AMD EPYC 9755 vs Intel Xeon 6960P Comparison

The AMD EPYC 9755 and Intel Xeon 6960P are both flagship server processors, yet benchmark data shows a decisive performance gap. The EPYC 9755 wins all 14 head-to-head comparisons, with margins ranging from 6.6% in single-threaded tests to 113% in integer math. This analysis breaks down the architectural and specification differences that drive those results, and what they mean for server workloads.

The Verdict

The data is unambiguous: the AMD EPYC 9755 outperforms the Intel Xeon 6960P in every single benchmark recorded. Across Cinebench R15, R20, and R23 multi-core tests, the EPYC 9755 consistently leads by 27.3%, with scores of 14250, 59378, and 141378 versus the Xeon’s 11194, 46645, and 111060 respectively. That consistency suggests a fundamental throughput advantage rather than workload-specific tuning.

The largest victory comes in PassMark integer math, where the EPYC 9755 scores 1549946 against the Xeon’s 727750 — a 113% advantage. Floating-point math also shows a massive 75% gap (922900 vs 527473), and data encryption is 75.9% ahead (284927 vs 162013). These are not marginal differences; they indicate the EPYC 9755 delivers roughly double the raw compute throughput in certain integer-heavy scenarios.

For single-threaded performance, the EPYC 9755 also leads, scoring 3503 in PassMark single-thread versus 3287 for the Xeon — a 6.6% edge. While smaller, this matters for workloads that depend on per-core responsiveness. The EPYC 9755’s boost clock of 4.10 GHz versus the Xeon’s 3.90 GHz aligns with this result, though the architecture differences also play a role.

The average benchmark score tells the broader story: the EPYC 9755 averages 505778, while the Xeon 6960P averages 365194. The EPYC 9755 sits 18.7% above its nearest rival, the AMD EPYC 9745, and 22.7% above the Ryzen Threadripper PRO 9995WX. The Xeon 6960P, by contrast, is nearly tied with the AMD EPYC 9754 (0.2% delta) and trails the EPYC 9655 by 2.2%. Both processors rank in the 100th percentile among all CPUs, but that percentile masks the gulf between them.

Who should choose which? The EPYC 9755 is the clear choice for compute-intensive, multi-threaded workloads where every percentage point of throughput translates to real time savings — data compression, encryption, and floating-point simulations all favor it heavily. The Xeon 6960P, while still a top-tier processor, offers no benchmarked advantage in any category. If a workload is already bounded by memory bandwidth, the Xeon’s 614.4 GB/s versus the EPYC’s 576.0 GB/s might matter, but no test in this dataset reflects that. For pure compute, the EPYC 9755 dominates.

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9755 has 128 cores and 256 threads, while the Intel Xeon 6960P has 72 cores and 144 threads. The EPYC 9755 offers 56 more cores and 112 more threads, which directly contributes to its multi-threaded benchmark leads.

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

A: The EPYC 9755 wins all three Cinebench multi-core tests by exactly 27.3%: R15 scores 14250 vs 11194, R20 scores 59378 vs 46645, and R23 scores 141378 vs 111060. The consistent delta across all three versions indicates a stable performance ratio.

Q: Is the single-threaded performance difference significant?

A: The EPYC 9755 leads by 6.6% in PassMark single-thread (3503 vs 3287), and also holds a 6.6% edge in the duplicate singlethread test. This is the smallest margin of any benchmark, but it still favors AMD, with the EPYC 9755’s 4.10 GHz boost clock versus the Xeon’s 3.90 GHz.

Q: What are the biggest performance gaps between the two?

A: The largest margin is in PassMark integer math, where the EPYC 9755 scores 1549946 versus 727750 — a 113% advantage. Floating-point math and data encryption both show 75% and 75.9% leads respectively, while data compression is 61.5% ahead.

Q: How does the Xeon 6960P compare to other AMD parts?

A: The Xeon 6960P’s average score of 365194 is nearly identical to the AMD EPYC 9754 (364371, a 0.2% delta) and trails the EPYC 9655 (373479) by 2.2% and the EPYC 9535 (379408) by 3.7%. It sits 4.1% above the EPYC 9475F.

Q: Do both processors support ECC memory and DDR5?

A: Yes, both support ECC memory and DDR5. Both have twelve-channel memory buses, though the Xeon 6960P’s bandwidth is 614.4 GB/s versus the EPYC 9755’s 576.0 GB/s.

