AMD EPYC 9255 vs Intel Xeon 658X Comparison

Both the AMD EPYC 9255 and the Intel Xeon 658X are 24-core server processors aimed at the same dense compute segment, and the benchmark data shows they are remarkably close overall. The AMD EPYC 9255 holds a slim 0.3% lead in average benchmark score (116,388 vs. 116,060), yet the two chips achieve that parity through very different architectural and performance profiles. This analysis breaks down the data to help you decide which processor fits specific workloads.

FAQ

Q: Which processor has the higher overall average benchmark score?

A: The AMD EPYC 9255 leads with an average benchmark score of 116,388, which is 0.3% higher than the Intel Xeon 658X's 116,060. Both sit at the 97th percentile among all CPUs.

Q: How do their core and thread counts compare?

A: They are identical: both have 24 cores and 48 threads. The difference lies in how each architecture uses those resources, not in the raw counts.

Q: Which CPU wins in single-threaded performance?

A: The Intel Xeon 658X takes the single-thread crown in Passmark, scoring 3,728 vs. the EPYC 9255's 3,655, a 2% lead. However, the EPYC 9255 wins every Cinebench single-core test by 3%.

Q: What is the biggest performance gap between the two in any benchmark?

A: The AMD EPYC 9255 dominates the Passmark physics test with a score of 9,740 versus Intel's 6,470, a 50.5% advantage. This is the largest delta in the entire head-to-head dataset.

Q: Which processor has a higher boost clock?

A: The Intel Xeon 658X boosts to 4.90 GHz, while the AMD EPYC 9255 boosts to 4.80 GHz. Intel also has a lower base clock at 3.00 GHz compared to AMD's 3.25 GHz.

Q: Are both processors currently in production?

A: Yes, both are listed as "Active" in production status. The AMD EPYC 9255 was released on 2024-10-09, while the Intel Xeon 658X has a release date of 2026-02-01.

Architecture Differences

The two processors could not be more different under the hood, despite matching core counts. The AMD EPYC 9255 uses the Zen 5 architecture on a 4nm TSMC process, built on the "Turin" codename. It features a multi-die design with four 70.6 mm² dies. The Intel Xeon 658X uses Granite Rapids architecture on Intel's 5nm process, built on two much larger 598 mm² dies.

Cache configurations diverge significantly. AMD provides 80 KB of L1 and 1 MB of L2 per core, with 128 MB of shared L3 cache. Intel counters with larger per-core caches — 112 KB L1 and 2 MB L2 — plus a larger 144 MB shared L3. This gives Intel a 16 MB advantage in L3, but AMD's smaller dies may offer better memory latency characteristics.

Memory subsystems also differ. AMD runs a twelve-channel DDR5 memory bus delivering 576.0 GB/s of bandwidth, while Intel uses an eight-channel configuration providing 409.6 GB/s. Both support DDR5 and ECC memory, and both offer Gen 5 PCIe with 128 lanes from the CPU. The AMD part is a 200W TDP chip, while the Intel part draws 250W. Both fit in different sockets: AMD uses Socket SP5, Intel uses Socket 4710.

The Verdict

The data shows two distinct philosophies. The AMD EPYC 9255 is the multi-threaded workhorse, winning 11 of 17 head-to-head benchmarks. It excels in integer math (16.1% ahead), physics (50.5% ahead), and random string sorting (25.4% ahead). It also wins every Cinebench test by 3%, including both multi-core and single-core variants. For general server workloads involving cryptography, physics simulation, or data sorting, the EPYC 9255 is the clear pick.

The Intel Xeon 658X is the specialist for floating-point and instruction-heavy workloads. It wins the Passmark floating point math test by 12.9%, extended instructions by 11.2%, and data compression by 4.1%. It also edges out AMD in Passmark single-thread (2% lead) and prime number finding (10.6% lead). If your workload is heavy on AVX-512-style instructions or compression, the Xeon 658X is the better choice.

