Intel Core Ultra 9 285HX vs Intel Xeon 638 Comparison

Head-to-Head Benchmarks

The head-to-head comparison between the Intel Xeon 638 and the Intel Core Ultra 9 285HX reveals a split personality. The Xeon 638 wins 7 of the 17 recorded tests, while the Core Ultra 9 285HX takes 10. The margin of victory tells the real story: the Xeon dominates in several single-thread and cache-sensitive workloads, while the Ultra 9 excels in floating-point and encryption tasks.

The most striking result is Cinebench R23 single-core, where the Xeon 638 scores 6663 against the Ultra 9's 2187.5, a 204.6% advantage. This is not a small gap; it is a generational statement about the Xeon's per-core efficiency. The same pattern appears in Cinebench R15 single-core, where the Xeon leads 671 to 323.5, a 107.4% delta. These are the largest wins in either direction across the entire benchmark suite.

Multi-core results are more nuanced. In Cinebench R23 multi-core, the Xeon 638 scores 47202 versus 36429.5 for the Ultra 9, a 29.6% lead. However, in Cinebench R15 multi-core, the Ultra 9 wins 5656.5 to 4757, a 15.9% margin. Cinebench R20 multi-core is nearly tied: the Ultra 9 edges ahead 20236 to 19824, a 2% difference. The database shows that the Xeon's advantage grows as the workload becomes more sustained and cache-heavy, while the Ultra 9 handles shorter multi-threaded bursts better.

PassMark results further separate the two. The Xeon 638 wins data compression (725818 to 631885, 14.9%), extended instructions (56498 to 49148, 15%), integer math (184884 to 155076, 19.2%), and physics (4704 to 3476, 35.3%). The Ultra 9 counters with data encryption (48567 to 36030, 25.8%), floating-point math (194998 to 144757, 25.8%), find prime numbers (460 to 381, 17.2%), and random string sorting (77196 to 74318, 3.7%). The single-thread PassMark test goes to the Ultra 9, 4618 to 3670, a 20.5% margin, which contrasts sharply with the Cinebench single-core results.

The aggregate picture: the Xeon 638 has a higher average benchmark score of 80723 compared to the Ultra 9's 76155, a 5.9% difference. Both sit at the 95th percentile of all CPUs in the database, meaning they are elite performers, but they achieve that status through different strengths.

Architecture Differences

The architectural split between these two processors is fundamental. The Intel Xeon 638 uses Granite Rapids architecture on a 5 nm process, built by Intel. The Intel Core Ultra 9 285HX uses Arrow Lake-HX on a 3 nm process, fabricated by TSMC. The die sizes reflect this: the Xeon measures 598 mm², while the Ultra 9 is 243 mm². The Ultra 9 packs 17,800 million transistors, a figure the database does not provide for the Xeon.

Core configurations differ sharply. The Xeon 638 has 16 cores and 32 threads, meaning hyperthreading is enabled. The Ultra 9 has 24 cores and 24 threads, with no hyperthreading. This explains why the Xeon can maintain strong multi-threaded scores despite fewer physical cores. The Xeon's base clock is 3.20 GHz with a boost of 4.80 GHz. The Ultra 9 runs a lower base of 2.80 GHz but boosts higher to 5.50 GHz.

Cache hierarchies tell a story of different design goals. The Xeon 638 offers 112 KB of L1 per core, 2 MB of L2 per core, and a massive 72 MB of shared L3. The Ultra 9 has 192 KB of L1 per core, 3 MB of L2 per core, but only 36 MB of shared L3. The Xeon's 72 MB L3 is double the Ultra 9's, which directly supports its wins in data compression and extended instructions, workloads that benefit from large resident data sets.

