AMD Ryzen AI Embedded P164 vs Intel Core Ultra 9 386H Comparison

FAQ

Q: Which processor has the higher overall benchmark average?

A: The AMD Ryzen AI Embedded P164 records an average benchmark score of 52901, while the Intel Core Ultra 9 386H records 43210. Despite this, the Intel part wins 10 of the 11 head-to-head comparisons, a contradiction that reflects the different benchmark suites used for each unit.

Q: How does the AMD part compare to its nearest rivals?

A: The AMD Ryzen AI Embedded P164 sits at the 91st percentile of all CPUs. Its average score of 52901 is 0.1% above the AMD Ryzen 5 9500F (52873), 0.1% below the Intel Xeon 634 (52974), 0.6% below the AMD EPYC 7313P (53206), and 0.7% below the AMD Ryzen 9 7900X (53288).

Q: Where does the Intel part rank among all CPUs?

A: The Intel Core Ultra 9 386H sits at the 88th percentile. Its average score of 43210 is 0.3% below the AMD Ryzen AI Max PRO 385 (43326), 0.5% below the AMD Ryzen AI 9 465 (43431), 0.7% above the Intel Core i9-12900 (42906), and 0.9% above the Intel Core i9-12900KF (42830).

Q: What is the largest single-test margin between the two?

A: The largest margin is in PassMark find prime numbers, where the Intel Core Ultra 9 386H scores 341 versus the AMD's 71, a 79.2% advantage. The second-largest is in physics, with Intel at 3028 versus AMD at 1210, a 60% gap.

Q: Does the AMD processor win any head-to-head test?

A: Yes, exactly one. The AMD Ryzen AI Embedded P164 wins PassMark integer math with 87940 against the Intel's 87284, a slim 0.8% margin. All other ten tests go to Intel.

Q: What is the single-thread performance difference?

A: The Intel Core Ultra 9 386H leads single-thread performance with 4218 versus the AMD's 4029, a 4.5% advantage. Both processors are mobile parts, but the Intel unit's higher per-core efficiency shows in this metric.

Architecture Differences

The AMD Ryzen AI Embedded P164 and Intel Core Ultra 9 386H diverge sharply in their internal designs. The AMD part uses the Gorgon Point codename and belongs to the Ryzen AI Embedded generation built on Zen 5 / Zen 5c cores. It is manufactured on a 4 nm process at TSMC, with a die size of 233 mm². The Intel part uses the Panther Lake codename, belongs to the Core Ultra Series 3, and is built on a 3 nm process at Intel's own foundry. The process node difference gives Intel a density advantage, though the AMD part's larger die suggests a different core complex layout.

Core counts differ significantly. The AMD processor has 8 cores and 16 threads, while the Intel processor has 16 cores and 16 threads. This means the Intel part has twice as many physical cores but no hyperthreading, while the AMD part relies on simultaneous multithreading to reach its thread count. The cache hierarchy also differs: the AMD part has 80 KB of L1 per core, 1 MB of L2 per core, and 8 MB of L3. The Intel part has 192 KB of L1 per core, 2.5 MB of L2 per core, and 18 MB of shared L3. Intel's larger cache allocation likely contributes to its strong performance in compression and encryption workloads.

Clock speeds are close. The AMD part has a base clock of 2.00 GHz and a boost clock of 5.00 GHz. The Intel part has a base clock of 2.10 GHz and a boost clock of 4.90 GHz. The AMD part boosts slightly higher, but the Intel part runs faster at base. Thermal design power favors Intel at 25 watts versus AMD's 28 watts, a modest difference for mobile platforms.

Memory support is identical in type: both support DDR5 and LPDDR5X with dual-channel buses. However, the Intel part has a higher memory bandwidth at 115.2 GB/s versus the AMD part's 89.6 GB/s. The AMD part supports ECC memory, while the Intel part does not. PCIe connectivity differs: the AMD part offers Gen 4 with 16 CPU lanes, while the Intel part offers Gen 5 with 12 CPU lanes. Integrated graphics also differ, with the AMD part using Radeon 880M and the Intel part using Intel Xe3 Graphics.

The socket and packaging are entirely different. The AMD part uses AMD Socket FP8, while the Intel part uses Intel BGA 2540. Both are mobile segments and currently in active production. The Intel part has a part number (SA4R5Q9EH), while the AMD part's part number is unknown. Neither has a launch MSRP in the database.

Head-to-Head Benchmarks

The head-to-head comparison reveals a dominant pattern. The Intel Core Ultra 9 386H wins 10 of 11 tests, with the AMD Ryzen AI Embedded P164 taking only one. The data shows Intel's advantages are largest in compute-heavy and encryption workloads, while AMD's single win is narrow.

The most striking result is PassMark find prime numbers. Intel scores 341 against AMD's 71, a 79.2% advantage. This test measures raw integer arithmetic and loop efficiency, and the Intel part's 16 physical cores clearly outperform AMD's 8 cores with simultaneous multithreading. The physics test shows a similar pattern: Intel scores 3028 against AMD's 1210, a 60% gap. Physics simulations are heavily parallel, and Intel's core count advantage dominates.

Floating point math shows Intel at 108527 versus AMD's 55799, a 48.6% margin. This is a large absolute difference and suggests Intel's floating-point units are substantially more efficient per clock or that the core count advantage carries through. Data encryption is another major win: Intel scores 27150 versus AMD's 16055, a 40.9% gap. Encryption workloads often benefit from dedicated instruction sets, and Intel's implementation appears stronger.

