Intel Core Ultra 9 290K Plus vs Intel Xeon 638 Comparison

The Intel Core Ultra 9 290K Plus and the Intel Xeon 638 represent two distinct Intel design philosophies for high-performance computing. The Ultra 9 is a desktop-focused part built for low latency and high frequency, while the Xeon 638 is a server/workstation chip optimized for throughput and massive parallel scaling. The benchmark data reveals a clear split: the Ultra 9 dominates single-threaded and floating-point workloads, while the Xeon 638 holds advantages in specific integer and compression tasks.

Head-to-Head Benchmarks

The most decisive victories for the Intel Core Ultra 9 290K Plus come in Cinebench tests, where it wins all six rounds by a consistent margin. In Cinebench R23 multi-core, the Ultra 9 scores 51,731 against the Xeon’s 47,202, a 9.6% advantage. The single-core R23 result shows a similar 9.6% lead, with the Ultra 9 posting 7,303 versus 6,663. This pattern repeats across R15 and R20, with the Ultra 9 winning every multi-core and single-core iteration by either 9.6% or 9.7%. These results indicate that the Arrow Lake Refresh architecture delivers substantially better per-thread performance and superior multi-core scaling in rendering workloads.

The gap widens dramatically in PassMark’s floating-point math test. The Ultra 9 scores 214,760, which is 48.4% higher than the Xeon 638’s 144,757. This is the largest percentage differential in the entire comparison. The Ultra 9 also wins the PassMark single-thread test by a massive 31.4%, scoring 4,823 against 3,670. Data encryption shows another strong Ultra 9 win, with a 45.9% lead (52,563 vs. 36,030). Prime number finding favors the Ultra 9 by 32% (503 vs. 381), and multi-threaded performance goes to the Ultra 9 by 9.4% (60,860 vs. 55,651). Random string sorting also goes to the Ultra 9, with a 7.3% edge (79,744 vs. 74,318).

The Intel Xeon 638 secures four wins, and they are notable for their specificity. The largest Xeon victory is in PassMark physics, where it scores 4,704 against the Ultra 9’s 3,315, a 29.5% lead. This suggests the Xeon’s thread scheduling and memory subsystem handle physics simulation more efficiently. The Xeon also wins integer math by 10.1%, posting 184,884 versus 166,194. Data compression goes to the Xeon by 3.8% (725,818 vs. 698,346), and extended instructions favor the Xeon by 7.4% (56,498 vs. 52,338). These wins are concentrated in workloads that benefit from the Xeon’s larger L3 cache and higher TDP envelope.

FAQ

Q: Which processor wins the most benchmark comparisons?

A: The Intel Core Ultra 9 290K Plus wins 13 of the 17 head-to-head tests, while the Intel Xeon 638 wins 4.

Q: How large is the single-threaded performance gap?

A: In PassMark single-thread testing, the Ultra 9 scores 4,823 compared to the Xeon’s 3,670, giving the Ultra 9 a 31.4% advantage. Cinebench R23 single-core shows a smaller but consistent 9.6% lead (7,303 vs. 6,663).

Q: In which workload does the Xeon 638 have its biggest advantage?

A: The Xeon’s largest win is in PassMark physics, where it scores 4,704 versus 3,315, a 29.5% lead over the Ultra 9.

Q: Does the Xeon 638 outperform the Ultra 9 in any multi-core rendering test?

A: No. The Ultra 9 wins all three Cinebench multi-core tests (R15, R20, R23) by 9.6% each.

Q: What is the difference in PassMark floating-point performance?

A: The Ultra 9 scores 214,760 in floating-point math, which is 48.4% higher than the Xeon 638’s 144,757.

Q: Which processor has a higher average benchmark score?

A: The Ultra 9 has an average benchmark score of 84,003, while the Xeon 638 averages 80,723.

Architecture Differences

The two processors are built on fundamentally different silicon. The Intel Core Ultra 9 290K Plus uses a 3 nm process node from TSMC, while the Intel Xeon 638 uses a 5 nm node from Intel’s own foundry. The Ultra 9’s die size is 243 mm², considerably smaller than the Xeon’s 598 mm². The Ultra 9 packs 17,800 million transistors, whereas the Xeon’s transistor count is not listed in the available data.

