AMD EPYC 9745 vs Intel Xeon 6960P Comparison

# AMD EPYC 9745 vs Intel Xeon 6960P

The AMD EPYC 9745 and Intel Xeon 6960P are both flagship server processors, and while they land in the same performance tier on aggregate metrics, their individual benchmark profiles diverge sharply. The EPYC 9745 wins 10 of 14 head-to-head tests, while the Xeon 6960P counters with four wins, including a decisive single-thread advantage. The EPYC 9745's average benchmark score of 425,973 places it ahead of the Xeon 6960P's 365,194, and both chips sit at the 100th percentile among all CPUs, but the nature of their workloads dictates the winner.

Head-to-Head Benchmarks

The most striking result is the multi-threaded Cinebench parity. In Cinebench R15 multi-core, the EPYC 9745 scores 11,198 against the Xeon 6960P's 11,194 — a 0% delta. Cinebench R20 multi-core shows the same story: 46,659 versus 46,645. Cinebench R23 multi-core follows with 111,093 versus 111,060. Across all three Cinebench versions, the EPYC 9745 edges ahead by a negligible margin of 0%. These are effectively tied scores, meaning raw multi-threaded rendering capability is indistinguishable between the two.

The EPYC 9745 dominates in memory-sensitive and encryption workloads. In PassMark data compression, the AMD chip scores 3,929,890 versus 2,797,724 — a 40.5% advantage. Data encryption shows a 41.6% lead with 229,447 versus 162,013. Extended instructions (SIMD-heavy workloads) favor the EPYC 9745 by 45% (280,477 versus 193,404). Floating-point math sees a 44.3% gap (761,219 versus 527,473). The largest win for AMD comes in integer math: 1,224,315 versus 727,750, a 68.2% lead. Random string sorting goes to the EPYC 9745 by 26.1% (468,975 versus 371,795). PassMark multi-threaded is nearly tied at 130,698 versus 130,659 (0% delta).

The Xeon 6960P wins four tests. In PassMark find prime numbers, it scores 1,484 versus 979 — a 34% lead for Intel. PassMark physics shows 24,937 versus 17,122, a 31.3% advantage. Single-threaded performance goes to Intel decisively: 3,287 versus 2,806 in both PassMark single-thread and single-thread variants, a 14.6% lead. That single-thread gap is the most important Intel win, as it suggests better per-core performance despite fewer cores.

Architecture Differences

The two processors are built on fundamentally different designs. The AMD EPYC 9745 uses the Zen 5 architecture (codename Turin) on a 3 nm TSMC process, while the Intel Xeon 6960P uses Granite Rapids on Intel's 5 nm process with a die size of 3x 598 mm². The EPYC 9745 packs 128 cores and 256 threads; the Xeon 6960P has 72 cores and 144 threads. That core count difference explains the AMD's dominance in parallel integer and floating-point workloads.

Cache hierarchies differ significantly. The EPYC 9745 has 80 KB of L1 per core, 1 MB of L2 per core, and 256 MB of shared L3 cache. The Xeon 6960P has 112 KB of L1 per core, 2 MB of L2 per core, and a much larger 432 MB of shared L3 cache. The Intel chip's larger L3 (432 MB versus 256 MB) does not translate to benchmark wins in the memory-heavy tests, likely because the AMD chip's higher core count compensates.

Clock speeds favor Intel: the Xeon 6960P has a 2.70 GHz base clock and 3.90 GHz boost clock, versus 2.40 GHz base and 3.70 GHz boost for the EPYC 9745. That 0.2 GHz boost advantage contributes to Intel's single-thread win. Power envelopes differ too: the Xeon 6960P has a 500 W TDP versus the EPYC 9745's 400 W. Both support DDR5 memory with twelve-channel buses, but the Xeon 6960P has higher memory bandwidth at 614.4 GB/s versus 576.0 GB/s for the EPYC 9745. Both support ECC memory.

PCIe lanes favor AMD: the EPYC 9745 offers Gen 5 with 128 lanes (CPU only), while the Xeon 6960P provides Gen 5 with 96 lanes. Sockets are incompatible — AMD Socket SP5 for the EPYC 9745, Intel Socket 7529 for the Xeon 6960P. Neither has integrated graphics. The EPYC 9745 was released on 2024-10-09, while the Xeon 6960P launched earlier on 2024-09-23. The launch MSRP is $12,141 for the EPYC 9745 and $9,625 for the Xeon 6960P.

Where Each One Wins

The AMD EPYC 9745 is the clear choice for compute-heavy parallel workloads. It wins by 40.5% in data compression, 41.6% in encryption, 45% in extended instructions, 44.3% in floating-point math, 68.2% in integer math, and 26.1% in random string sorting. These are the workloads that scale with core count — database operations, scientific computing, financial modeling, and any task that can saturate 256 threads. The 128-core count makes it ideal for virtualization, cloud compute, and HPC clusters where thread-level parallelism is the bottleneck.