For most mixed server workloads, the AMD EPYC 9255's wider benchmark dominance, lower 200W TDP, and higher memory bandwidth (576.0 GB/s vs. 409.6 GB/s) make it the safer default. The Intel part's strengths are real but narrower, and its 250W TDP means higher cooling demands. Both are capable 24-core parts, but the EPYC 9255's balanced performance profile gives it the edge for general-purpose deployment.

Specification Differences

| Specification | AMD EPYC 9255 | Intel Xeon 658X |

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

| Base Clock | 3.25 GHz | 3.00 GHz |

| Boost Clock | 4.80 GHz | 4.90 GHz |

| TDP | 200W | 250W |

| Socket | AMD Socket SP5 | Intel Socket 4710 |

| Architecture | Zen 5 | Granite Rapids |

| Process Node | 4 nm | 5 nm |

| Foundry | TSMC | Intel |

| Die Size | 4x 70.6 mm² | 2x 598 mm² |

| L1 Cache (per core) | 80 KB | 112 KB |

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

| L3 Cache (shared) | 128 MB | 144 MB |

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

| Memory Bandwidth | 576.0 GB/s | 409.6 GB/s |

| Multiplier Unlocked | No | Yes |

| Launch MSRP | $2495 | $1699 |

Head-to-Head Benchmarks

The Cinebench suite is a clean sweep for AMD. The EPYC 9255 wins all six Cinebench tests by exactly 3% each. In R23 multi-core, AMD scores 64,318 vs. Intel's 62,466; in R23 single-core, AMD scores 9,080 vs. 8,818. The R20 and R15 tests mirror this pattern, with AMD consistently ahead in both multi-core (27,013 vs. 26,235 in R20) and single-core (3,813 vs. 3,703 in R20).

Passmark tells a different story. AMD's biggest wins are in physics (9,740 vs. 6,470, a 50.5% advantage), random string sorting (129,202 vs. 103,028, up 25.4%), and integer math (306,442 vs. 263,995, up 16.1%). AMD also wins data encryption (59,668 vs. 52,357, up 14%) and Passmark multi-thread (76,580 vs. 73,490, up 4.2%).

Intel's counterattacks come in floating-point-heavy tasks. The Xeon 658X wins floating point math by 12.9% (210,480 vs. 183,367), extended instructions by 11.2% (84,626 vs. 75,185), and find prime numbers by 10.6% (649 vs. 580). It also takes data compression (1,062,062 vs. 1,018,904, up 4.1%) and Passmark single-thread (3,728 vs. 3,655, up 2%).

Where Each One Wins

The AMD EPYC 9255 wins for: Physics simulation, integer-heavy database workloads, data encryption, random string sorting, and sustained multi-threaded rendering. Its 50.5% physics lead and 25.4% sorting advantage are decisive for scientific computing and data processing tasks. The 16.1% integer math win also favors OLTP database workloads. All Cinebench wins by 3% suggest consistent performance across both single and multi-threaded rendering tasks. The lower 200W TDP also makes it easier to cool in dense server chassis.

The Intel Xeon 658X wins for: Floating-point math, extended instruction set workloads, data compression, and single-threaded Passmark tasks. The 12.9% floating point win and 11.2% extended instructions advantage point to HPC and scientific workloads that leverage AVX-512. The 4.1% data compression win helps with storage and archival workloads. The 2% single-thread Passmark lead, combined with the higher 4.90 GHz boost clock, gives it an edge in latency-sensitive single-threaded applications. The unlocked multiplier also allows for manual overclocking, which the AMD part does not support.

The 24-core, 48-thread parity means the choice comes down to workload character. AMD wins the majority of tests and offers more memory bandwidth and lower power draw. Intel wins the specific tests where its architecture's floating-point and instruction processing shines. For a general-purpose server, the data favors the EPYC 9255; for specialized math-heavy compute, the Xeon 658X is the better fit.