Memory architecture reinforces the server versus mobile positioning. The Xeon 638 supports quad-channel DDR5 with 204.8 GB/s of bandwidth. The Ultra 9 is dual-channel with 102.4 GB/s, exactly half. Both support ECC memory, a rare feature for a mobile processor. PCIe lanes also diverge: the Xeon provides 80 Gen 5 lanes, while the Ultra 9 offers 20 Gen 5 lanes. The Xeon has no integrated graphics, while the Ultra 9 includes Arc Xe-LPG Graphics with 64 execution units.

Power envelopes are dramatically different. The Xeon 638 has a TDP of 180 W, while the Ultra 9 is rated at 55 W. This explains the Xeon's need for a server/workstation socket (Intel Socket 4710) versus the Ultra 9's mobile BGA 2114 package. The Xeon's release date is listed as 2026-02-01, while the Ultra 9 launched on 2025-01-12. The Xeon carries a launch MSRP of $899; the Ultra 9's launch MSRP is not recorded.

The Verdict

The data points to a clear division of labor. The Intel Xeon 638 is the choice for workloads that demand massive L3 cache, high memory bandwidth, and sustained integer throughput. Its 204.6% lead in Cinebench R23 single-core and 35.3% lead in PassMark physics are not anomalies; they reflect a design optimized for per-thread efficiency and data locality.

The Intel Core Ultra 9 285HX wins where floating-point math and encryption matter. Its 25.8% advantages in both PassMark floating-point math and data encryption are substantial. The 20.5% lead in PassMark single-thread suggests better raw clock speed utilization, which aligns with its 5.50 GHz boost clock.

The average benchmark scores place the Xeon 638 at 80723, slightly ahead of the Ultra 9's 76155. The Xeon's nearest rival, the AMD Ryzen AI Max+ PRO 395, scores 80762 with a 0% delta, meaning the Xeon is effectively tied with that chip. The Ultra 9's nearest rival, the AMD Ryzen 9 8945HX, scores 76212 with a -0.1% delta, also effectively tied.

For a system builder, the decision hinges on platform constraints. The Xeon 638 requires a server/workstation motherboard with Socket 4710 and can use 80 PCIe lanes. The Ultra 9 is a mobile part on BGA 2114, found in laptops or compact systems. The Xeon's 180 W TDP demands serious cooling, while the Ultra 9's 55 W TDP is far more manageable. The Xeon's 204.8 GB/s memory bandwidth versus the Ultra 9's 102.4 GB/s is decisive for memory-bound server workloads. The Ultra 9's integrated graphics may eliminate the need for a discrete GPU in some mobile configurations.

Specification Differences

| Specification | Intel Xeon 638 | Intel Core Ultra 9 285HX |

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

| Cores | 16 | 24 |

| Threads | 32 | 24 |

| Base Clock | 3.20 GHz | 2.80 GHz |

| Boost Clock | 4.80 GHz | 5.50 GHz |

| TDP | 180 W | 55 W |

| Socket | Intel Socket 4710 | Intel BGA 2114 |

| Architecture | Granite Rapids | Arrow Lake |

| Process Node | 5 nm | 3 nm |

| Foundry | Intel | TSMC |

| Die Size | 598 mm² | 243 mm² |

| Transistors | Not recorded | 17,800 million |

| L1 Cache | 112 KB per core | 192 KB per core |

| L2 Cache | 2 MB per core | 3 MB per core |

| L3 Cache | 72 MB shared | 36 MB shared |

| Memory Bus | Quad-channel | Dual-channel |

| Memory Bandwidth | 204.8 GB/s | 102.4 GB/s |

| PCIe Lanes | Gen 5, 80 lanes | Gen 5, 20 lanes |

| Integrated Graphics | N/A | Arc Xe-LPG Graphics 64EU |

| Market Segment | Server/Workstation | Mobile |

| Release Date | 2026-02-01 | 2025-01-12 |

| Launch MSRP | $899 | Not recorded |

FAQ

Q: Which processor has better single-thread performance?