The Intel Xeon 6960P wins in prime number finding (34% lead), physics simulation (31.3% lead), and single-threaded tasks (14.6% lead). The physics win is particularly notable — it suggests better per-core efficiency in workloads with limited parallelism. The Xeon 6960P's higher boost clock (3.90 GHz versus 3.70 GHz) and larger L3 cache (432 MB versus 256 MB) likely drive this. For legacy workloads with poor thread scaling, or single-threaded database queries, the Xeon 6960P has an edge.

In mixed workloads, the data shows near-parity. Cinebench multi-core tests are statistically tied across all three versions (0% delta), and PassMark multi-threaded scores are effectively equal (130,698 versus 130,659). This means for general multi-threaded rendering or video encoding, either processor delivers equivalent results. The EPYC 9745's wins in integer math (68.2%) and floating-point math (44.3%) are far larger than any Intel win, suggesting AMD has a broader advantage across typical server workloads.

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9745 has 128 cores and 256 threads, while the Intel Xeon 6960P has 72 cores and 144 threads.

Q: Is the AMD EPYC 9745 faster in single-threaded performance?

A: No. The Intel Xeon 6960P wins PassMark single-thread with a score of 3,287 versus 2,806, a 14.6% advantage.

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

A: They are effectively tied. Cinebench R15, R20, and R23 multi-core scores show a 0% delta, with the EPYC 9745 leading by 4 to 33 points.

Q: Which processor has higher memory bandwidth?

A: The Intel Xeon 6960P has 614.4 GB/s memory bandwidth versus 576.0 GB/s for the AMD EPYC 9745, though both use twelve-channel DDR5.

Q: What are the process node differences?

A: The AMD EPYC 9745 uses a 3 nm TSMC process, while the Intel Xeon 6960P uses Intel's 5 nm process.

Q: Which processor has more PCIe lanes?

A: The AMD EPYC 9745 has 128 Gen 5 lanes (CPU only), while the Intel Xeon 6960P has 96 Gen 5 lanes.

The Verdict

The data points to the AMD EPYC 9745 as the superior processor for most server workloads. It wins 10 of 14 head-to-head benchmarks, with its largest margins coming in integer math (68.2% ahead), extended instructions (45%), and floating-point math (44.3%). The aggregate benchmark score is 425,973 versus 365,194 — a 60,779-point gap. The EPYC 9745 achieves this with a lower TDP (400 W versus 500 W) and more PCIe lanes (128 versus 96), making it more efficient per watt and more expandable.

The Intel Xeon 6960P should be chosen only for specific workloads that favor its strengths: single-threaded performance (14.6% lead), physics simulation (31.3% lead), and prime number finding (34% lead). Its 432 MB L3 cache and 3.90 GHz boost clock make it attractive for latency-sensitive applications with limited thread scaling. The Xeon 6960P also has a lower launch MSRP of $9,625 versus $12,141.

For most buyers, the EPYC 9745 is the better investment. The 68.2% integer math lead and 40.5% compression advantage translate directly to faster database and data-processing tasks, which dominate real-world server usage. The near-tied Cinebench results mean rendering performance is a wash, and the AMD's higher core count provides more headroom for virtualized environments. The Xeon 6960P's wins are real but narrow — no Intel victory exceeds 34%, while AMD wins by 40-68% in four separate tests. Unless the workload is heavily single-threaded or physics-based, the EPYC 9745's aggregate superiority (16.6% higher average score) makes it the default recommendation.

The nearest rival data reinforces this: the EPYC 9745 sits 3.4% ahead of the AMD Ryzen Threadripper PRO 9995WX and 7.1% ahead of the AMD EPYC 9655P. The Xeon 6960P, by contrast, is only 0.2% ahead of the older AMD EPYC 9754 and is beaten by the AMD EPYC 9655 (2.2%) and EPYC 9535 (3.7%). This suggests the Xeon 6960P is competitive with the previous AMD generation but not with the current EPYC 9005 series.

Specification Differences

| Specification | AMD EPYC 9745 | Intel Xeon 6960P |

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

| Cores | 128 | 72 |

| Threads | 256 | 144 |

| Base Clock | 2.40 GHz | 2.70 GHz |

| Boost Clock | 3.70 GHz | 3.90 GHz |

| TDP | 400 W | 500 W |

| Socket | AMD Socket SP5 | Intel Socket 7529 |

| Architecture | Zen 5 | Granite Rapids |

| Codename | Turin | Granite Rapids |

| Process Node | 3 nm | 5 nm |

| Foundry | TSMC | Intel |

| Die Size | Not specified | 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 | 256 MB (shared) | 432 MB (shared) |

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

| PCIe | Gen 5, 128 Lanes | Gen 5, 96 Lanes |

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

| Launch MSRP | $12,141 | $9,625 |

| Part Number | 100-000001460 | SRPKX |