AMD Ryzen AI 5 440G vs Intel Core Ultra 5 338H Comparison

The Verdict

The recorded data separates these two processors into distinct roles. The AMD Ryzen AI 5 440G wins only two of the eleven head-to-head benchmark comparisons, while the Intel Core Ultra 5 338H takes the remaining nine. The Intel part posts a higher PassMark multithread score of 28717 against 23487 for the AMD, a gap of 18.2 percent. The AMD chip counters with a 9.7 percent win in integer math, scoring 71251 versus 64934, and a 5.9 percent win in data compression, 292735 versus 276539. The overall database percentile places the AMD at 89 percent of all CPUs, while the Intel sits at 84 percent, yet the Intel part carries a higher average benchmark score in its own rival group. The data suggests the Intel Core Ultra 5 338H delivers broader performance across most measured workloads, while the AMD Ryzen AI 5 440G holds specific advantages in integer-heavy and compression tasks.

The AMD part targets the desktop segment with an AMD Socket AM5, a 65 TDP, and an unlocked multiplier, which indicates a user-installable, overclockable desktop processor. The Intel part is a mobile chip on Intel BGA 2540 with a 25 TDP and a locked multiplier, designed for thin-and-light or mainstream laptops. The choice between them depends entirely on platform: the data does not support a single universal winner, it supports a desktop part with narrow but real wins and a mobile part with wider general performance.

Where Each One Wins

The AMD Ryzen AI 5 440G wins only where its architecture favors integer throughput and compression. Its PassMark integer math score of 71251 beats the Intel score of 64934 by 9.7 percent. Its data compression result of 292735 beats 276539 by 5.9 percent. These two wins align with the AMD chip's 6 cores and 12 threads, where each core carries 80 KB of L1 cache and 1 MB of L2 cache, and the shared L3 is 8 MB. The smaller core count with higher per-core cache appears to translate into efficient integer execution and compression work.

The Intel Core Ultra 5 338H wins every other measured category. Its multithread score leads by 18.2 percent, 28717 versus 23487. The floating point math gap is the largest single delta, with Intel scoring 84067 against 45711, a 45.6 percent advantage. Physics simulation shows Intel ahead by 50.7 percent, 2697 versus 1329. Prime number finding shows Intel at 304 versus 90, a 70.4 percent lead. Data encryption favors Intel by 36.4 percent, 21367 versus 13585. Extended instructions favor Intel by 13.6 percent, 23906 versus 20648. Random string sorting favors Intel by 8.7 percent, 34082 versus 31106. Even single-thread performance, often a desktop strength, favors Intel by 2.9 percent, 4180 versus 4060.

The Intel part achieves these wins with 12 cores and 12 threads, no hyperthreading, larger per-core L1 at 192 KB and L2 at 2.5 MB, plus a shared 18 MB L3. The physical core count advantage appears decisive in multithread, floating point, and physics workloads. The AMD part cannot match the core count, and its wins are confined to integer math and compression, where its architecture and cache arrangement seem more efficient per thread.

Architecture Differences

The two processors come from different foundries and process nodes. The AMD Ryzen AI 5 440G uses a 4 nm process from TSMC, with a codename of Gorgon Point and a generation label of Ryzen AI 400 (Zen 5 / Zen 5c). The Intel Core Ultra 5 338H uses a 3 nm process from Intel's own foundry, with the Panther Lake architecture and a generation label of Ultra 5 (Panther Lake-H). The process node difference, 4 nm versus 3 nm, does not by itself explain the benchmark deltas, but it does correlate with the Intel part's lower 25 TDP versus the AMD's 65 TDP.

Core counts differ significantly. The AMD chip has 6 cores and 12 threads, indicating simultaneous multithreading. The Intel chip has 12 cores and 12 threads, indicating no SMT but double the physical core count. Cache layouts differ as well: the AMD provides 80 KB L1 per core, 1 MB L2 per core, and 8 MB shared L3. The Intel provides 192 KB L1 per core, 2.5 MB L2 per core, and 18 MB shared L3. The Intel part has more cache at every level, which aligns with its wins in floating point math and physics.

Memory support diverges. The AMD supports DDR5 over a dual-channel bus with 89.6 GB/s bandwidth and ECC memory support. The Intel supports LPDDR5X over a dual-channel bus with 136.5 GB/s bandwidth and no ECC. The Intel memory bandwidth is 52.4 percent higher, a plausible contributor to its multithread and floating point leads. PCIe generations differ: the AMD uses Gen 4 with 12 CPU lanes, while the Intel uses Gen 5 with 4 CPU lanes. The Intel has a newer PCIe generation but far fewer lanes, a tradeoff that matters for desktop expansion versus mobile integration.

Integrated graphics differ. The AMD uses a Radeon 840M, while the Intel uses an Arc B370. No graphics benchmarks appear in the database, so the comparison cannot extend to iGPU performance. The AMD part has an unlocked multiplier, the Intel is locked. The AMD socket is AM5, the Intel is BGA 2540. The AMD die size is recorded at 195 mm², while the Intel die size is not recorded. Release dates place the Intel part about two months earlier, with the AMD launching on 2026-02-28 and the Intel on 2026-01-04.

FAQ

Q: Which processor has the higher PassMark multithread score?

A: The Intel Core Ultra 5 338H scores 28717, which is 18.2 percent higher than the AMD Ryzen AI 5 440G score of 23487.

