AMD Ryzen AI Embedded P164 vs Intel Core Ultra 5 225T Comparison

Where Each One Wins

The benchmark data splits the two processors into distinct roles. The AMD Ryzen AI Embedded P164 wins 5 of the 11 recorded head-to-head tests, while the Intel Core Ultra 5 225T wins 6. The nature of those wins, however, is not evenly distributed across workload types.

The AMD part dominates in data compression, extended instruction sets, integer math, multithreaded throughput, and random string sorting. Its data compression score of 327891 is 40.1% ahead of Intel's 233998, which is the largest single-test margin in either direction. Integer math shows a similar pattern, with AMD scoring 87940 against Intel's 59543, a 47.7% advantage. These are classic CPU-bound tasks that respond well to high thread counts and strong per-core integer execution.

The Intel Core Ultra 5 225T takes the other side of the ledger. It wins in data encryption, prime number finding, floating point math, physics simulation, and single threaded workloads. The most striking result is in prime number generation, where Intel scores 284 versus AMD's 71, a 75% deficit for AMD. Floating point math also goes Intel's way by 32.6%, with 82751 against 55799. Physics simulation, which often relies on floating point throughput, shows Intel ahead at 2053 versus 1210, a 41.1% margin.

The overall average benchmark scores reflect a broader gap. The AMD processor records an average benchmark score of 52901, placing it in the 91st percentile of all CPUs. The Intel part averages 30468, landing in the 82nd percentile. That difference is substantial, but it is driven largely by AMD's enormous win in data compression and integer math, not by uniform superiority across all tests.

The multithreaded result is close. AMD's 25889 edges Intel's 25358 by a narrow 2.1% margin. Single threaded performance, however, favors Intel at 4348 versus 4029, a 7.3% difference. This split creates a clear use-case separation: AMD for integer-heavy parallel workloads, Intel for floating point, encryption, and lightly threaded tasks.

The Verdict

The data supports a straightforward selection rule based on workload character. For users whose primary tasks involve data compression, integer math, extended instruction processing, or random string sorting, the AMD Ryzen AI Embedded P164 delivers measurably higher throughput. Its 40.1% lead in data compression and 47.7% lead in integer math are not marginal differences; they represent a different performance class for those operations.

For tasks that depend on floating point execution, physics simulation, encryption, or single thread responsiveness, the Intel Core Ultra 5 225T is the stronger choice. Its 32.6% advantage in floating point math and 41.1% advantage in physics are equally decisive. The encryption result, 18289 versus 16055, gives Intel a 12.2% edge. Single thread performance at 4348 versus 4029 adds another 7.3% margin in Intel's favor.

The percentile ranking complicates a simple verdict. AMD's 91st percentile placement versus Intel's 82nd suggests that on average, the AMD part outperforms a larger fraction of the CPU population. But the average is skewed by the extreme compression and integer scores. A workload that never touches those operations would not see that advantage. The Intel part's nearest rivals include the Intel Core i9-11980HK with a 0.2% delta and the AMD Ryzen 5 7640HS with a 0.3% delta, while AMD's nearest rival is the AMD Ryzen 5 9500F at a 0.1% delta. Both processors sit tightly within their respective performance peer groups.

The selection should follow the dominant workload. Mixed workloads that include both integer-heavy and floating point tasks will see either processor win roughly half the time, but the magnitude of the losses matters. AMD loses 75% in prime number generation and 32.6% in floating point math, while Intel loses 47.7% in integer math and 40.1% in data compression. The larger penalty is asymmetric: AMD's losses in Intel's strong areas are bigger than Intel's losses in AMD's strong areas, except for integer math.

Head-to-Head Benchmarks

The largest AMD victory is in integer math. The score of 87940 against Intel's 59543 represents a 47.7% advantage. This is the kind of result that shows up in general productivity, database operations, and any workload with heavy arithmetic on whole numbers. Data compression follows closely, with AMD's 327891 beating Intel's 233998 by 40.1%. Compression algorithms rely on integer operations and memory access patterns, both of which favor AMD's architecture in these measurements.

Random string sorting gives AMD a 20.9% win, scoring 34801 versus 28774. Extended instructions, which test SIMD and specialized instruction paths, show AMD ahead by 20.5% with 24193 against 20083. The multithreaded benchmark is the narrowest AMD win: 25889 versus 25358, a 2.1% margin. This indicates that overall parallel throughput is similar, and the larger AMD wins come from specific instruction types rather than general thread scaling.

Intel's largest win is in prime number generation, where the 284 score dwarfs AMD's 71. The 75% margin is the most lopsided result in the entire comparison. Prime number finding is heavily dependent on integer division and modular arithmetic, which suggests Intel's implementation of those operations is significantly more efficient. Floating point math gives Intel a 32.6% win with 82751 against 55799. Physics simulation, which uses floating point heavily, follows with a 41.1% Intel advantage, scoring 2053 versus 1210.

Data encryption favors Intel by 12.2%, with 18289 against 16055. Encryption workloads often use specialized instructions for cryptographic operations, and the result indicates Intel's implementation is faster. Single thread performance, recorded twice in the database as passmark_single_thread and passmark_singlethread, both show Intel at 4348 against AMD's 4029, a 7.3% margin. This confirms that for lightly threaded applications, Intel has a consistent but modest lead.

The win counts are close, 5 for AMD and 6 for Intel, but the score distributions are not. AMD's wins tend to be large in compression and integer math, while Intel's wins are large in prime numbers and floating point. The single thread and multithread results are comparatively small margins on both sides.

FAQ

Q: Which processor has the higher single thread score?

