Intel Core Ultra 7 265H vs Intel Xeon 6357P Comparison

The database pairs two very different Intel processors here: the Core Ultra 7 265H, a mobile Arrow Lake-H part built for thin machines, and the Xeon 6357P, a Raptor Lake-R workstation chip aimed at socket platforms. Despite targeting different markets, their aggregate results land almost on top of each other (an average benchmark score of 41621 for the Ultra 7 against 40630 for the Xeon), which makes the head-to-head data unusually instructive. The recorded results split 12 wins to 5 in favor of the Core Ultra 7 265H, but the Xeon takes some of the most visible single-core tests, so the correct pick depends heavily on workload shape.

FAQ

Q: Which CPU wins the overall head-to-head?

A: The Core Ultra 7 265H takes 12 of 17 recorded benchmarks against 5 for the Xeon 6357P, and holds the higher average benchmark score, 41621 versus 40630.

Q: Which one is faster in single-core workloads?

A: It depends on the test. The Xeon leads in Cinebench R15 single-core (372 versus 307) and dominates Cinebench R23 single-core (3691 versus 2080), while the Ultra 7 265H edges ahead in Cinebench R20 single-core (1712 versus 1550) and PassMark single thread (4334 versus 4233).

Q: Which CPU is more power efficient on paper?

A: The Core Ultra 7 265H carries a 28 W TDP versus 80 W for the Xeon 6357P, a substantial gap that reflects the mobile-versus-workstation split between the two parts.

Q: Do both processors support ECC memory?

A: Yes, both list ECC memory support in the database, which is notable for a mobile-class chip.

Q: Which platform offers more PCIe connectivity?

A: The Xeon 6357P provides Gen 5 with 16 CPU lanes, double the 8 Gen 5 lanes listed for the Core Ultra 7 265H.

Q: How do the two rank against all CPUs in the database?

A: The Core Ultra 7 265H sits in the 88th percentile; the Xeon 6357P sits in the 87th, effectively the same tier.

Architecture Differences

These two parts come from opposite ends of Intel's recent product stack. The Core Ultra 7 265H is an Arrow Lake-H design, fabricated by TSMC on a 3 nm process node, part of the Core Ultra Series 2 and released in January 2025. The Xeon 6357P is a Raptor Lake-R design on Intel's own 10 nm process, released in February 2025 for the server and workstation segment.

The core topologies diverge sharply. The Ultra 7 265H offers 16 cores with 16 threads, one thread per core. The Xeon 6357P offers 8 cores with 16 threads via simultaneous multithreading, giving both chips the same 16-thread capacity but reached through different means. That structural difference explains much of the benchmark spread: parallel workloads behave differently when spread across 16 physical cores versus 8 cores each running two threads.

Cache configurations also differ. The Ultra 7 265H carries 192 KB of L1 per core and 3 MB of L2 per core; the Xeon carries 80 KB of L1 per core and 2 MB of L2 per core. Both share an identical 24 MB of L3 cache.

The Ultra 7 265H includes integrated Arc Graphics 140T, while the Xeon 6357P has none and requires a discrete GPU. Memory support splits as well: the Ultra 7 takes DDR5 and LPDDR5X on a dual-channel bus with 102.4 GB/s of listed bandwidth, while the Xeon accepts DDR4 or DDR5, also dual-channel, with no bandwidth figure recorded. Both chips have locked multipliers. The Xeon's die size is recorded at 257 mm²; no die size is listed for the Ultra 7.

Where Each One Wins

The Core Ultra 7 265H owns throughput-oriented compute. It wins floating point math by 46.6 percent, encryption by 41.9 percent, prime number search by 124.8 percent, extended instructions by 9.5 percent, string sorting by 14.8 percent, physics by 8.3 percent, and overall PassMark multithread by 10.6 percent. It also takes Cinebench R15 multi-core by 13.4 percent and R20 multi-core by 10.5 percent. Any workload that scales across many threads, or that leans on floating point and encryption, favors the mobile chip decisively.

The Xeon 6357P wins where raw per-thread speed and specific integer paths matter. Its strongest results are Cinebench R23 single-core, where it leads by 43.6 percent, Cinebench R15 single-core at 17.5 percent ahead, integer math at 12.2 percent ahead, and data compression at 5.3 percent ahead. It also takes the modern Cinebench R23 multi-core test by 23.7 percent, suggesting that under the sustained, well-threaded rendering load of R23 the Xeon's higher boost clock and threaded cores pull ahead of the older-generation R15 and R20 runs.

