AMD Ryzen AI Embedded P164 vs Intel Core Ultra 5 338H Comparison

The AMD Ryzen AI Embedded P164 and Intel Core Ultra 5 338H are both active mobile processors, but the recorded data separates them clearly. AMD fields 8 cores with 16 threads, while Intel fields 12 cores with 12 threads. Across the head-to-head PassMark tests, AMD wins 4 and Intel wins 7, yet AMD holds the higher average benchmark score at 52901 versus 33989 and the higher percentile ranking at 91st versus 84th. These two processors are not direct substitutes; they are optimized for different workload clusters.

Head-to-Head Benchmarks

AMD's biggest win is integer math. The Ryzen AI Embedded P164 scores 87940 to Intel's 64934, a 35.4 percent lead. Data compression also goes to AMD by a wide margin: 327891 versus 276539, an 18.6 percent lead. AMD's other wins are smaller: extended instructions 24193 to 23906, a 1.2 percent lead, and random string sorting 34801 to 34082, a 2.1 percent lead.

Intel's wins cover more tests and include the largest single gap. In find prime numbers, Intel scores 304 to AMD's 71, leaving AMD 76.6 percent behind. Physics follows a similar pattern: Intel 2697, AMD 1210, a 55.1 percent gap. Floating point math favors Intel 84067 to 55799, a 33.6 percent gap. Data encryption favors Intel 21367 to 16055, a 24.9 percent gap. Intel also wins multithread 28717 to 25889, a 9.8 percent gap, and single-thread 4180 to 4029, a 3.6 percent gap. The single-thread result is notable because AMD has the higher boost clock at 5.00 GHz versus 4.70 GHz.

Average scores place these wins in context. AMD's 52901 average is 0.1 percent above the Ryzen 5 9500F (52873) and 0.1 percent below the Xeon 634 (52974). Intel's 33989 average is 0.3 percent below the Core Ultra 7 165H (34083) and 0.3 percent above the Core i7-12800HX (33875). The AMD part also carries a 91st percentile ranking, while Intel sits in the 84th percentile.

Where Each One Wins

AMD's wins cluster in integer and compression work. Integer math, data compression, extended instructions, and random string sorting all favor the Ryzen AI Embedded P164. The AMD part records these wins while carrying 16 threads and a 5.00 GHz boost clock.

Intel's wins cluster in floating-point, physics, encryption, and prime number tests. Those four tests all favor the Core Ultra 5 338H, as do the PassMark multithread and single-thread totals. The Intel part also has recorded Cinebench results: Cinebench R15 multi-core 2504 and single-core 305, Cinebench R20 multi-core 10213 and single-core 1441, and Cinebench R23 multi-core 16331 and single-core 2044. No Cinebench result is recorded for the AMD part, so the Cinebench comparison cannot be extended.

Architecture Differences

The two parts come from different foundries and nodes. AMD uses a 4 nm TSMC process with a 233 mm² die, while Intel uses a 3 nm Intel process with no die size recorded. AMD's generation is Ryzen AI Embedded (Zen 5 / Zen 5c) under the Gorgon Point codename. Intel's generation is Ultra 5 (Panther Lake-H) under the Panther Lake architecture and codename.

Core and cache structures differ sharply. AMD has 8 cores and 16 threads, with 80 KB L1 and 1 MB L2 per core and 8 MB of L3. Intel has 12 cores and 12 threads, with 192 KB L1 and 2.5 MB L2 per core and 18 MB of shared L3. Intel therefore records more cache at every level, while AMD records more threads.

Memory and I/O also differ. AMD supports DDR5 and LPDDR5X through a dual-channel bus with 89.6 GB/s of bandwidth; Intel supports only LPDDR5X through a dual-channel bus with 136.5 GB/s of bandwidth. AMD has ECC memory support; Intel does not. AMD exposes PCIe Gen 4 with 16 CPU lanes; Intel exposes PCIe Gen 5 with 4 CPU lanes. Integrated graphics are Radeon 880M on AMD and Arc B370 on Intel.