A: It depends on the benchmark. The Xeon 638 wins Cinebench R23 single-core by 204.6% (6663 vs 2187.5) and Cinebench R15 single-core by 107.4% (671 vs 323.5). The Ultra 9 wins PassMark single-thread by 20.5% (4618 vs 3670). The Cinebench results suggest the Xeon's architecture extracts more work per clock in certain rendering tasks, while the Ultra 9's higher boost clock helps in PassMark's specific test.

Q: Which processor has more cores and threads?

A: The Ultra 9 has 24 cores and 24 threads. The Xeon 638 has 16 cores and 32 threads. The Xeon's hyperthreading doubles its thread count, giving it 32 threads from 16 cores. The Ultra 9 does not use hyperthreading, so its thread count equals its core count.

Q: How do the memory bandwidth figures compare?

A: The Xeon 638 offers 204.8 GB/s via quad-channel DDR5. The Ultra 9 provides 102.4 GB/s via dual-channel DDR5. The Xeon's bandwidth is exactly double the Ultra 9's, which contributes to its wins in data-heavy workloads like PassMark data compression.

Q: Do both processors support ECC memory?

A: Yes, both the Xeon 638 and the Core Ultra 9 285HX have ECC memory support recorded as true. This is notable for the Ultra 9, as ECC is uncommon in mobile processors.

Q: What is the difference in L3 cache size?

A: The Xeon 638 has 72 MB of shared L3 cache. The Ultra 9 has 36 MB of shared L3 cache. The Xeon's L3 is double the size, which likely explains its 14.9% lead in data compression and 15% lead in extended instructions.

Q: Which processor has a higher average benchmark score?

A: The Xeon 638 has an average benchmark score of 80723. The Ultra 9 has an average score of 76155. The Xeon is ahead by roughly 5.9% in this aggregate metric, and both processors sit at the 95th percentile of all CPUs in the database.

Where Each One Wins

The Intel Xeon 638 wins in scenarios that stress cache capacity, integer throughput, and sustained multi-threaded rendering. Its 72 MB L3 cache and quad-channel memory give it a decisive edge in data compression (14.9% ahead), extended instructions (15%), and integer math (19.2%). The 29.6% lead in Cinebench R23 multi-core and 35.3% lead in PassMark physics point to workloads that scale with thread count and cache residency. Server virtualization, database query processing, and scientific computing with large working sets would favor the Xeon.

The Intel Core Ultra 9 285HX wins in floating-point-heavy and encryption-focused tasks. Its 25.8% advantages in both PassMark floating-point math and data encryption are the largest Ultra 9 wins. The 17.2% lead in find prime numbers and 3.7% lead in random string sorting indicate strengths in algorithmic workloads. The 20.5% lead in PassMark single-thread suggests responsiveness in lightly threaded applications. Mobile workstations handling 3D rendering, cryptography, or financial modeling would benefit from the Ultra 9's characteristics.

The multi-threaded results are mixed. The Ultra 9 wins Cinebench R15 multi-core by 15.9% and PassMark multithread by 2.2%, but loses Cinebench R23 multi-core by 29.6%. This inconsistency suggests the Ultra 9 performs well in shorter multi-threaded bursts, while the Xeon maintains performance over longer sustained loads. The 55 W TDP of the Ultra 9 versus the 180 W TDP of the Xeon means the Ultra 9 achieves competitive multi-threaded scores with far less power draw, but the Xeon pulls ahead when the workload exceeds the Ultra 9's thermal headroom.

For PCIe expansion, the Xeon's 80 lanes versus the Ultra 9's 20 lanes is a decisive factor. Systems needing multiple GPUs, NVMe arrays, or network adapters require the Xeon's connectivity. The Ultra 9's integrated graphics and 55 W TDP make it suitable for compact, power-efficient designs where the Xeon's server platform is impractical. The database shows no scenario where both win simultaneously: it is a choice between server-grade throughput and mobile efficiency.