AMD EPYC 9135 vs Intel Xeon 638 Comparison

Head-to-Head Benchmarks

The benchmark data shows a clear overall victory for the AMD EPYC 9135, which wins 14 of the 17 head-to-head comparisons. The most decisive AMD advantage appears in random string sorting, where it scores 90,064 against the Intel Xeon 638's 74,318, a 21.2% margin. Physics performance also heavily favors the AMD part, with a score of 5,477 versus 4,704, a 16.4% lead. Data encryption shows an 11.8% advantage for AMD, scoring 40,295 against Intel's 36,030.

The Cinebench suite is uniformly favorable to the AMD EPYC 9135. Across all six Cinebench R15, R20, and R23 tests—both single-core and multi-core—the AMD part wins by margins between 4.1% and 4.2%. For instance, in Cinebench R23 multi-core, AMD scores 49,136 while Intel scores 47,202; in single-core R23, the scores are 6,936 and 6,663 respectively. This consistency suggests a fundamental architectural edge rather than a workload-specific quirk.

The Intel Xeon 638 does claim three wins, and they are worth examining closely. The largest Intel victory is in find prime numbers, where it scores 381 versus AMD's 292—a 23.4% advantage that is the single biggest delta in either direction across all tests. Floating point math also goes to Intel, 144,757 to 126,679, a 12.5% lead. Extended instructions narrowly favor Intel, 56,498 to 55,822, a 1.2% margin.

Other notable AMD wins include integer math at 9.8% (202,962 vs. 184,884) and data compression at 1.9% (739,277 vs. 725,818). The PassMark multi-thread score shows AMD ahead by 2.7% (57,170 vs. 55,651), while single-thread performance is effectively a tie: 3,672 vs. 3,670, a 0.1% delta. The average benchmark score reinforces the overall picture: AMD's 82,980 places it in the 96th percentile of all CPUs, while Intel's 80,723 sits in the 95th percentile.

Architecture Differences

The two processors diverge sharply in their underlying designs. The AMD EPYC 9135 is built on a 4 nm process at TSMC, using the Zen 5 architecture with the codename Turin, and is part of the EPYC 9005 series. It packs 16,630 million transistors across a dual-chiplet design with each die measuring 70.6 mm². In contrast, the Intel Xeon 638 uses Intel's 5 nm process with the Granite Rapids architecture and a single large die of 598 mm²; transistor count is not listed.

Clock speeds tell an interesting story. The AMD part has a higher base clock at 3.65 GHz versus Intel's 3.20 GHz, but Intel's boost clock reaches 4.80 GHz compared to AMD's 4.30 GHz. Both processors feature 16 cores and 32 threads, yet their cache hierarchies differ substantially. Intel allocates 112 KB of L1 cache per core and 2 MB of L2 per core, while AMD provides 80 KB of L1 and 1 MB of L2 per core. However, AMD's shared L3 cache is 64 MB, while Intel's is 72 MB—a smaller difference than the per-core figures suggest.

Memory architecture is a major differentiator. Both support DDR5 and ECC memory, but AMD implements a twelve-channel memory bus delivering 576.0 GB/s of bandwidth, while Intel uses a quad-channel bus limited to 204.8 GB/s. This nearly threefold bandwidth advantage for AMD is likely a significant factor in workloads that scale with memory throughput. PCIe connectivity also favors AMD, which offers 128 Gen 5 lanes (CPU only) versus Intel's 80 Gen 5 lanes.

The sockets are incompatible: AMD uses Socket SP5, while Intel uses Socket 4710. The Intel part is multiplier-unlocked, whereas the AMD part is not. Release dates differ by over a year, with AMD launching on 2024-10-09 and Intel on 2026-02-01; both are listed as Active in production and neither includes integrated graphics.

Where Each One Wins

The AMD EPYC 9135 is the stronger choice for encryption-heavy and memory-bandwidth-sensitive workloads. Its 11.8% lead in data encryption and 21.2% lead in random string sorting point to strengths in data manipulation and cryptographic tasks. The 16.4% physics advantage suggests better performance in simulation and collision-detection style computations. Integer math also favors AMD by 9.8%, which is relevant for general server-side processing and database operations. The consistent 4.1% Cinebench wins across both single-core and multi-core tests indicate well-rounded performance in rendering and CPU-bound productivity workloads.

The Intel Xeon 638 wins in three specific areas that hint at a different strength profile. The 23.4% lead in find prime numbers and 12.5% lead in floating point math show that Intel's architecture handles certain mathematical workloads more efficiently. These are classic compute-heavy operations that often benefit from higher boost clocks—Intel's 4.80 GHz versus AMD's 4.30 GHz may play a role here. The extended instructions win, while narrow at 1.2%, suggests Intel has an edge in workloads that leverage newer or specific instruction set extensions.

