AMD EPYC 9655P vs Intel Xeon 6960P Comparison

The AMD EPYC 9655P and Intel Xeon 6960P are flagship server processors that both sit at the 100th percentile of all CPUs, but their benchmark profiles are far from identical. The data shows a decisive performance advantage for the AMD part across every single measured workload, with the largest gaps appearing in integer math and encryption. This analysis breaks down the head-to-head results, where each chip fits best, and what the architectural differences mean for a server purchase.

Head-to-Head Benchmarks

The EPYC 9655P wins all 14 head-to-head benchmark comparisons included in the data. This is not a narrow victory; the margin varies significantly by workload. In Cinebench multi-core tests, the AMD chip holds a consistent 22.8% lead across R15, R20, and R23 versions. The EPYC 9655P scores 13,744 in R15, 57,268 in R20, and 136,354 in R23, while the Xeon 6960P trails at 11,194, 46,645, and 111,060 respectively. This uniform delta suggests a fundamental throughput advantage rather than a workload-specific quirk.

The largest single margin appears in PassMark integer math, where the EPYC 9655P scores 1,225,251 against the Xeon's 727,750 — a 68.4% advantage. Floating-point math also shows a substantial gap at 35.7% (715,866 vs 527,473). Data encryption is another standout, with the AMD part delivering 220,074 versus 162,013, a 35.8% lead. These are not marginal differences; they indicate that the EPYC 9655P processes arithmetic and cryptographic workloads at a fundamentally higher rate.

Other PassMark tests show double-digit leads for the AMD part across the board. Data compression comes in 24.6% higher (3,486,158 vs 2,797,724), random string sorting is 21.5% higher (451,824 vs 371,795), and the extended instructions test shows a 19.2% advantage (230,609 vs 193,404). Single-thread performance is also firmly in AMD's favor at 17.1% (3,849 vs 3,287), which is notable for a 96-core part against a 72-core rival. The multithread score follows the Cinebench pattern at 22.8% (160,490 vs 130,659).

The closest contest is in PassMark physics, where the EPYC 9655P wins by just 3.6% (25,847 vs 24,937). Prime number finding is also relatively close at 13.6% (1,686 vs 1,484). These narrower margins suggest that the Xeon 6960P is more competitive in latency-sensitive or branch-heavy tasks, but it still loses every race. The average benchmark score reflects this: the EPYC 9655P averages 397,773 across all tests, while the Xeon 6960P averages 365,194.

Where Each One Wins

Based strictly on the benchmark data, the AMD EPYC 9655P wins in every measured category. There is no workload in the fact pack where the Intel Xeon 6960P comes out ahead. The AMD chip is faster in multi-threaded rendering, integer and floating-point math, encryption, compression, sorting, physics simulation, and even single-threaded tasks.

That said, the degree of the AMD advantage varies, which matters for workload prioritization. If a server's primary job is integer-heavy database or application logic, the 68.4% lead in integer math is the strongest argument for the EPYC 9655P. For encryption-heavy workloads like secure communications or VPN termination, the 35.8% encryption lead is decisive. Floating-point workloads — common in scientific computing and financial modeling — see a 35.7% advantage.

The Xeon 6960P does not win any benchmark, but its relative competitiveness in physics (3.6% gap) and prime number finding (13.6% gap) means that in these specific areas, it gets closer to AMD's performance. Still, the data does not support any use case where the Intel part is the better choice on raw performance. The Xeon 6960P's only advantage in the fact pack is its lower launch MSRP of $9,625 versus the EPYC 9655P's $10,811, but that is a pricing detail, not a performance win.

The Verdict

The verdict from the data is unambiguous: the AMD EPYC 9655P is the faster processor. It wins every benchmark in the head-to-head set, with leads ranging from 3.6% to 68.4%. For any workload that benefits from high core counts, fast math, or strong encryption, the EPYC 9655P is the superior choice.

Who should pick the AMD EPYC 9655P? Anyone running multi-threaded rendering workloads, scientific simulations with heavy floating-point math, or data services that rely on integer and encryption performance. The 22.8% lead in Cinebench multi-core tests and the 68.4% lead in integer math make it the clear pick for compute-heavy server fleets. The 96 cores and 192 threads provide the raw parallelism needed to keep these workloads saturated.

