single-thread-bound processes. The Intel Xeon 6747P’s single-thread win does not compensate for the 19.6% multi-thread deficit, the 47.2% compression gap, or the 96.6% encryption gap. The benchmark results are unambiguous in their overall assessment.

Specification Differences

| Specification | AMD EPYC 9684X | Intel Xeon 6747P |

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

| Cores | 96 | 48 |

| Threads | 192 | 96 |

| Base Clock | 2.55 GHz | 2.70 GHz |

| Boost Clock | 3.70 GHz | 3.90 GHz |

| TDP | 400 W | 330 W |

| Socket | AMD Socket SP5 | Intel Socket 4710 |

| Architecture | Zen 4 | Granite Rapids |

| Codename | Genoa-X | Granite Rapids |

| Process Node | 5 nm | 5 nm |

| Foundry | TSMC | Intel |

| Transistors | 135,240 million | N/A |

| Die Size | 12x 72 mm² | 2x 598 mm² |

| L1 Cache | 64 KB (per core) | 112 KB (per core) |

| L2 Cache | 1 MB (per core) | 2 MB (per core) |

| L3 Cache | 1152 MB (shared) | 288 MB (shared) |

| Memory Bus | Twelve-channel | Eight-channel |

| Memory Bandwidth | 460.8 GB/s | 409.6 GB/s |

| PCIe | Gen 5, 128 Lanes | Gen 5, 88 Lanes |

| Integrated Graphics | N/A | N/A |

| Release Date | 2023-06-12 | 2025-02-23 |

| Part Number | 100-100000892 | SRVEZ |

Where Each One Wins

The AMD EPYC 9684X wins in every multi-threaded category tested. It is the clear choice for Cinebench multicore workloads, with a 19.6% lead across R15, R20, and R23. For data-heavy operations, it wins decisively: data compression by 47.2%, data encryption by 96.6%, integer math by 80.2%, random string sorting by 93.5%, and find prime numbers by 75.5%. Floating-point math favors AMD by 29%, physics by 84.3%, and extended instructions by 24.8%. The Passmark multithread test confirms the pattern with a 19.6% advantage. Any workload that uses multiple cores, large data sets, or heavy computation leans toward the AMD part.

The Intel Xeon 6747P wins only in Passmark single-thread tests, with a 10.7% margin over AMD. This advantage applies to workloads that cannot use more than one or two cores effectively, such as certain legacy applications, serialized scripting, or low-latency single-threaded services. The Intel part’s higher base and boost clocks support this profile. For everything else, the benchmark data indicates the AMD EPYC 9684X is the stronger performer, despite its higher TDP of 400 W versus 330 W for Intel. The AMD processor’s 12 benchmark wins versus 2 for Intel make the use-case split straightforward: choose Intel only for single-thread-bound tasks, choose AMD for all multi-threaded and throughput-intensive workloads.