AMD EPYC 7H12 vs Intel Core Ultra 5 115U Comparison

The Verdict

The AMD EPYC 7H12 and Intel Core Ultra 5 115U occupy completely different corners of the processor market, and benchmark data reflects that divide. The EPYC 7H12 is a server-class monster with 64 cores and 128 threads, built for parallel workloads that scale across many compute units. The Core Ultra 5 115U is a mobile-focused chip with 8 cores and 10 threads, designed for efficiency in thin-and-light laptops. In head-to-head Cinebench testing, the EPYC 7H12 wins all six recorded comparisons, with deltas ranging from 445.2% to 446.8%. That is not a close contest; it is a categorical difference in compute capacity.

For buyers, the choice depends entirely on use case. The EPYC 7H12, with a 280 W TDP and AMD Socket SP3, belongs in servers or workstations where multi-threaded rendering, virtualization, or heavy database workloads dominate. Its 256 MB shared L3 cache and eight-channel DDR4 memory bus give it massive data throughput. The Core Ultra 5 115U, with a 15 W TDP and Intel BGA 2049 socket, targets ultraportable notebooks where battery life and modest thermals matter more than raw core counts. Its integrated Arc Xe-LPG 48EU graphics add display and light media capabilities that the EPYC lacks entirely.

The average benchmark scores tell a similar story. The EPYC 7H12 posts an average score of 17120, while the Core Ultra 5 115U averages 16753. That difference of roughly 2.2% in average score is small, but it masks the enormous variance in workload types. The EPYC excels in multi-core tests (Cinebench R23 multi-core: 59188 vs. 10855), while the Core Ultra's PassMark results show strengths in data compression (141945), integer math (41043), and floating-point math (30538), though those tests have no direct EPYC counterpart in the database. In percentile terms, the EPYC ranks in the 71st percentile of all CPUs, while the Core Ultra sits in the 70th percentile, meaning both are above-average chips, but for entirely different markets.

Architecture Differences

The architectural gap between these two processors is as wide as their performance gap. The EPYC 7H12 uses AMD's Zen 2 architecture under the Rome codename, fabricated on a 7 nm process by TSMC. It packs 3,800 million transistors into a 74 mm² die. The Core Ultra 5 115U uses Intel's Meteor Lake architecture, also on a 7 nm process, but fabricated by Intel's own foundry. The database does not list a transistor count or die size for the Intel part.

Core counts differ dramatically. The EPYC offers 64 cores and 128 threads, while the Core Ultra provides 8 cores and 10 threads. That thread advantage (128 vs. 10) is the primary driver of the EPYC's multi-core dominance. Cache hierarchies also diverge. The EPYC allocates 96 KB of L1 per core, 512 KB of L2 per core, and a massive 256 MB shared L3. The Core Ultra has 112 KB L1 per core, 2 MB L2 per core, and just 10 MB shared L3. The EPYC's 256 MB L3 is 25.6 times larger than the Core Ultra's 10 MB, which matters for working sets that exceed small caches.

Memory architecture reinforces the server-versus-mobile split. The EPYC supports DDR4 over an eight-channel interface with 204.8 GB/s bandwidth and ECC memory. The Core Ultra supports DDR5, but the database notes it "depends on motherboard," uses a dual-channel bus with 89.6 GB/s bandwidth, and lacks ECC support. PCIe connectivity also differs: both use Gen 4, but the Core Ultra specifies 12 lanes (CPU only), while the EPYC's lane count is not listed. The EPYC has no integrated graphics; the Core Ultra includes Arc Xe-LPG 48EU. The EPYC targets the server/workstation segment, while the Core Ultra is explicitly a mobile part. Release dates differ too: the EPYC launched on September 17, 2019, while the Core Ultra arrived on December 13, 2023.

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 7H12 has 64 cores and 128 threads. The Intel Core Ultra 5 115U has 8 cores and 10 threads. The EPYC offers 8 times the core count and roughly 12.8 times the thread count.

Q: What is the TDP difference between these two chips?

A: The EPYC 7H12 has a 280 W TDP, while the Core Ultra 5 115U has a 15 W TDP. That is an 18.7-fold difference in thermal design power, reflecting their distinct market positions.

Q: Which processor supports ECC memory?

A: The EPYC 7H12 supports ECC memory. The Core Ultra 5 115U does not. This makes the EPYC suitable for error-sensitive server workloads where data integrity is critical.

Q: How do their single-core Cinebench R23 scores compare?

