AMD Ryzen 5 40 vs AMD Ryzen AI Embedded P164 Comparison

Where Each One Wins

The benchmark data presents a completely one-sided comparison. Out of 11 recorded head-to-head tests, the AMD Ryzen AI Embedded P164 wins every single one. The AMD Ryzen 5 40 records zero wins across the entire test suite. This is not a close contest with trade-offs between workloads; the P164 dominates every category measured, from single-threaded responsiveness to multi-threaded throughput.

Looking at the PassMark sub-tests, the P164 shows its largest advantages in compute-heavy workloads. In extended instructions, it scores 24193 against 6437 for the Ryzen 5 40, a delta of 73.4%. Floating-point math shows a 72.8% gap, with the P164 at 55799 versus 15194. Integer math follows closely at 64.1% ahead, scoring 87940 against 31598. Prime number finding, a test that stresses branch prediction and integer iteration, gives the P164 a 71.8% edge with 71 points versus 20.

Memory and data tasks also favor the P164 decisively. Data compression scores 327891 against 141533, a 56.8% delta. Random string sorting shows a 56.5% gap, with 34801 versus 15124. Data encryption reaches 16055 against 6646, a 58.6% difference. The multi-threaded PassMark score of 25889 versus 9341 represents a 63.9% advantage, while the physics sub-test shows 1210 versus 432, a 64.3% gap.

Even the closest contest, single-thread performance, still favors the P164 by 38.5%. It scores 4029 in PassMark single-thread against 2477 for the Ryzen 5 40. The overall average benchmark score tells the same story: 52901 for the P164 against 15882 for the Ryzen 5 40. The Ryzen 5 40 sits at the 70th percentile of all CPUs in the database, while the P164 reaches the 91st percentile.

The Ryzen 5 40 does have some favorable characteristics in the raw specifications. It runs at a lower 15 W TDP against 28 W for the P164, and its base clock of 2.80 GHz is higher than the P164's 2.00 GHz. But in actual recorded performance, none of these translate into a single benchmark victory.

Architecture Differences

The two processors come from different generations and use different manufacturing processes. The Ryzen 5 40 uses the Zen 2 architecture under the Mendocino codename, built on a 6 nm process at TSMC. The Ryzen AI Embedded P164 uses the newer Gorgon Point codename with a Zen 5 / Zen 5c hybrid configuration, built on a 4 nm process, also at TSMC. The die size reflects this generational gap: the Ryzen 5 40 measures 100 mm² while the P164 is substantially larger at 233 mm².

Core and thread counts differ significantly. The Ryzen 5 40 provides 4 cores and 8 threads. The P164 doubles that to 8 cores and 16 threads. Cache hierarchies also scale up. The Ryzen 5 40 has 64 KB of L1 per core, 512 KB of L2 per core, and 4 MB of shared L3 cache. The P164 offers 80 KB of L1 per core, 1 MB of L2 per core, and 8 MB of L3 cache. Every level of cache favors the P164, both in per-core capacity and total shared capacity.

Clock behavior shows a trade-off. The Ryzen 5 40 has a higher base clock at 2.80 GHz but a lower boost clock at 4.30 GHz. The P164 starts lower at 2.00 GHz base but boosts higher to 5.00 GHz. The higher boost ceiling helps explain the P164's 38.5% single-thread advantage despite its lower base clock.

Memory support diverges as well. The Ryzen 5 40 supports LPDDR5 only, while the P164 supports both DDR5 and LPDDR5X. Both use dual-channel memory buses. Memory bandwidth is close: 88.0 GB/s for the Ryzen 5 40 and 89.6 GB/s for the P164. ECC memory is not supported on the Ryzen 5 40 but is available on the P164, which matters for embedded and reliability-focused deployments.

PCIe connectivity shows a major difference. The Ryzen 5 40 uses PCIe Gen 3 with 4 lanes from the CPU. The P164 uses PCIe Gen 4 with 16 lanes. This quadruples the available CPU-attached lane count and doubles the per-lane bandwidth generation, which affects expandability for embedded platforms.

Integrated graphics differ as well. The Ryzen 5 40 pairs with a Radeon 610M, while the P164 uses a Radeon 880M. Both target mobile and embedded segments, but the P164's newer and larger iGPU aligns with its higher overall positioning.

Sockets are not interchangeable. The Ryzen 5 40 uses AMD Socket FT6, while the P164 uses AMD Socket FP8. Neither chip has an unlocked multiplier. The Ryzen 5 40 was released in 2025, while the P164 carries a 2026 release date. Both remain in active production.

Head-to-Head Benchmarks

The recorded data shows the P164 ahead in every measured test, but the size of the gap varies meaningfully by workload type.

The largest margin comes in extended instructions, where the P164 leads by 73.4%. This test typically measures vectorized and SIMD-heavy code paths. The P164's score of 24193 against 6437 suggests a significantly wider execution pipeline or better instruction-level parallelism per cycle. Floating-point math follows at 72.8% ahead, with the P164 scoring 55799 versus 15194. This aligns with the Zen 5 architecture's wider floating-point units compared to Zen 2.

