CPU Comparison

Head-to-Head Benchmarks

The Intel Xeon 6960P dominates the AMD EPYC 9734 in overall benchmark wins, taking 9 of the 14 head-to-head comparisons. The most decisive victory comes in PassMark physics, where the Xeon 6960P scores 24,937 against the EPYC 9734's 6,747, a staggering 269.6% advantage. This is the single largest performance gap between the two processors in any tested workload.

The Xeon 6960P also sweeps the Cinebench multicore tests with consistent 27.7% margins across R15, R20, and R23. In Cinebench R23 multicore, the Intel part scores 111,060 versus 86,943 for the AMD chip. The same 27.7% delta appears in PassMark multithread scoring, where the Xeon 6960P posts 130,659 against 102,286. This consistency across rendering and general multithreaded workloads suggests the Intel processor has a structural advantage in heavily threaded scenarios.

Single-thread performance strongly favors Intel as well. The Xeon 6960P scores 3,287 in PassMark single-thread tests, a 42.3% improvement over the EPYC 9734's 2,310. The EPYC 9734's Cinebench single-core results, 1,237 in R15, 5,155 in R20, and 12,274 in R23, have no direct Intel comparison in the head-to-head data, but the PassMark single-thread delta tells a clear story.

The AMD EPYC 9734 wins the remaining five tests, but with narrower margins. Its best result comes in PassMark integer math, where it scores 823,150 versus 727,750, an 11.6% advantage. Data encryption shows a 9.7% lead (179,390 vs. 162,013), and extended instructions favor AMD by 6.1% (205,925 vs. 193,404). Floating-point math and data compression round out AMD's wins with 3.9% and 3.5% edges, respectively.

The Xeon 6960P also wins random string sorting by 4% (371,795 vs. 357,638) and find prime numbers by a massive 79% (1,484 vs. 829). Interestingly, the prime number test is the only workload where the Intel chip's per-core advantage translates into an outsized win, the 42.3% single-thread lead becomes 79% in this specific calculation.

Architecture Differences

The two processors represent fundamentally different design philosophies despite sharing a 5 nm process node. The Intel Xeon 6960P uses Granite Rapids architecture with three 598 mm² dies fabricated by Intel, while the AMD EPYC 9734 uses Zen 4c (Bergamo) with eight 73 mm² dies from TSMC. AMD packs 71,000 million transistors into its chiplet design, while Intel's transistor count is not listed.

Core counts diverge sharply. The EPYC 9734 offers 112 cores and 224 threads, compared to the Xeon 6960P's 72 cores and 144 threads. This 40-core difference gives AMD a raw thread-count advantage, yet the benchmark data shows Intel winning most multithreaded tests, a testament to the Xeon's higher clock speeds. The Intel part runs at 2.70 GHz base and 3.90 GHz boost, versus 2.20 GHz base and 3.00 GHz boost for the AMD chip.

Cache hierarchies also differ substantially. Intel allocates 112 KB of L1 and 2 MB of L2 per core, with a massive 432 MB of shared L3 cache. AMD uses 64 KB L1 and 1 MB L2 per core, with 256 MB shared L3. The Intel chip's larger per-core caches likely contribute to its single-thread and physics performance advantages.

Memory bandwidth favors Intel at 614.4 GB/s versus AMD's 460.8 GB/s, though both use DDR5 over twelve channels. PCIe lane counts reverse the trend: the EPYC 9734 provides 128 Gen 5 lanes from the CPU, while the Xeon 6960P offers 96. Both support ECC memory and lack integrated graphics.

Power profiles also differ. The Xeon 6960P carries a 500 W TDP, while the EPYC 9734 draws 340 W. The Intel part's higher power envelope correlates with its higher clock speeds and benchmark wins. Socket compatibility is split: Intel uses Socket 7529, AMD uses SP5.

Where Each One Wins

The Intel Xeon 6960P clearly targets workloads that demand high per-core performance combined with massive multicore throughput. Its physics test dominance (269.6% ahead) and 42.3% single-thread lead make it the choice for simulation, scientific computing, and any workload where individual thread performance matters. The consistent 27.7% wins across all three Cinebench versions and PassMark multithread indicate strong rendering and general parallel compute capabilities. The 79% lead in prime number finding further confirms Intel's arithmetic throughput advantages.