Context from the rivals list reinforces how close these two are in overall standing. The Ultra 7 265H's nearest rivals include the Intel Core 7 251TE (-0.1 percent), Core i7-14650HX (+0.1 percent), Core i7-12850HX (-0.4 percent), and AMD Ryzen 9 5900X (+0.6 percent). The Xeon's neighbors are the Intel Core 5 223PE (0.1 percent), Core 7 253PE (0.2 percent), AMD Ryzen AI 5 PRO 435G (-0.2 percent), and Intel Core Ultra X7 368H (0.3 percent). Both chips sit in crowded company at nearly identical performance tiers.

Specification Differences

• Cores/threads: Ultra 7 265H has 16 cores and 16 threads; Xeon 6357P has 8 cores and 16 threads.
• Clocks: Ultra 7 265H runs a 2.20 base and 5.30 boost; Xeon 6357P runs a 3.00 base and 5.40 boost.
• TDP: 28 W for the Ultra 7 265H versus 80 W for the Xeon 6357P.
• Socket: Intel BGA 2049 (soldered mobile) versus Intel Socket 1700.
• Process: 3 nm at TSMC versus 10 nm at Intel.
• Cache: 192 KB L1 and 3 MB L2 per core on the Ultra 7 versus 80 KB L1 and 2 MB L2 per core on the Xeon; 24 MB shared L3 on both.
• Memory: DDR5/LPDDR5X with 102.4 GB/s bandwidth on the Ultra 7 versus DDR4/DDR5 on the Xeon; both dual-channel with ECC.
• PCIe: Gen 5 with 8 CPU lanes on the Ultra 7 versus Gen 5 with 16 CPU lanes on the Xeon.
• Graphics: Arc Graphics 140T integrated on the Ultra 7; none on the Xeon.
• Segment: Mobile versus Server/Workstation.
• Launch MSRP: recorded only for the Xeon 6357P, at $556.

Head-to-Head Benchmarks

Starting with Cinebench, the results flip by version. The Ultra 7 265H wins R15 multi-core 2989 to 2635 and R20 multi-core 12131 to 10980, but the Xeon 6357P reverses the picture in R23 multi-core, 26145 to 19940, a 23.7 percent margin. Single-core results are equally split: the Xeon takes R15 single-core 372 to 307 and R23 single-core 3691 to 2080, the largest gap in the dataset at 43.6 percent, while the Ultra 7 takes R20 single-core 1712 to 1550.

The PassMark suite tilts heavily toward the mobile chip. Its biggest win is find prime numbers, 335 versus 149, a 124.8 percent blowout. Floating point math goes to the Ultra 7 109123 to 74460, and encryption 26005 to 18324. Multithread lands at 34027 versus 30759, physics at 2497 versus 2305, extended instructions 26805 versus 24490, string sorting 40742 versus 35495, and single thread 4334 versus 4233. The Xeon answers in integer math, 97375 versus 85479, and data compression, 353521 versus 334711.

The pattern across the data is consistent: the 16-physical-core Arrow Lake design dominates workloads that scale with cores and with floating point or encryption throughput, while the higher-clocking, SMT-based Raptor Lake part pulls ahead in per-thread rendering performance and a subset of integer-bound tasks.

The Verdict

The data supports a clear workload split. Buyers running heavily threaded compute, floating point work, encryption, or physics simulation should choose the Core Ultra 7 265H: it wins 12 of 17 recorded tests, posts the higher average score, and does so within a 28 W envelope on a soldered BGA 2049 mobile platform with integrated Arc Graphics 140T and 102.4 GB/s of memory bandwidth.

The Xeon 6357P is the pick when single-thread responsiveness dominates, when DDR4 compatibility matters, when twice the Gen 5 lanes are needed for expansion, or when a Socket 1700 workstation platform with ECC is the target. Its 43.6 percent R23 single-core win and its R23 multi-core lead show it holds its own in modern rendering loads, and its 87th percentile placement trails the Ultra 7's 88th by barely a step. Neither chip is unlocked for overclocking, so the recorded numbers are effectively the ceiling for both.