AMD EPYC 9454 vs Intel Core Ultra 5 238V Comparison

Head-to-Head Benchmarks

The benchmark data presents an unusually one-sided comparison. Across every recorded test in the head-to-head set, the AMD EPYC 9454 dominates the Intel Core Ultra 5 238V. The margin is consistent and striking: a uniform 78.7% advantage for AMD in every single Cinebench test, both single-core and multi-core.

Starting with multi-core performance, the gap is enormous. In Cinebench R23 multi-core, the EPYC 9454 scores 73,375 points against the Ultra 5 238V's 15,645. That is a 78.7% delta, meaning the AMD part delivers roughly 4.7 times the multi-threaded rendering throughput. The pattern repeats in Cinebench R20 multi-core, where the EPYC scores 30,817 versus 6,570, and in Cinebench R15 multi-core, where it scores 7,396 versus 1,576. These are not marginal differences; they represent entirely different performance tiers.

What is more surprising is the single-core results. The Ultra 5 238V has a higher boost clock (4.70 GHz versus 3.80 GHz) and a newer architecture, yet it still loses decisively. In Cinebench R23 single-core, the EPYC 9454 scores 10,358 versus 2,208 for the Intel part, a 78.7% deficit. The same margin appears in Cinebench R20 single-core (4,350 versus 927) and Cinebench R15 single-core (1,044 versus 222). This uniformity across all six tests suggests a fundamental performance ceiling for the Ultra 5 238V in these workloads, not just a multi-core scaling issue.

The average benchmark scores tell a similar story, though with less drama. The Intel part averages 21,981 across all recorded benchmarks, while the AMD part averages 21,223. That places the Ultra 5 238V slightly ahead on average, but this is misleading. The Intel chip has many more PassMark tests recorded, including data compression (176,532), encryption (13,072), and integer math (38,889), which inflate its average. The EPYC 9454 only has Cinebench results in the database, so its average reflects purely rendering workloads. The head-to-head tests, which use identical workloads, are the more reliable comparison.

Both processors sit at the 75th percentile against all CPUs in the database, which is a curious coincidence. It suggests that while the EPYC 9454 is a server-class monster, its overall standing is pulled down by the limited test set, while the Ultra 5 238V benefits from a broader range of lighter workloads.

Where Each One Wins

The data points to a clear split: the EPYC 9454 wins every rendering and compute-heavy benchmark, while the Ultra 5 238V has no head-to-head victories to claim. That said, the Intel part does have strengths in other recorded tests that the AMD chip does not compete in.

For multi-threaded productivity, the EPYC 9454 is the obvious choice. Its 48 cores and 96 threads allow it to crush Cinebench workloads, which are highly parallel. The 78.7% margin in every multi-core test indicates that the EPYC scales far better with thread count, which is expected given its 6x core advantage (48 versus 8). Any workload that can utilize more than a handful of threads will favor the AMD part.

For single-threaded responsiveness, the data shows that the EPYC 9454 also wins, despite the Intel chip's higher boost clock. This is notable because single-core performance typically correlates with clock speed and IPC. The EPYC's Zen 4 architecture at 3.80 GHz outperforms the Lunar Lake chip at 4.70 GHz by 78.7%, which suggests that the EPYC's per-core efficiency is substantially higher in these specific tests.

The Ultra 5 238V does have recorded wins in PassMark tests, but those are not head-to-head comparisons. Its PassMark single-thread score of 3,890 and multi-thread score of 18,407 show reasonable performance for a mobile chip. The data compression score of 176,532 and floating-point math score of 53,160 indicate solid capabilities for everyday tasks. However, since the EPYC 9454 has no PassMark results, these cannot be directly compared.

The use-case split is therefore stark: the EPYC 9454 is for server and workstation workloads where rendering, simulation, and heavy compute dominate. The Ultra 5 238V is for mobile devices where power efficiency and lighter tasks matter, but the data cannot claim it wins any head-to-head benchmark.

Architecture Differences

The two processors come from opposite ends of the computing spectrum. The Intel Core Ultra 5 238V is built on Lunar Lake architecture using a 3 nm process from TSMC, while the AMD EPYC 9454 uses Zen 4 architecture on a 5 nm process, also from TSMC. The smaller node gives Intel a theoretical transistor density advantage, but the EPYC compensates with raw scale.

Core counts differ massively: the Ultra 5 238V has 8 cores and 8 threads, while the EPYC 9454 has 48 cores and 96 threads. The EPYC also has a higher base clock (2.75 GHz versus 2.10 GHz), though the Intel chip boosts higher (4.70 GHz versus 3.80 GHz). The EPYC's transistor count is recorded at 52,560 million, with a die size of 8x 72 mm², while the Intel part has no recorded transistor or die size data.

Cache hierarchies are also divergent. The Intel chip has 192 KB of L1 per core, 2.5 MB of L2 per core, and 8 MB of shared L3. The EPYC has 64 KB of L1 per core, 1 MB of L2 per core, and a massive 256 MB of shared L3. That 32x difference in L3 cache is a major factor in server workloads that repeatedly access large datasets.