The AMD EPYC 9734 wins where raw core count and specific instruction efficiency matter. Its 11.6% lead in integer math, 9.7% in data encryption, and 6.1% in extended instructions suggest strengths in database operations, cryptography, and workloads using specialized instruction sets. The 3.5% data compression win and 3.9% floating-point math edge make it competitive for archival, compression-heavy, and certain scientific workloads. The EPYC's 112 cores provide more parallel execution units, which helps in these specific throughput-oriented tasks.

Random string sorting slightly favors Intel (4% delta), indicating better memory access patterns for sorting algorithms. The overall pattern shows Intel winning broad-spectrum performance while AMD excels in targeted, core-count-heavy workloads.

FAQ

Q: Which processor has more cores?

A: The AMD EPYC 9734 has 112 cores and 224 threads, while the Intel Xeon 6960P has 72 cores and 144 threads.

Q: How much faster is the Intel Xeon 6960P in Cinebench R23 multicore?

A: The Xeon 6960P scores 111,060 versus 86,943 for the EPYC 9734, a 27.7% advantage.

Q: Does the AMD EPYC 9734 win any benchmarks?

A: Yes, AMD wins five of 14 head-to-head tests: data compression, data encryption, extended instructions, floating-point math, and integer math.

Q: What is the biggest single benchmark margin between the two?

A: PassMark physics shows the largest delta, with the Intel Xeon 6960P scoring 24,937 versus 6,747 for the EPYC 9734, a 269.6% lead.

Q: How do their memory bandwidth figures compare?

A: Intel's Xeon 6960P provides 614.4 GB/s, while AMD's EPYC 9734 offers 460.8 GB/s, both over twelve-channel DDR5.

Q: Which processor has a higher boost clock?

A: The Intel Xeon 6960P boosts to 3.90 GHz, compared to 3.00 GHz for the AMD EPYC 9734.

Specification Differences

| Specification | Intel Xeon 6960P | AMD EPYC 9734 |

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

| Cores | 72 | 112 |

| Threads | 144 | 224 |

| Base Clock | 2.70 GHz | 2.20 GHz |

| Boost Clock | 3.90 GHz | 3.00 GHz |

| TDP | 500 W | 340 W |

| Socket | Intel Socket 7529 | AMD Socket SP5 |

| Architecture | Granite Rapids | Zen 4c (Bergamo) |

| Foundry | Intel | TSMC |

| Transistors | Not listed | 71,000 million |

| Die Size | 3x 598 mm² | 8x 73 mm² |

| L1 Cache | 112 KB (per core) | 64 KB (per core) |

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

| L3 Cache | 432 MB (shared) | 256 MB (shared) |

| Memory Bandwidth | 614.4 GB/s | 460.8 GB/s |

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

| Release Date | 2024-09-23 | 2023-06-12 |

| Launch MSRP | $9625 | $9600 |

| Part Number | SRPKX | 100-000001235 |

The Verdict

The Intel Xeon 6960P is the superior processor for most compute-intensive workloads. Its 27.7% wins across all Cinebench multicore tests and PassMark multithread, combined with a 42.3% single-thread lead, make it the clear choice for rendering, simulation, physics, and general high-performance computing. The 269.6% physics benchmark margin is unprecedented between two modern server CPUs. The 79% prime-number finding advantage further cements Intel's dominance in arithmetic-heavy tasks.

The AMD EPYC 9734 should be selected when specific workload characteristics align with its strengths. Its 11.6% integer math lead, 9.7% encryption advantage, and 6.1% extended instructions win make it suitable for database workloads, cryptographic operations, and specialized instruction-heavy applications. The 40-core advantage and 128 PCIe lanes also make it attractive for highly parallel, I/O-heavy server deployments where raw core count and connectivity matter more than per-core speed.

The data shows Intel winning 9 of 14 benchmarks with an average benchmark score of 365,194 versus AMD's 310,619. Intel's percentile ranking at 100 versus AMD's 99 reflects this overall superiority. However, the EPYC 9734's 340 W TDP versus Intel's 500 W means AMD offers competitive performance in select workloads at lower power consumption. For users prioritizing the broadest performance across diverse workloads, the Xeon 6960P is the data-backed choice; for those with core-count-hungry, specific-use-case workloads, the EPYC 9734 holds distinct advantages.