Power and packaging are different as well. AMD is rated at TDP 28, runs at 2.00 GHz base and 5.00 GHz boost, and uses AMD Socket FP8. Intel is rated at TDP 25, runs at 1.90 GHz base and 4.70 GHz boost, and uses Intel BGA 2540. Both parts are locked, active, mobile processors released in 2026, with Intel dated January 4 and AMD dated March 8.

FAQ

Q: Which CPU has the higher PassMark multithread score?

A: Intel Core Ultra 5 338H scores 28717, while AMD Ryzen AI Embedded P164 scores 25889. AMD trails by 9.8 percent.

Q: Which CPU wins integer math and data compression?

A: AMD wins both. Integer math is 87940 to 64934, a 35.4 percent lead, and data compression is 327891 to 276539, an 18.6 percent lead.

Q: Does the AMD part support ECC memory?

A: Yes. The Ryzen AI Embedded P164 has ECC memory support. The Intel Core Ultra 5 338H does not.

Q: Which CPU has more memory bandwidth?

A: Intel records 136.5 GB/s versus AMD's 89.6 GB/s. Both use dual-channel memory, but Intel only supports LPDDR5X while AMD supports DDR5 and LPDDR5X.

Q: How do the average benchmark scores compare?

A: AMD averages 52901 with a 91st percentile ranking. Intel averages 33989 with an 84th percentile ranking. AMD's nearest rivals include Ryzen 5 9500F at 52873 and Xeon 634 at 52974; Intel's nearest rivals include Core Ultra 7 165H at 34083 and Core i7-12800HX at 33875.

Q: Which CPU has more cores and threads?

A: Intel has 12 cores and 12 threads. AMD has 8 cores and 16 threads. Intel has more cores, AMD has more threads.

The Verdict

The recorded data supports a workload-based split. AMD is the part to choose for integer math, data compression, extended instructions, and random string sorting, where its leads range from 1.2 percent to 35.4 percent. AMD also holds the higher average benchmark score and the higher percentile ranking, which places it in a stronger overall position in the database.

Intel is the part to choose for floating-point math, physics, prime number finding, data encryption, and the PassMark multithread and single-thread totals. Intel's wins are more numerous, and its largest gaps in prime numbers and physics are substantial. For workloads that depend on memory bandwidth or PCIe Gen 5 CPU lanes, Intel has the advantage. For workloads that require ECC memory, AMD is the only option in this pair.

Specification Differences

| Field | AMD Ryzen AI Embedded P164 | Intel Core Ultra 5 338H |

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

| Manufacturer | AMD | Intel |

| Series | Not recorded | Core Ultra Series 3 |

| Architecture | Not recorded | Panther Lake |

| Cores | 8 | 12 |

| Threads | 16 | 12 |

| Base clock | 2.00 GHz | 1.90 GHz |

| Boost clock | 5.00 GHz | 4.70 GHz |

| TDP | 28 | 25 |

| Socket | AMD Socket FP8 | Intel BGA 2540 |

| Codename | Gorgon Point | Panther Lake |

| Generation | Ryzen AI Embedded (Zen 5 / Zen 5c) | Ultra 5 (Panther Lake-H) |

| Process node | 4 nm | 3 nm |

| Foundry | TSMC | Intel |

| Die size | 233 mm² | Not recorded |

| L1 cache | 80 KB (per core) | 192 KB (per core) |

| L2 cache | 1 MB (per core) | 2.5 MB (per core) |

| L3 cache | 8 MB | 18 MB (shared) |

| Memory support | DDR5, LPDDR5X | LPDDR5X |

| Memory bandwidth | 89.6 GB/s | 136.5 GB/s |

| ECC memory | Yes | No |

| PCIe | Gen 4, 16 Lanes (CPU only) | Gen 5, 4 Lanes (CPU only) |

| Integrated graphics | Radeon 880M | Arc B370 |

| Release date | 2026-03-08 | 2026-01-04 |

| Part number | Unknown | SA4REQ9EW |