Q: Does the AMD chip win any benchmark categories?

A: Yes. The AMD wins passmark_integer_math with 71251 versus 64934, a 9.7 percent lead, and passmark_data_compression with 292735 versus 276539, a 5.9 percent lead.

Q: What explains the Intel part's large floating point advantage?

A: The Intel scores 84067 in passmark_floating_point_math against 45711 for the AMD, a 45.6 percent lead. The Intel part has 12 physical cores versus 6, a larger 18 MB shared L3 versus 8 MB, and higher memory bandwidth at 136.5 GB/s versus 89.6 GB/s.

Q: Are both processors on the same process node?

A: No. The AMD uses a 4 nm TSMC process, while the Intel uses a 3 nm process from Intel's own foundry.

Q: Which processor supports ECC memory?

A: The AMD Ryzen AI 5 440G supports ECC memory, while the Intel Core Ultra 5 338H does not.

Q: How do the CPU percentiles compare?

A: The AMD sits at the 89th percentile of all CPUs, while the Intel sits at the 84th percentile. The AMD's higher percentile coexists with lower head-to-head benchmark scores because the AMD's average benchmark score of 46187 in its rival group exceeds the Intel's average of 33989 in its own rival group.

Head-to-Head Benchmarks

The largest Intel win appears in passmark_find_prime_numbers. The Intel score of 304 dwarfs the AMD score of 90, a 70.4 percent delta. This workload rewards raw core count and cache, and the Intel part's 12 physical cores and 18 MB L3 dominate the AMD's 6 cores and 8 MB L3. The second-largest Intel win is passmark_physics, where Intel scores 2697 versus 1329, a 50.7 percent lead. Physics simulation typically scales with multithread throughput, and the Intel multithread score of 28717 confirms that pattern.

Floating point math shows a 45.6 percent Intel lead, 84067 versus 45711. This is the largest absolute score gap in the dataset. The Intel's higher memory bandwidth, 136.5 GB/s versus 89.6 GB/s, likely feeds this workload more effectively. Data encryption favors Intel by 36.4 percent, 21367 versus 13585. This workload often relies on AES instructions and memory bandwidth, both of which favor the Intel part's newer architecture and wider memory path.

The AMD's wins are narrower but real. Integer math goes to AMD by 9.7 percent, 71251 versus 64934. Data compression goes to AMD by 5.9 percent, 292735 versus 276539. These two wins suggest the AMD's Zen 5 / Zen 5c cores execute integer operations and compression algorithms more efficiently per thread, even with fewer cores. The AMD also avoids a loss in random string sorting by a modest margin, but still loses that test by 8.7 percent, 34082 versus 31106.

Single-thread performance goes to Intel by a slim 2.9 percent, 4180 versus 4060. This is the closest contest in the dataset. The Intel part's larger per-core L1 and L2 caches, 192 KB and 2.5 MB versus 80 KB and 1 MB, appear to provide a small but consistent single-thread edge. Extended instructions favor Intel by 13.6 percent, 23906 versus 20648, a category that includes SIMD and specialized instruction throughput.

The multithread test, which is often the headline comparison, favors Intel by 18.2 percent, 28717 versus 23487. This aligns with the core count difference: 12 physical cores versus 6 cores with SMT. The AMD's SMT gives it 12 threads, matching the Intel's thread count, but the Intel's physical cores deliver more sustained throughput in the aggregate. The data compression test is the only multithread-adjacent category where the AMD pulls ahead, which reinforces the interpretation that the AMD's efficiency advantage is workload-specific rather than general.

Specification Differences

The two processors differ across nearly every recorded specification field. The AMD has 6 cores and 12 threads; the Intel has 12 cores and 12 threads. Base clocks are close, with AMD at 2.00 GHz and Intel at 1.90 GHz. Boost clocks are also close, with AMD at 4.80 GHz and Intel at 4.70 GHz. The TDP differs substantially: AMD at 65, Intel at 25. The AMD uses AMD Socket AM5; the Intel uses Intel BGA 2540. The AMD architecture is recorded as Gorgon Point with a generation of Ryzen AI 400 (Zen 5 / Zen 5c); the Intel architecture is Panther Lake with a generation of Ultra 5 (Panther Lake-H).

Process node and foundry differ: AMD on 4 nm at TSMC, Intel on 3 nm at Intel. Die size is recorded for AMD at 195 mm², while Intel's die size is not recorded. Cache differs at every level: AMD L1 is 80 KB per core, Intel L1 is 192 KB per core; AMD L2 is 1 MB per core, Intel L2 is 2.5 MB per core; AMD L3 is 8 MB, Intel L3 is 18 MB shared. Memory support differs: AMD uses DDR5, Intel uses LPDDR5X. Memory bandwidth differs: AMD at 89.6 GB/s, Intel at 136.5 GB/s. ECC support is present on AMD, absent on Intel. PCIe differs: AMD uses Gen 4 with 12 CPU lanes, Intel uses Gen 5 with 4 CPU lanes. Integrated graphics differ: AMD Radeon 840M versus Intel Arc B370. Market segment differs: AMD is Desktop, Intel is Mobile. The AMD multiplier is unlocked, the Intel is locked. Part numbers differ: AMD 100-000001918, Intel SA4REQ9EW. Release dates differ: AMD 2026-02-28, Intel 2026-01-04. Neither part has a recorded launch MSRP.