Multithread performance gives Intel 35399 against AMD's 25889, a 26.9% margin. This is the broadest measure of parallel throughput and confirms the core count advantage. Extended instructions show Intel at 29138 versus AMD's 24193, a 17% gap. Random string sorting has Intel at 42135 versus AMD's 34801, a 17.4% margin. Data compression gives Intel 352365 versus AMD's 327891, a 6.9% advantage, which is smaller than the other workloads but still favors Intel.

Single-thread performance is closer. Intel scores 4218 versus AMD's 4029, a 4.5% margin. This indicates that Intel's per-core architecture is slightly more efficient, but the gap is modest. The only AMD win is integer math: AMD scores 87940 versus Intel's 87284, a 0.8% margin. This is a narrow victory and may reflect AMD's Zen 5 integer pipeline efficiency or better thread scheduling.

The data indicates a clear pattern: Intel dominates in every workload that scales with core count or benefits from its larger cache, while AMD's sole win is a marginal integer math result. The average benchmark scores tell a different story because they include different test suites, but the head-to-head results are unambiguous.

The Verdict

The benchmark data supports a clear division of strengths. The Intel Core Ultra 9 386H outperforms the AMD Ryzen AI Embedded P164 in 10 of 11 shared tests, with margins ranging from 4.5% in single-thread to 79.2% in prime number finding. The Intel part also has a higher memory bandwidth (115.2 GB/s versus 89.6 GB/s), a larger L3 cache (18 MB versus 8 MB), and a more advanced process node (3 nm versus 4 nm). For workloads such as physics simulation, encryption, floating-point math, and multithreaded applications, the Intel part is the stronger choice.

The AMD Ryzen AI Embedded P164 wins only in integer math, with a 0.8% margin. It offers ECC memory support, which the Intel part lacks, and it has a higher boost clock (5.00 GHz versus 4.90 GHz). The AMD part also records a higher average benchmark score (52901 versus 43210) and a higher percentile ranking (91st versus 88th). However, these average scores include separate benchmark suites, so they do not directly reflect the head-to-head results.

The data suggests the Intel part is better for compute-intensive parallel tasks, while the AMD part is better for ECC-dependent embedded applications. The AMD part's lower TDP (28 watts versus 25 watts for Intel) is slightly higher, so power efficiency favors Intel. The Intel part's PCIe Gen 5 support offers newer connectivity, while the AMD part's Gen 4 with more CPU lanes (16 versus 12) provides more lanes. The choice depends on whether the workload prioritizes raw parallel performance (Intel) or ECC reliability and higher boost clocks (AMD).

Specification Differences

The two processors differ in several key specification fields. The AMD Ryzen AI Embedded P164 has 8 cores and 16 threads, while the Intel Core Ultra 9 386H has 16 cores and 16 threads. Base clocks are 2.00 GHz for AMD and 2.10 GHz for Intel. Boost clocks are 5.00 GHz for AMD and 4.90 GHz for Intel. TDP is 28 watts for AMD and 25 watts for Intel.

The sockets differ: AMD uses Socket FP8, Intel uses BGA 2540. The process nodes differ: AMD uses 4 nm at TSMC, Intel uses 3 nm at Intel's foundry. The AMD die size is 233 mm², while the Intel die size is not recorded. Cache configurations differ: AMD has 80 KB L1 per core, 1 MB L2 per core, and 8 MB L3; Intel has 192 KB L1 per core, 2.5 MB L2 per core, and 18 MB shared L3.

Memory bandwidth differs: AMD at 89.6 GB/s, Intel at 115.2 GB/s. ECC support is present on AMD, absent on Intel. PCIe differs: AMD has Gen 4 with 16 CPU lanes, Intel has Gen 5 with 12 CPU lanes. Integrated graphics differ: AMD uses Radeon 880M, Intel uses Intel Xe3 Graphics. The Intel part has a part number (SA4R5Q9EH), while the AMD part's is unknown. Release dates differ: AMD releases on 2026-03-08, Intel on 2026-01-04. Neither has a launch MSRP.

Where Each One Wins

The Intel Core Ultra 9 386H wins in the majority of benchmark categories. Its largest margins are in prime number finding (79.2% ahead), physics (60% ahead), floating-point math (48.6% ahead), and encryption (40.9% ahead). These results indicate a clear advantage for scientific computing, financial modeling, cryptography, and any workload that stresses parallel arithmetic. The Intel part also wins in multithread performance (26.9% ahead), extended instructions (17% ahead), random string sorting (17.4% ahead), and data compression (6.9% ahead). Its single-thread advantage of 4.5% covers general responsiveness and lightly threaded applications.

The AMD Ryzen AI Embedded P164 wins in integer math by 0.8%, a narrow but real margin. This suggests a slight edge in workloads dominated by integer operations, such as database indexing or certain types of data processing. The AMD part also offers ECC memory support, which is not present on the Intel part. This makes the AMD processor suitable for embedded systems that require error-correcting memory for data integrity. The AMD part's higher average benchmark score (52901 versus 43210) and higher percentile (91st versus 88th) indicate that in its own broader test suite, it performs competitively against a wider range of CPUs.

The data does not support a universal recommendation. For parallel compute throughput, memory bandwidth, and cache-heavy workloads, the Intel part is consistently ahead. For ECC-enabled embedded deployments and integer-focused tasks, the AMD part has specific advantages. The recorded benchmarks show a trade-off between raw performance and feature-specific reliability, with Intel leading in performance metrics and AMD leading in memory integrity features.