AMD EPYC 9734 vs Intel Xeon 6780E Comparison

The AMD EPYC 9734 and Intel Xeon 6780E are both 99th-percentile server processors, yet they represent fundamentally different design philosophies. The EPYC 9734, built on Zen 4c (Bergamo), leverages 112 cores and 224 threads, while the Xeon 6780E, based on Sierra Forest, counters with 144 cores but only 144 threads. Benchmark results show the AMD part winning 14 of 16 head-to-head tests, but the Intel chip claims two decisive victories that reveal its specialized strengths. The data indicates a clear split: AMD dominates general compute, rendering, and math workloads, while Intel holds a significant edge in physics simulation and encryption.

Where Each One Wins

The AMD EPYC 9734 is the overwhelming winner in multi-threaded rendering and CPU arithmetic. In Cinebench R23 multi-core, the EPYC 9734 scores 86,943 versus the Xeon 6780E’s 73,723, a 17.9% advantage. This pattern repeats across all Cinebench versions—R15, R20, and R23—with identical 17.9% deltas in both multi-core and single-core tests. The AMD chip also leads in PassMark’s integer math (823,150 vs 641,817, up 28.3%) and floating-point math (549,045 vs 433,862, up 26.5%). For workloads that stress extended instruction sets, the gap widens dramatically: the EPYC 9734 scores 205,925 versus 114,008 in PassMark extended instructions, a massive 80.6% lead. This suggests AMD’s core architecture handles AVX-style workloads far more efficiently.

The Intel Xeon 6780E wins exactly two benchmarks, but they are telling. In PassMark physics, the Intel chip scores 10,951 versus 6,747 for AMD, a 38.4% advantage—the largest margin of any test in either direction. This indicates the Xeon’s core design excels at physics simulation, likely due to different scheduling or cache behavior. Intel also wins PassMark data encryption with 193,004 versus 179,390, a 7.1% edge. For servers running cryptographic workloads or physics-heavy simulations, the Xeon 6780E presents a compelling alternative despite losing the overall benchmark count.

Architecture Differences

The two processors diverge sharply in core organization. The AMD EPYC 9734 uses 112 cores with 224 threads, meaning every core supports two threads via SMT. The Intel Xeon 6780E has 144 cores but only 144 threads, indicating a single-thread-per-core design. This explains why the AMD chip wins multi-threaded tests despite having fewer physical cores—its 224 threads provide more parallel work items. Both chips share the same 2.20 GHz base clock and 3.00 GHz boost clock, so frequency is not a differentiator.

Cache hierarchies are also distinct. AMD allocates 64 KB L1 and 1 MB L2 per core, plus a shared 256 MB L3. Intel uses 96 KB L1 per core and 4 MB L2 per module, with only 108 MB shared L3. The AMD part’s 256 MB L3 cache is more than double Intel’s, which likely contributes to its data-compression win (2,900,008 vs 2,557,582, up 13.4%). The process nodes differ despite both being 5 nm: AMD uses TSMC as foundry, while Intel uses its own fabs. The die sizes reflect this—AMD’s is 8x 73 mm² chiplets versus Intel’s monolithic 578 mm².

Memory architecture also separates them. The EPYC 9734 supports twelve-channel DDR5 with 460.8 GB/s bandwidth, while the Xeon 6780E uses eight-channel DDR5 with 409.6 GB/s. This 12.5% bandwidth advantage for AMD helps in memory-intensive workloads like random string sorting, where AMD wins 357,638 vs 326,954 (up 9.4%). PCIe connectivity differs too: AMD offers 128 Gen 5 lanes, Intel only 88. Both support ECC memory, and neither has integrated graphics (Intel lists "N/A" explicitly).

Head-to-Head Benchmarks

The single most lopsided result is PassMark extended instructions, where the EPYC 9734’s 205,925 beats the Xeon 6780E’s 114,008 by 80.6%. This is not a marginal win but a categorical one, suggesting the Zen 4c cores execute SIMD and cryptographic extension instructions with far greater throughput. The smallest AMD win is in random string sorting at 9.4%, still a clear margin. Data compression shows a 13.4% AMD lead, while integer math (28.3%) and floating-point math (26.5%) are both substantial.

The Intel wins are equally pronounced. The physics benchmark shows a 38.4% Intel advantage, the largest delta in the entire dataset. Data encryption yields a more modest 7.1% Intel edge. Notably, the Cinebench single-core tests show AMD leading by 17.9% (1,237 vs 1,049 in R15; 5,155 vs 4,371 in R20), which contradicts the common assumption that lower-core-count designs have higher single-thread performance. The EPYC 9734’s single-thread score of 2,310 in PassMark is 20.1% higher than the Xeon’s 1,923.