Cache organization differs significantly. The Ultra 9 features 192 KB of L1 cache per core and 3 MB of L2 cache per core, with a 36 MB shared L3 cache. The Xeon 638 has smaller per-core caches—112 KB L1 and 2 MB L2—but compensates with a much larger 72 MB shared L3 cache. This doubling of L3 capacity likely explains the Xeon’s wins in integer math and data compression, where larger working sets can be held on-die.

The memory controllers also diverge. The Ultra 9 uses a dual-channel DDR5 interface with 115.2 GB/s of bandwidth, while the Xeon 638 uses a quad-channel DDR5 setup delivering 204.8 GB/s. PCIe connectivity is another major differentiator: the Ultra 9 provides 20 PCIe Gen 5 lanes from the CPU, while the Xeon 638 provides 80 Gen 5 lanes. The Ultra 9 includes integrated Arc Xe-LPG Graphics with 64 execution units, while the Xeon 638 has no integrated graphics. Both support ECC memory.

Specification Differences

The core and thread counts are a key divergence. The Ultra 9 has 24 cores and 24 threads, indicating no hyper-threading, while the Xeon 638 has 16 cores and 32 threads, meaning each core supports two threads. Clock speeds favor the Ultra 9, with a base clock of 3.70 GHz and a boost clock of 5.80 GHz, versus the Xeon’s 3.20 GHz base and 4.80 GHz boost. Power envelopes are inverted relative to performance: the Ultra 9 has a 125 W TDP, while the Xeon 638 draws 180 W.

The socket platforms are incompatible. The Ultra 9 uses Intel Socket 1851, while the Xeon 638 uses Intel Socket 4710. The Ultra 9 is part of the Core Ultra Series 2 generation with the Arrow Lake Refresh codename, while the Xeon 638 belongs to the Xeon 600 series (Granite Rapids-WS) with the Granite Rapids architecture. The Xeon 638 has a release date of February 1, 2026, and a launch MSRP of $899; the Ultra 9’s release date and MSRP are not listed. Both processors have unlocked multipliers. The Xeon 638 has a part number of SA2DN, while the Ultra 9’s part number is unknown.

Where Each One Wins

The Intel Core Ultra 9 290K Plus is the clear choice for any workload that prioritizes single-threaded speed or floating-point arithmetic. The 31.4% PassMark single-thread lead and 48.4% floating-point advantage make it superior for interactive desktop use, scientific simulations, and rendering tasks. Its consistent 9.6% wins across all Cinebench versions indicate that it is the better processor for 3D modeling and video editing. The 45.9% encryption advantage also positions it well for security-focused desktop applications. With 13 wins out of 17 tests, the Ultra 9 is the more versatile processor for general high-performance computing.

The Intel Xeon 638 wins in a narrower set of scenarios, but those scenarios are meaningful for server workloads. Its 29.5% physics advantage suggests it handles rigid-body dynamics and collision detection more efficiently. The 10.1% integer math win and 3.8% data compression win point to strengths in database operations and file archiving. The 7.4% extended instructions win indicates better support for specialized instruction sets used in enterprise software. The Xeon’s 80 PCIe Gen 5 lanes and quad-channel memory bandwidth make it the better platform for systems requiring massive I/O expansion, even though this is not directly reflected in the benchmark scores.

The Verdict

The benchmark data is unambiguous about overall performance: the Intel Core Ultra 9 290K Plus is the faster processor in the majority of tests. Its 96th percentile ranking among all CPUs, compared to the Xeon 638’s 95th percentile, reinforces this conclusion. The Ultra 9 also has a higher average benchmark score of 84,003 versus 80,723. For desktop users, content creators, and professionals running single-threaded or floating-point-heavy applications, the Ultra 9 is the superior choice based on its 9.6% Cinebench wins and 48.4% floating-point advantage.

The Intel Xeon 638 should be selected by users whose workloads specifically match its four winning categories. If physics simulation, integer math, data compression, or extended instruction sets dominate the application mix, the Xeon’s 29.5% physics lead and 10.1% integer math win are decisive. Server administrators requiring 80 PCIe Gen 5 lanes or quad-channel memory bandwidth will find the Xeon 638’s platform capabilities essential, even if it loses in raw rendering performance. The Xeon’s 72 MB L3 cache provides a capacity advantage that can benefit large data sets. Ultimately, the Ultra 9 is the better all-around processor, but the Xeon 638 has a purpose-built niche where its specific strengths outweigh its overall benchmark deficit.