Prime number finding shows a 71.8% gap, with the P164 at 71 versus 20. This test is sensitive to branch prediction and integer division throughput. The P164's 38.5% single-thread lead contributes here, but the 71.8% margin exceeds the single-thread gap, indicating that multi-threaded scaling amplifies the advantage in this workload.

Integer math shows a 64.1% delta. The P164 scores 87940 against 31598. Multi-threaded performance sits at 63.9% ahead, scoring 25889 versus 9341. Physics simulation, another heavily multi-threaded workload, shows 1210 versus 432, a 64.3% gap. These three tests cluster tightly between 63.9% and 64.3%, suggesting that the P164's advantage in general multi-threaded integer work is consistent.

Data compression and random string sorting show the smallest multi-threaded deltas at 56.8% and 56.5% respectively. The P164 scores 327891 and 34801 against 141533 and 15124. These workloads often depend on memory bandwidth and cache behavior. The P164's larger 8 MB L3 cache and newer memory controller appear to help, but the relative gap is narrower than in pure compute tests.

Data encryption shows a 58.6% delta, with the P164 at 16055 versus 6646. Encryption workloads frequently rely on specific instruction extensions, which ties back to the large extended-instruction advantage.

Single-thread performance is the closest contest at 38.5%. The P164 scores 4029 against 2477. This gap is substantial but smaller than the multi-threaded margins. It reflects the combination of the P164's 5.00 GHz boost clock against 4.30 GHz, plus the architectural improvements from Zen 2 to Zen 5. The duplicate single-thread entries in the database, passmark_single_thread and passmark_singlethread, both record identical results, confirming the 38.5% figure.

The average benchmark scores place the two chips in different competitive tiers. The Ryzen 5 40 averages 15882, sitting within 0.4% of the AMD EPYC 75F3 and 0.4% of the AMD EPYC 9354P. It trails the Intel Core Ultra 5 134U by 0.2% and the AMD EPYC 9334 by 0.4%. The P164 averages 52901, coming within 0.1% of the AMD Ryzen 5 9500F and trailing the Intel Xeon 634 by 0.1%, the AMD EPYC 7313P by 0.6%, and the AMD Ryzen 9 7900X by 0.7%. The P164's closest rivals are desktop and server parts, while the Ryzen 5 40 competes with a mix of mobile and server chips.

The Verdict

The data points to a clear performance hierarchy. The AMD Ryzen AI Embedded P164 is the stronger processor in every benchmark category recorded. Its 91st percentile ranking versus the Ryzen 5 40's 70th percentile places them in different performance classes entirely. The 52901 average benchmark score against 15882 represents a roughly 3.3x overall difference, driven by the P164's doubling of cores and threads, larger caches, higher boost clock, and newer architecture.

The Ryzen 5 40 retains relevance for scenarios where its 15 W TDP and smaller 100 mm² die matter more than raw throughput. It draws 13 W less than the P164 and uses a lower-power socket. For thermally constrained mobile designs, that difference can be decisive. The Ryzen 5 40 also supports LPDDR5 memory, which suits compact, low-power systems.

The P164 targets embedded platforms that need compute headroom. Its 8 cores and 16 threads, 8 MB L3 cache, PCIe Gen 4 with 16 lanes, and ECC memory support make it suitable for workloads that require sustained multi-threaded performance plus reliability features. The 28 W TDP is higher, but the performance returns are proportionally larger across every measured metric.

Users prioritizing single-thread responsiveness should pick the P164, as it leads by 38.5%. Users prioritizing multi-threaded throughput should also pick the P164, as it leads by 63.9% in PassMark multi-thread. Users prioritizing power efficiency might consider the Ryzen 5 40, but the database contains no efficiency benchmarks to quantify that trade-off.

The release dates matter for platform planning. The Ryzen 5 40 released in 2025, while the P164 follows in 2026. Both remain active in production. The P164's newer release aligns with its newer architecture and process node.

FAQ

Q: Which processor has more cores and threads?

A: The AMD Ryzen AI Embedded P164 has 8 cores and 16 threads. The AMD Ryzen 5 40 has 4 cores and 8 threads.

Q: How large is the single-thread performance gap?

A: The P164 leads by 38.5% in PassMark single-thread tests, scoring 4029 against 2477 for the Ryzen 5 40.

Q: What is the memory bandwidth difference?

A: The Ryzen 5 40 provides 88.0 GB/s, while the P164 provides 89.6 GB/s. Both use dual-channel memory buses.

Q: Does either processor support ECC memory?

A: The P164 supports ECC memory. The Ryzen 5 40 does not.

Q: What PCIe connectivity does each chip offer?

A: The Ryzen 5 40 uses PCIe Gen 3 with 4 CPU lanes. The P164 uses PCIe Gen 4 with 16 CPU lanes.

Q: Which chip has the higher boost clock?

A: The P164 boosts to 5.00 GHz, while the Ryzen 5 40 boosts to 4.30 GHz. The Ryzen 5 40 has the higher base clock at 2.80 GHz versus 2.00 GHz for the P164.

Q: How do the cache sizes compare?

A: The Ryzen 5 40 has 64 KB L1 per core, 512 KB L2 per core, and 4 MB shared L3. The P164 has 80 KB L1 per core, 1 MB L2 per core, and 8 MB L3.