Intel Core Ultra 5 235A vs Intel Core Ultra 5 338H Comparison

Head-to-Head Benchmarks

The benchmark comparison between the Intel Core Ultra 5 235A and the Intel Core Ultra 5 338H is overwhelmingly one-sided. Out of 17 recorded tests, the 235A wins 16, with the 338H securing only a single victory. The average benchmark score for the 235A is 48201, placing it at the 90th percentile of all CPUs in the database, while the 338H averages 33989 and sits at the 84th percentile.

The largest margin comes in Cinebench R23 single-core, where the 235A scores 4607 against 2044 for the 338H, a 125.4% advantage. That is not a minor gap; it is a generational difference in raw single-thread capability. Multi-core performance tells a similar story in R23, with the 235A posting 32633 versus 16331, a 99.8% lead. The 235A essentially doubles the 338H in that workload.

Cinebench R20 shows a consistent pattern. Multi-core scores are 13705 for the 235A and 10213 for the 338H, a 34.2% difference. Single-core R20 also lands at 34.2%, with 1934 against 1441. In R15, the margins are 31.3% multi-core (3289 versus 2504) and 52.1% single-core (464 versus 305). The single-core deltas grow larger in older Cinebench versions, indicating that the 235A's clock advantage scales well even in lightly threaded tasks.

PassMark tests reinforce the dominance. Data compression shows 393800 for the 235A versus 276539 for the 338H, a 42.4% lead. Data encryption is 30136 against 21367, a 41% gap. Extended instructions score 31625 versus 23906, up 32.3%. Floating point math delivers 118778 against 84067, a 41.3% advantage. Integer math follows at 36.5% (88626 versus 64934). Random string sorting is 49489 versus 34082, a 45.2% gap. Find prime numbers shows 392 versus 304, a 28.9% lead. The multithread PassMark score is 38392 versus 28717, or 33.7% higher.

Single-thread PassMark is the closest mainstream test, with 4557 versus 4180, a 9% edge for the 235A. That result stands in contrast to the enormous Cinebench R23 single-core gap and suggests the two processors are closer in some everyday lightly threaded workloads than the R23 numbers imply, though the 235A still wins clearly.

The one test the 338H takes is PassMark physics, scoring 2697 against 2437 for the 235A, a 9.6% margin in favor of the mobile chip. Physics simulation in PassMark can favor certain cache or scheduling behaviors, and here the 338H's architecture appears to handle that specific workload more efficiently. It is the only recorded loss for the 235A across the entire benchmark set.

Looking at nearest rivals for context, the 235A sits within 0.6% of the AMD Ryzen 9 3900 and within 0.1% of the AMD Ryzen AI Max PRO 380. It trades blows with the Intel Core Ultra 5 245HX, which is only 0.2% higher on average. The 338H, by contrast, lands within 0.3% of the Intel Core Ultra 7 165H and 0.3% of the Intel Core i7-12800HX. These groupings place the 235A in a higher performance tier overall.

Architecture Differences

The two processors come from different Intel lineups. The 235A belongs to Core Ultra Series 2, uses the Arrow Lake architecture on the Arrow Lake-S codename, and is built on a 3 nm process by TSMC. The 338H is from Core Ultra Series 3, uses Panther Lake on the Panther Lake-H codename, and is also on a 3 nm process but manufactured by Intel's own foundry. Both use the same 3 nm node generation, but the foundry differs, and that distinction matters for how each chip is optimized.

Core counts differ significantly. The 235A has 14 cores and 14 threads, while the 338H has 12 cores and 12 threads. Neither processor uses simultaneous multithreading, so thread counts equal core counts. The 235A therefore brings two additional physical cores to any workload that scales with core count.

Cache hierarchies are not identical. Both share a 192 KB L1 per core. The L2 cache is 3 MB per core on the 235A versus 2.5 MB per core on the 338H. The 235A also has a larger shared L3 at 24 MB, while the 338H has 18 MB. The 338H has 12 cores and 12 threads, and with 2.5 MB per core, its total L2 is smaller. The larger per-core L2 and larger L3 on the 235A contribute to its advantage in memory-sensitive workloads like data compression and string sorting.

Clock speeds set the two far apart. The 235A has a base clock of 3.40 GHz and a boost clock of 5.00 GHz. The 338H has a base clock of only 1.90 GHz with a boost of 4.70 GHz. That 1.5 GHz base clock gap is substantial and explains much of the single-thread performance difference, especially in workloads that hold the processor near base clocks.

The integrated graphics differ as well. The 235A uses Arc Xe-LPG Graphics with 24 execution units. The 338H uses an Arc B370 iGPU, which is a newer and more capable graphics block. For users who rely on the integrated GPU, the 338H is likely the stronger choice, though the database does not provide direct graphics benchmarks for either chip.

Memory support also diverges. The 235A supports DDR5 with a dual-channel bus and 102.4 GB/s of memory bandwidth. The 338H supports LPDDR5X, also dual-channel, but with 136.5 GB/s of bandwidth. The mobile chip has higher peak memory bandwidth despite its lower overall CPU performance.

PCIe connectivity is another major difference. The 235A provides Gen 5 with 20 lanes from the CPU. The 338H provides Gen 5 with only 4 lanes from the CPU. That makes the 235A far better suited for desktop builds with multiple Gen 5 devices, while the 338H is constrained to a thin-and-light mobile platform.

The transistor count and die size are only recorded for the 235A, at 17,800 million transistors on a 243 mm² die. The 338H has no listed transistor count or die size in the database. The 235A is a desktop part on Intel Socket 1851, while the 338H is a mobile part on Intel BGA 2540. The market segments are fundamentally different.