Data compression is nearly a wash, with AMD ahead by just 1.9%. Single-thread performance is statistically tied. The PassMark multithread score also lands close, with AMD ahead by 2.7%. This granularity matters: a workload dominated by prime number generation or heavy floating-point math may see Intel come out ahead, but most mixed server workloads would likely favor AMD based on the broader benchmark distribution.

The Verdict

The benchmark data presents a straightforward recommendation for most buyers. The AMD EPYC 9135 wins 14 of 17 comparisons, including all Cinebench tests, and offers substantially higher memory bandwidth (576.0 GB/s vs. 204.8 GB/s) plus more PCIe lanes (128 vs. 80). Its average benchmark score of 82,980 sits in the 96th percentile, edging out Intel's 80,723 in the 95th percentile. For users prioritizing encryption, sorting, integer math, physics, and general rendering tasks, the AMD part is the clear choice.

The Intel Xeon 638 should be considered by users with specific workloads that match its three wins. The 23.4% advantage in find prime numbers and 12.5% lead in floating point math are substantial margins that would translate directly to speedups in those niche areas. Its higher boost clock of 4.80 GHz and larger L3 cache of 72 MB may also appeal to users running latency-sensitive single-threaded tasks, even though the single-thread benchmark shows only a 0.1% difference. The lower launch MSRP of $899 versus AMD's $1,214 may factor into procurement decisions, though the performance data shows AMD leading in the majority of tests.

Neither processor is a poor choice; both are active server/workstation parts with strong overall performance. The decision hinges on whether the user's workload resembles the AMD-favorable majority (encryption, sorting, integer math, Cinebench-style rendering) or the Intel-favorable minority (prime number computation, floating-point math, extended instructions). For general-purpose server deployment, AMD's broader win count and higher average score make it the safer recommendation.

FAQ

Q: Which processor has a higher average benchmark score?

A: The AMD EPYC 9135 has an average benchmark score of 82,980, compared to the Intel Xeon 638's 80,723.

Q: How many benchmark comparisons does each processor win?

A: The AMD EPYC 9135 wins 14 of the 17 head-to-head comparisons, while the Intel Xeon 638 wins 3.

Q: What is the largest single benchmark margin in either direction?

A: The Intel Xeon 638's 23.4% lead in the PassMark find prime numbers test is the largest delta, with Intel scoring 381 against AMD's 292.

Q: Do both processors support the same memory type and ECC?

A: Yes, both support DDR5 memory and ECC, but AMD uses a twelve-channel bus with 576.0 GB/s bandwidth, while Intel uses a quad-channel bus with 204.8 GB/s.

Q: What are the boost clocks of each processor?

A: The Intel Xeon 638 boosts to 4.80 GHz, while the AMD EPYC 9135 boosts to 4.30 GHz. The AMD part has a higher base clock at 3.65 GHz versus Intel's 3.20 GHz.

Q: Which processor has more PCIe lanes?

A: The AMD EPYC 9135 offers 128 Gen 5 lanes (CPU only), while the Intel Xeon 638 offers 80 Gen 5 lanes.

Specification Differences

| Specification | AMD EPYC 9135 | Intel Xeon 638 |

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

| Architecture | Zen 5 | Granite Rapids |

| Codename | Turin | Granite Rapids |

| Generation | EPYC (Zen 5 (Turin)) | Xeon 600 (Granite Rapids-WS) |

| Process Node | 4 nm (TSMC) | 5 nm (Intel) |

| Base Clock | 3.65 GHz | 3.20 GHz |

| Boost Clock | 4.30 GHz | 4.80 GHz |

| TDP | 200 W | 180 W |

| Socket | AMD Socket SP5 | Intel Socket 4710 |

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

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

| L3 Cache | 64 MB (shared) | 72 MB (shared) |

| Memory Bus | Twelve-channel | Quad-channel |

| Memory Bandwidth | 576.0 GB/s | 204.8 GB/s |

| PCIe | Gen 5, 128 Lanes (CPU only) | Gen 5, 80 Lanes (CPU only) |

| Die Size | 2x 70.6 mm² | 598 mm² |

| Transistors | 16,630 million | Not listed |

| Multiplier Unlocked | No | Yes |

| Launch MSRP | $1214 | $899 |

| Release Date | 2024-10-09 | 2026-02-01 |

| Part Number | 100-000001150 | SA2DN |