A: The EPYC 7H12 scores 8355 in Cinebench R23 single-core, while the Core Ultra 5 115U scores 1532. The EPYC is 445.4% ahead in this test, a surprising result given the Core Ultra's higher boost clock of 4.20 GHz versus the EPYC's 3.30 GHz boost.

Q: Do these processors use the same socket?

A: No. The EPYC 7H12 uses AMD Socket SP3, while the Core Ultra 5 115U uses Intel BGA 2049. They are not interchangeable in any system.

Q: What is the memory bandwidth available to each processor?

A: The EPYC 7H12 provides 204.8 GB/s over eight DDR4 channels. The Core Ultra 5 115U provides 89.6 GB/s over dual-channel DDR5. The EPYC offers 2.29 times the memory bandwidth.

Specification Differences

| Specification | AMD EPYC 7H12 | Intel Core Ultra 5 115U |

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

| Cores | 64 | 8 |

| Threads | 128 | 10 |

| Base clock | 2.60 GHz | 1500.00 MHz (1.50 GHz) |

| Boost clock | 3.30 GHz | 4.20 GHz |

| TDP | 280 W | 15 W |

| Socket | AMD Socket SP3 | Intel BGA 2049 |

| Architecture | Zen 2 | Meteor Lake |

| Codename | Rome | Meteor Lake |

| Process node | 7 nm (TSMC) | 7 nm (Intel) |

| Foundry | TSMC | Intel |

| Transistors | 3,800 million | Not listed |

| Die size | 74 mm² | Not listed |

| L1 cache | 96 KB per core | 112 KB per core |

| L2 cache | 512 KB per core | 2 MB per core |

| L3 cache | 256 MB shared | 10 MB shared |

| Memory support | DDR4 | DDR5 (depends on motherboard) |

| Memory bus | Eight-channel | Dual-channel |

| Memory bandwidth | 204.8 GB/s | 89.6 GB/s |

| ECC memory | Yes | No |

| PCIe | Gen 4 | Gen 4, 12 lanes (CPU only) |

| Integrated graphics | None | Arc Xe-LPG 48EU |

| Market segment | Server/Workstation | Mobile |

| Release date | 2019-09-17 | 2023-12-13 |

| Part number | 100-000000055 | SRN6F |

| Multiplier unlocked | No | No |

Head-to-Head Benchmarks

The recorded Cinebench data shows a complete sweep. The AMD EPYC 7H12 wins all six head-to-head tests, with no victories for the Intel Core Ultra 5 115U. The margins are consistent and enormous, all falling between 445.2% and 446.8%. This uniformity suggests the difference is structural, driven by the EPYC's 64-core, 128-thread design versus the Core Ultra's 8-core, 10-thread layout.

In Cinebench R15 multi-core, the EPYC scores 5965 against the Core Ultra's 1094, a 445.2% advantage. Single-core R15 shows 842 versus 154, a 446.8% delta, the largest margin in the entire comparison. Cinebench R20 multi-core follows the same pattern: 24858 versus 4559, a 445.3% gap. R20 single-core gives 3509 versus 643, or 445.7% in favor of the EPYC. Cinebench R23 multi-core, the most demanding test in the suite, shows 59188 versus 10855, another 445.3% lead. R23 single-core closes at 8355 versus 1532, a 445.4% margin.

What stands out is that even in single-core tests, the EPYC dominates. The Core Ultra has a significantly higher boost clock (4.20 GHz vs. 3.30 GHz), yet still loses single-core Cinebench by over 400%. This indicates that Zen 2's per-core efficiency at the time of the EPYC's release, combined with its larger cache hierarchy, overcame the raw clock-speed disadvantage. The 256 MB L3 cache likely plays a role, as Cinebench workloads can benefit from larger cache residency.

The PassMark data for the Core Ultra, though not directly comparable in a head-to-head table, provides context for its strengths. Scores like 141945 in data compression, 41043 in integer math, and 30538 in floating-point math show that the Core Ultra is not a weak chip; it simply operates in a different performance envelope. Its 3057 single-thread PassMark score is respectable for a mobile part. But none of those results can close the gap created by 64 cores versus 8 cores in multi-threaded render workloads.

The database records 6 wins for the EPYC and 0 for the Core Ultra in head-to-head benchmarks. That is a decisive outcome. For anyone evaluating these two processors, the data makes the verdict unambiguous: if multi-core compute is the priority, the EPYC 7H12 is the only choice. If low-power mobile operation with integrated graphics is required, the Core Ultra 5 115U serves that role, but it cannot compete with the EPYC in raw throughput. The benchmark record leaves no room for ambiguity.