The multi-threaded PassMark score (102,286 vs 86,734) and the Cinebench R23 multi-core result both show the same 17.9% delta, reinforcing that the AMD chip’s thread advantage translates consistently across synthetic loads. Find prime numbers shows a 17.1% AMD win (829 vs 708), another arithmetic test where AMD prevails. Overall, the data paints a picture of AMD leading in 14 of 16 tests, with Intel only countering in physics and encryption.

FAQ

Q: Why does the Intel Xeon 6780E have more cores but lose most benchmarks?

A: The Xeon 6780E has 144 cores but only 144 threads, while the AMD EPYC 9734 has 112 cores and 224 threads. The AMD chip’s SMT capability provides 80 additional threads, which outweighs Intel’s 32-core physical advantage in multi-threaded tests like Cinebench R23 (86,943 vs 73,723).

Q: Which processor is better for physics simulation workloads?

A: The Intel Xeon 6780E. It scores 10,951 in PassMark physics versus 6,747 for the EPYC 9734, a 38.4% higher result. This is Intel’s largest win across all benchmarks.

Q: Is the AMD EPYC 9734 better at encryption?

A: No. The Intel Xeon 6780E wins PassMark data encryption with 193,004 versus 179,390, a 7.1% advantage. However, AMD leads extended instructions by 80.6%, so the choice depends on the specific cryptographic algorithm.

Q: How do the memory systems compare?

A: The EPYC 9734 uses twelve-channel DDR5 with 460.8 GB/s bandwidth, while the Xeon 6780E uses eight-channel DDR5 with 409.6 GB/s. The AMD chip’s higher bandwidth likely contributes to its win in random string sorting (357,638 vs 326,954).

Q: What is the largest performance gap in either direction?

A: The biggest delta is AMD’s 80.6% lead in PassMark extended instructions (205,925 vs 114,008). The largest Intel advantage is 38.4% in PassMark physics (10,951 vs 6,747).

Q: Do both processors support the same PCIe generation?

A: Yes, both support PCIe Gen 5, but AMD provides 128 lanes (CPU only) while Intel provides 88 lanes. This gives the EPYC 9734 more expansion headroom for accelerators or storage.

Specification Differences

| Specification | AMD EPYC 9734 | Intel Xeon 6780E |

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

| Cores | 112 | 144 |

| Threads | 224 | 144 |

| Socket | AMD Socket SP5 | Intel Socket 4710 |

| Architecture | Zen 4 | Sierra Forest |

| Codename | Bergamo | Sierra Forest |

| Generation | EPYC (Zen 4c (Bergamo)) | Xeon 6 (Sierra Forest-SP) |

| Foundry | TSMC | Intel |

| Transistors | 71,000 million | (not specified) |

| Die Size | 8x 73 mm² | 578 mm² |

| L1 Cache | 64 KB (per core) | 96 KB (per core) |

| L2 Cache | 1 MB (per core) | 4 MB (per module) |

| L3 Cache | 256 MB (shared) | 108 MB (shared) |

| Memory Bus | Twelve-channel | Eight-channel |

| Memory Bandwidth | 460.8 GB/s | 409.6 GB/s |

| PCIe | Gen 5, 128 Lanes (CPU only) | Gen 5, 88 Lanes (CPU only) |

| Integrated Graphics | (none listed) | N/A |

| Release Date | 2023-06-12 | 2024-06-02 |

| Launch MSRP | $9600 | $11350 |

| Part Number | 100-000001235 | SRPG3 |

The specification sheet confirms that the AMD EPYC 9734 carries more threads, more cache (256 MB L3 vs 108 MB), a wider memory bus, and more PCIe lanes. The Intel Xeon 6780E counters with more physical cores, a larger L1 (96 KB vs 64 KB) and L2 (4 MB vs 1 MB) per core/module, and a later release date. The Xeon’s higher launch MSRP of $11,350 versus AMD’s $9,600 suggests Intel positions it as a premium part, but the benchmark data shows AMD winning the majority of performance tests. The Xeon 6780E’s strengths are narrow—physics and encryption—while the EPYC 9734 delivers broader superiority across rendering, math, compression, and single-thread performance. For mixed server workloads, the AMD part is the data-backed choice; for specialized physics or crypto tasks, Intel’s design holds a clear edge.