Memory support further separates them. The Ultra 5 238V supports dual-channel memory, with the database noting that capacity depends on the motherboard. The EPYC 9454 supports twelve-channel DDR5 memory with a recorded bandwidth of 460.8 GB/s. The EPYC also supports ECC memory, while the Intel part does not. PCIe lanes differ as well: the EPYC offers Gen 5 with 128 lanes, while the Intel chip offers Gen 5 with only 4 lanes.

The Intel part includes integrated graphics (Arc 130V), while the EPYC has none recorded. The EPYC is a server/workstation chip on AMD Socket SP5, while the Ultra 5 238V is a mobile chip on Intel BGA 2833. Power envelopes are unsurprisingly different: the EPYC has a TDP of 290 watts versus 17 watts for the Intel part.

The Verdict

The recorded data makes the verdict straightforward for compute-heavy workloads: the AMD EPYC 9454 is the superior processor. It wins every head-to-head benchmark by a consistent 78.7% margin, including single-core tests where the Intel chip's higher clock speed does not help. For rendering, simulation, or any parallel workload, the EPYC's 48 cores and 96 threads deliver roughly 4.7 times the multi-core performance of the Ultra 5 238V.

The Intel Core Ultra 5 238V is not without merit, but its strengths lie outside the head-to-head tests. Its PassMark scores show competent single-threaded performance (3,890) and respectable multi-threaded throughput (18,407) for a mobile chip. Its 17 watt TDP makes it suitable for laptops and compact devices, whereas the EPYC's 290 watt TDP demands server infrastructure. The Intel chip also includes integrated graphics, which the EPYC lacks entirely.

The launch MSRP for the EPYC 9454 is $5,225, a figure that reflects its enterprise positioning. The Ultra 5 238V has no recorded launch MSRP, but its mobile segment and lower TDP imply a different market entirely.

Who should pick which? The data suggests that anyone running Cinebench-class workloads on a server should choose the EPYC 9454. Its single-core advantage, despite a lower boost clock, indicates strong per-core efficiency in Zen 4. The Ultra 5 238V is the choice for mobile users who need a capable processor with integrated graphics and minimal power draw, but the benchmark data cannot justify it for heavy compute tasks.

FAQ

Q: Why does the AMD EPYC 9454 win every head-to-head benchmark by the same 78.7% margin?

A: The uniform delta across all six Cinebench tests suggests a consistent performance ceiling difference between the two architectures. The EPYC's 48 cores and 96 threads, combined with its 256 MB L3 cache, give it a structural advantage in rendering workloads.

Q: Does the Intel chip's higher boost clock (4.70 GHz) help it in single-core tests?

A: No. In Cinebench R23 single-core, the EPYC 9454 scores 10,358 versus 2,208 for the Ultra 5 238V, a 78.7% deficit for Intel despite its 0.90 GHz clock advantage. The EPYC's Zen 4 architecture delivers better per-clock performance in these tests.

Q: Which processor has better memory bandwidth?

A: The EPYC 9454 has a recorded memory bandwidth of 460.8 GB/s with twelve-channel DDR5 support. The Ultra 5 238V has dual-channel memory with no bandwidth figure recorded, and its memory capacity depends on the motherboard.

Q: Is the Ultra 5 238V suitable for server workloads?

A: The data does not support this. It has only 4 PCIe Gen 5 lanes, no ECC support, and loses every head-to-head benchmark by a large margin. Its 17 watt TDP and mobile socket indicate it is designed for portable devices.

Q: Does the EPYC 9454 have integrated graphics?

A: No integrated graphics are recorded for the EPYC 9454. The Ultra 5 238V includes Arc 130V integrated graphics, which makes it self-sufficient for display output in mobile systems.

Q: How do their average benchmark scores compare?

A: The Ultra 5 238V averages 21,981 across all recorded tests, while the EPYC 9454 averages 21,223. However, the Intel chip has many more PassMark tests recorded, which skews its average upward. The head-to-head Cinebench tests are the more direct comparison.

Specification Differences

| Specification | Intel Core Ultra 5 238V | AMD EPYC 9454 |

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

| Cores | 8 | 48 |

| Threads | 8 | 96 |

| Base Clock | 2.10 GHz | 2.75 GHz |

| Boost Clock | 4.70 GHz | 3.80 GHz |

| TDP | 17 W | 290 W |

| Socket | Intel BGA 2833 | AMD Socket SP5 |

| Architecture | Lunar Lake | Zen 4 |

| Process Node | 3 nm | 5 nm |

| Transistors | Not recorded | 52,560 million |

| Die Size | Not recorded | 8x 72 mm² |

| L1 Cache | 192 KB (per core) | 64 KB (per core) |

| L2 Cache | 2.5 MB (per core) | 1 MB (per core) |

| L3 Cache | 8 MB (shared) | 256 MB (shared) |

| Memory Support | Depends on motherboard | DDR5 |

| Memory Bus | Dual-channel | Twelve-channel |

| Memory Bandwidth | Not recorded | 460.8 GB/s |

| ECC Memory | No | Yes |

| PCIe | Gen 5, 4 Lanes | Gen 5, 128 Lanes |

| Integrated Graphics | Arc 130V | None |

| Market Segment | Mobile | Server/Workstation |

| Release Date | 2024-09-23 | 2022-11-09 |

| Launch MSRP | Not recorded | $5,225 |

| Part Number | SRPN5SRPN4 | 100-100000478 |