Who should pick the Intel Xeon 6960P? Based strictly on benchmark results, no one looking for maximum performance. The Xeon 6960P loses every single test. The data shows no scenario where it outruns the EPYC 9655P. If a buyer is constrained by platform compatibility or has a specific need for Intel's architecture, the Xeon 6960P is still a capable processor — it sits at the 100th percentile of all CPUs and beats its own nearest rival, the AMD EPYC 9754, by 0.2% in average score. But against the EPYC 9655P, it is slower in every measurable way.

FAQ

Q: How much faster is the AMD EPYC 9655P in multi-core Cinebench tests?

A: The AMD EPYC 9655P is 22.8% faster than the Intel Xeon 6960P in Cinebench R15, R20, and R23 multi-core tests, scoring 13,744, 57,268, and 136,354 respectively versus 11,194, 46,645, and 111,060.

Q: Which processor has the largest performance advantage in any single test?

A: The largest gap is in PassMark integer math, where the AMD EPYC 9655P scores 1,225,251 against the Xeon 6960P's 727,750, a 68.4% advantage.

Q: Does the Intel Xeon 6960P win any benchmark?

A: No. The data shows the Intel Xeon 6960P loses all 14 head-to-head benchmark comparisons, with its closest result being a 3.6% deficit in PassMark physics (24,937 vs 25,847).

Q: What is the single-thread performance difference?

A: The AMD EPYC 9655P scores 3,849 in PassMark single-thread, which is 17.1% higher than the Intel Xeon 6960P's 3,287.

Q: How do the two processors compare in memory bandwidth?

A: The Intel Xeon 6960P has a higher memory bandwidth at 614.4 GB/s, while the AMD EPYC 9655P offers 576.0 GB/s. Both use twelve-channel DDR5.

Q: Which processor has more cores and threads?

A: The AMD EPYC 9655P has 96 cores and 192 threads, while the Intel Xeon 6960P has 72 cores and 144 threads.

Architecture Differences

The two processors are built on fundamentally different architectures and manufacturing processes. The AMD EPYC 9655P uses the Zen 5 architecture with the "Turin" codename, manufactured on a 4 nm process by TSMC. The Intel Xeon 6960P uses the Granite Rapids architecture, built on Intel's 5 nm process. This process difference partially explains the performance gap; the smaller 4 nm node allows for higher density and efficiency.

The chip designs also differ structurally. The EPYC 9655P uses a multi-chiplet design with 12 dies of 70.6 mm² each, totaling 99,780 million transistors. The Xeon 6960P uses 3 dies of 598 mm² each, and the fact pack does not list a transistor count for Intel. The die size difference is substantial — Intel's monolithic-style dies are much larger, which can impact yield and thermal density.

Cache configurations diverge significantly. The AMD part has 80 KB of L1 and 1 MB of L2 per core, plus 384 MB of shared L3 cache. The Intel part has larger per-core caches at 112 KB L1 and 2 MB L2, but the shared L3 is 432 MB. Despite Intel's larger L3, the AMD chip's overall performance advantage suggests its cache hierarchy is better suited to the benchmark workloads.

Memory support is similar in structure — both use twelve-channel DDR5 — but the bandwidth differs, with Intel at 614.4 GB/s and AMD at 576.0 GB/s. The fact pack does not indicate that this bandwidth difference translates into a performance win for Intel in any benchmark. PCIe connectivity also differs: the EPYC 9655P offers 128 Gen 5 lanes, while the Xeon 6960P offers 96 Gen 5 lanes.

Specification Differences

The specifications where the two processors differ are substantial. Core and thread counts are the most obvious: the AMD EPYC 9655P has 96 cores and 192 threads, while the Intel Xeon 6960P has 72 cores and 144 threads. Clock speeds also favor AMD in boost, with 4.50 GHz versus Intel's 3.90 GHz, though Intel has a slightly higher base clock at 2.70 GHz versus 2.60 GHz.

Power consumption differs, with the AMD part rated at 400 W TDP and the Intel part at 500 W TDP. The fact pack does not include any measured power or efficiency numbers, so no analysis of performance-per-watt is possible from this data alone. Socket compatibility is completely different: the AMD part uses AMD Socket SP5, while the Intel part uses Intel Socket 7529.

Manufacturing details show the AMD chip on a 4 nm TSMC process with a 12-die layout, while the Intel chip is on a 5 nm Intel process with a 3-die layout. Cache layouts differ as detailed above, and PCIe lane count favors AMD (128 vs 96). The release dates are close, with the Intel part launching on 2024-09-23 and the AMD part on 2024-10-09. Neither processor includes integrated graphics, and both are marked as active production with ECC memory support.