Architecture Differences

The EPYC 9755 is built on AMD’s Zen 5 architecture, codenamed Turin, using a 4 nm process at TSMC. The Xeon 6960P uses Intel’s Granite Rapids architecture on a 5 nm process at Intel. This process node difference — 4 nm versus 5 nm — likely contributes to the EPYC 9755’s higher boost clock (4.10 GHz vs 3.90 GHz) despite identical base clocks of 2.70 GHz.

The transistor count further highlights the scale difference: the EPYC 9755 packs 133,040 million transistors across 16 dies, each 70.6 mm², while the Xeon 6960P’s transistor count is not listed, but its die size is 3x 598 mm². The EPYC 9755’s smaller, more numerous dies suggest a chiplet design that scales better for core count.

Cache hierarchies differ substantially. The EPYC 9755 provides 80 KB of L1 per core, 1 MB of L2 per core, and 512 MB of shared L3 cache. The Xeon 6960P offers 112 KB of L1 per core, 2 MB of L2 per core, and 432 MB of shared L3. While the Xeon has more per-core L1 and L2, the EPYC’s larger total L3 (512 MB vs 432 MB) provides more shared cache for the 128-core configuration.

Both processors use DDR5 memory with twelve-channel buses, and both have active production status. The EPYC 9755 supports PCIe Gen 5 with 128 lanes (CPU only), while the Xeon 6960P supports PCIe Gen 5 with 96 lanes. Neither has integrated graphics, and both are locked multipliers, which is typical for server parts.

Specification Differences

| Specification | AMD EPYC 9755 | Intel Xeon 6960P |

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

| Cores | 128 | 72 |

| Threads | 256 | 144 |

| Boost Clock | 4.10 GHz | 3.90 GHz |

| Socket | AMD Socket SP5 | Intel Socket 7529 |

| Architecture | Zen 5 (Turin) | Granite Rapids |

| Process Node | 4 nm | 5 nm |

| Foundry | TSMC | Intel |

| Transistors | 133,040 million | Not listed |

| Die Size | 16x 70.6 mm² | 3x 598 mm² |

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

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

| L3 Cache | 512 MB shared | 432 MB shared |

| Memory Bandwidth | 576.0 GB/s | 614.4 GB/s |

| PCIe Lanes | 128 (Gen 5) | 96 (Gen 5) |

| Release Date | 2024-10-09 | 2024-09-23 |

| Launch MSRP | $12984 | $9625 |

The Xeon 6960P has a higher per-core L1 and L2 cache, plus greater memory bandwidth (614.4 GB/s vs 576.0 GB/s). However, the EPYC 9755 compensates with 56 additional cores, 112 more threads, a larger L3 cache, more PCIe lanes, and a smaller process node.

Head-to-Head Benchmarks

The EPYC 9755 wins all 14 head-to-head benchmarks. The most striking result is PassMark integer math: 1549946 for the EPYC versus 727750 for the Xeon, a 113% delta. This suggests the EPYC 9755’s core architecture and cache design deliver nearly double the integer throughput, which is critical for database, compilation, and general server workloads.

Floating-point math shows a 75% advantage (922900 vs 527473), and data encryption is 75.9% ahead (284927 vs 162013). These results indicate that both scientific computing and cryptographic workloads would see substantial speedups on the EPYC 9755. Data compression follows at 61.5% (4517407 vs 2797724), a key metric for storage and analytics systems.

The Cinebench multi-core results are uniformly 27.3% higher for the EPYC 9755 across R15, R20, and R23. This consistent margin likely reflects the core count difference: 128 versus 72 cores, a 77.8% increase, though efficiency gains from the 4 nm process and Zen 5 architecture also play a role. The PassMark multithread score shows the same 27.3% delta (166328 vs 130659).

Extended instructions and random string sorting show large gaps as well — 56.8% (303321 vs 193404) and 53.6% (571185 vs 371795) respectively. These are indicative of SIMD-heavy and sorting-intensive workloads, where the EPYC 9755’s higher core count and cache size prove decisive. Find prime numbers is 37.9% ahead (2047 vs 1484), and physics is 11.5% ahead (27806 vs 24937).

The smallest margin is in single-threaded performance, where the EPYC 9755 leads by 6.6% (3503 vs 3287). While this is the only sub-double-digit delta, it still favors AMD, and the EPYC 9755’s higher boost clock of 4.10 GHz versus 3.90 GHz directly supports this result.

Across the board, the data shows the EPYC 9755 holds a commanding lead, with only memory bandwidth favoring the Xeon 6960P. No benchmark in this dataset gives the Xeon a win, making the EPYC 9755 the superior choice for virtually all compute-heavy server applications.