Where Each One Wins

The data points to the 235A as the clear winner in almost every CPU-bound task. Rendering workloads, represented by the Cinebench R15, R20, and R23 suites, all favor the 235A by wide margins. The multi-core R23 result is particularly telling: 32633 versus 16331. Anyone running CPU rendering, video encoding, or batch processing should expect the 235A to finish roughly twice as fast in that specific Cinebench workload.

Compression and encryption tasks also strongly favor the 235A. Data compression is 42.4% faster, and data encryption is 41% faster. These are common in file archiving, backup tools, and disk encryption scenarios. The larger L3 cache and higher core count help here.

Math-intensive workloads follow the same pattern. Floating point math is 41.3% faster, integer math is 36.5% faster, and extended instructions are 32.3% faster on the 235A. Scientific computing, spreadsheet calculations, and simulation workloads would all benefit from the desktop chip.

The 338H's single win in PassMark physics is worth examining. Physics simulation in that benchmark often relies on specific instruction patterns and cache behavior. The 338H's 9.6% advantage suggests that its Panther Lake architecture handles that particular workload more efficiently, possibly due to the higher memory bandwidth of LPDDR5X or some scheduling advantage. But this is one test out of 17, and no other physics-related benchmark in the database confirms a broader trend.

For single-thread responsiveness, the 235A still wins, but the margin narrows to 9% in PassMark single-thread. That is still a meaningful advantage for everyday application responsiveness, though not as dramatic as the 125.4% gap in Cinebench R23 single-core.

Specification Differences

The two chips differ in several recorded specifications. Core count is 14 for the 235A versus 12 for the 338H. Threads match core counts at 14 and 12, respectively. Base clock is 3.40 GHz on the 235A versus 1.90 GHz on the 338H. Boost clock is 5.00 GHz versus 4.70 GHz.

TDP is a major separator: 65 watts for the 235A versus 25 watts for the 338H. That reflects the desktop versus mobile design philosophy. The 235A has more thermal headroom, allowing it to sustain higher clocks, while the 338H is built for power efficiency in laptops.

The socket differs completely: Intel Socket 1851 for the 235A, Intel BGA 2540 for the 338H. The 338H is soldered, not upgradeable in the traditional sense. Memory support differs: DDR5 for the 235A, LPDDR5X for the 338H. Memory bandwidth is 102.4 GB/s for the 235A and 136.5 GB/s for the 338H.

L2 cache is 3 MB per core on the 235A versus 2.5 MB per core on the 338H. L3 cache is 24 MB shared versus 18 MB shared. PCIe lanes are 20 on the 235A versus 4 on the 338H, both Gen 5. Integrated graphics differ: Arc Xe-LPG Graphics 24EU on the 235A versus Arc B370 on the 338H.

Production status is Active for both. The 235A has a launch MSRP of $269 and a release date of 2025-07-28. The 338H has no listed MSRP and a release date of 2026-01-04. The 235A is a desktop part, the 338H is mobile. The 235A uses TSMC as foundry, the 338H uses Intel. The process node is 3 nm for both.

FAQ

Q: Which processor has better multi-core performance?

A: The Intel Core Ultra 5 235A. In Cinebench R23 multi-core, it scores 32633 versus 16331 for the 338H, a 99.8% advantage. The gap is 34.2% in Cinebench R20 multi-core and 31.3% in Cinebench R15 multi-core.

Q: Does the Core Ultra 5 338H win any benchmark?

A: Yes, one. It wins PassMark physics with a score of 2697 against 2437 for the 235A, a 9.6% margin. All other 16 recorded tests go to the 235A.

Q: How do their single-thread scores compare?

A: The 235A wins single-thread tests across the board. In Cinebench R23 single-core, it scores 4607 versus 2044, a 125.4% lead. In PassMark single-thread, the margin is narrower at 9%, with 4557 versus 4180.

Q: What are the core and thread counts?

A: The 235A has 14 cores and 14 threads. The 338H has 12 cores and 12 threads. Neither processor supports simultaneous multithreading.

Q: Which chip has more memory bandwidth?

A: The 338H has higher peak memory bandwidth at 136.5 GB/s with LPDDR5X, while the 235A has 102.4 GB/s with DDR5. The 235A still wins most memory-sensitive benchmarks despite the lower bandwidth.

Q: What is the TDP difference?

A: The 235A has a TDP of 65 watts, while the 338H has a TDP of 25 watts. This reflects the desktop versus mobile design targets.

The Verdict

The benchmark data is unambiguous. The Intel Core Ultra 5 235A is the faster processor in nearly every measured workload. It wins 16 of 17 tests, with margins ranging from 9% in PassMark single-thread to 125.4% in Cinebench R23 single-core. Its average benchmark score of 48201 places it at the 90th percentile of all CPUs, while the 338H sits at 33989 and the 84th percentile.

The 235A is the appropriate choice for desktop users who need maximum CPU throughput in rendering, compression, encryption, and math workloads. Its 14 cores, 14 threads, 5.00 GHz boost clock, and 24 MB L3 cache deliver consistent wins across the entire Cinebench and PassMark suites. The 20 PCIe Gen 5 lanes also make it suitable for high-bandwidth desktop expansion.

The 338H is a mobile processor with a 25 watt TDP, designed for laptops where power draw and heat matter more than raw performance. Its single win in PassMark physics, combined with higher memory bandwidth and a newer Arc B370 integrated GPU, gives it some niche appeal. The data does not support choosing the 338H for CPU-heavy desktop workloads.

The 235A belongs in a performance desktop build. The 338H belongs in a thin-and-light mobile system where efficiency is the priority. The recorded scores leave no room for ambiguity about which one computes faster.