A: The Intel Core Ultra 5 225T scores 4348 in both passmark_single_thread and passmark_singlethread, while the AMD Ryzen AI Embedded P164 scores 4029 in both. Intel leads by 7.3%.

Q: How large is AMD's advantage in data compression?

A: AMD scores 327891 against Intel's 233998, a 40.1% lead. This is the second largest margin in the comparison after Intel's prime number win.

Q: What explains the multithreaded performance difference?

A: The multithread score is nearly identical: AMD at 25889 and Intel at 25358, a 2.1% AMD advantage. Despite AMD having 16 threads against Intel's 10, the overall parallel throughput is close.

Q: Does Intel win every floating point related test?

A: Yes. Intel wins floating point math with 82751 versus 55799, a 32.6% margin, and physics with 2053 versus 1210, a 41.1% margin. Both tests rely heavily on floating point operations.

Q: How do the processors compare in encryption?

A: Intel scores 18289 in data encryption against AMD's 16055, a 12.2% advantage. This is one of the smaller Intel wins but still consistent with the pattern of Intel leading in specialized instruction workloads.

Q: What is the overall average benchmark score difference?

A: AMD averages 52901 across its benchmark suite, placing in the 91st percentile. Intel averages 30468, placing in the 82nd percentile. The gap is driven primarily by AMD's large wins in compression and integer math.

Architecture Differences

The two processors come from different design philosophies. AMD's Ryzen AI Embedded P164 uses the Gorgon Point codename and belongs to the Ryzen AI Embedded generation built on Zen 5 and Zen 5c cores. It is manufactured on a 4 nm process at TSMC with a die size of 233 mm². Intel's Core Ultra 5 225T is part of the Core Ultra Series 2, uses the Arrow Lake architecture with the Arrow Lake-S codename, and is built on a 3 nm process, also at TSMC, with a die size of 243 mm² and a transistor count of 17,800 million.

Core counts differ significantly. AMD provides 8 cores and 16 threads, while Intel provides 10 cores and 10 threads. Intel's higher core count does not translate into a thread advantage because it lacks simultaneous multithreading. AMD's 16 threads give it a parallel scheduling advantage in multithreaded benchmarks, though the measured multithread gap is only 2.1%.

Cache hierarchies are structured differently. AMD uses 80 KB of L1 per core, 1 MB of L2 per core, and 8 MB of L3 cache. Intel uses 192 KB of L1 per core, 3 MB of L2 per core, and 20 MB of shared L3 cache. Intel's larger L3 cache, 20 MB versus 8 MB, provides more shared on-die storage for frequently accessed data. Intel also has a larger per-core L2 allocation, which may contribute to its single thread performance lead.

Memory support differs. AMD supports DDR5 and LPDDR5X across a dual-channel bus with 89.6 GB/s of bandwidth and ECC memory support. Intel supports DDR5 on a dual-channel bus with 102.4 GB/s of bandwidth but no ECC support. Intel's higher memory bandwidth, 102.4 versus 89.6, gives it a theoretical advantage in memory-bound workloads, though the benchmark data does not isolate memory bandwidth effects.

PCIe connectivity differs. AMD provides Gen 4 with 16 CPU-only lanes, while Intel provides Gen 5 with 20 CPU-only lanes. Intel's newer PCIe generation and higher lane count support faster peripheral connectivity. Integrated graphics also differ: AMD uses the Radeon 880M, while Intel uses Arc Xe-LPG Graphics 16EU.

The market segments and sockets diverge. AMD targets mobile with the AMD Socket FP8, while Intel targets desktop with the Intel Socket 1851. AMD's TDP is 28 watts, and Intel's is 65 watts. Release dates differ as well, with AMD dated 2026-03-08 and Intel dated 2024-12-31. Neither processor has an unlocked multiplier.

Specification Differences

Clock speeds differ in both base and boost. AMD has a base clock of 2.00 GHz and a boost clock of 5.00 GHz. Intel has a base clock of 2.50 GHz and a boost clock of 4.90 GHz. Intel's higher base clock, 2.50 versus 2.00, contributes to its single thread advantage, while AMD's higher boost clock, 5.00 versus 4.90, provides a peak frequency edge.

Thread counts differ substantially. AMD has 16 threads from 8 cores, while Intel has 10 threads from 10 cores. This is the primary structural difference in parallel processing capability. Despite Intel having more physical cores, the thread count favors AMD.

Cache specifications differ across all levels. AMD's L1 is 80 KB per core, L2 is 1 MB per core, and L3 is 8 MB. Intel's L1 is 192 KB per core, L2 is 3 MB per core, and L3 is 20 MB shared. Intel has larger cache allocations at every level.

Memory bandwidth differs, with Intel at 102.4 GB/s and AMD at 89.6 GB/s. ECC memory support is present on AMD but absent on Intel. Memory type support also differs: AMD supports both DDR5 and LPDDR5X, while Intel supports only DDR5.

PCIe generation and lane counts differ. AMD uses Gen 4 with 16 lanes, Intel uses Gen 5 with 20 lanes. The integrated graphics are different models, with AMD using Radeon 880M and Intel using Arc Xe-LPG Graphics 16EU.

Manufacturing process nodes differ, with AMD on 4 nm and Intel on 3 nm, both at TSMC. Die sizes are close, 233 mm² for AMD and 243 mm² for Intel. Intel has a recorded transistor count of 17,800 million, while AMD does not have a recorded transistor count in the database.

Market segments differ, with AMD classified as mobile and Intel as desktop. This aligns with their sockets: AMD Socket FP8 for AMD and Intel Socket 1851 for Intel. TDP differs, with AMD at 28 watts and Intel at 65 watts. Production status for both is active. Neither processor has an unlocked multiplier.