AMD Ryzen AI Embedded P164 vs AMD Ryzen Embedded 9900X3D Comparison

Head-to-Head Benchmarks

The recorded data does not contain overlapping benchmark results for these two processors. The AMD Ryzen AI Embedded P164 has a full set of PassMark workload scores, while the AMD Ryzen Embedded 9900X3D has no recorded benchmark entries in the database. Consequently, a direct comparison of individual test scores is not possible from the available measurements.

What can be assessed is the overall standing of each part. The Ryzen AI Embedded P164 holds a 91st percentile ranking among all CPUs, with an average benchmark score of 52,901. Its nearest rivals in the database include the AMD Ryzen 5 9500F (average score 52,873, delta of +0.1%), the Intel Xeon 634 (average score 52,974, delta of -0.1%), the AMD EPYC 7313P (average score 53,206, delta of -0.6%), and the AMD Ryzen 9 7900X (average score 53,288, delta of -0.7%). The data shows the P164 sits within roughly one percent of these established desktop and server parts, indicating its multi-threaded throughput is competitive with mainstream high-core-count processors.

The Ryzen Embedded 9900X3D lists no benchmark scores and no nearest rivals, resulting in an average benchmark score of zero and a 50th percentile placeholder. Without measured data, any numeric comparison between the two would be speculative. The database currently offers no head-to-head results for these items, so the analysis below focuses on architectural and specification differences that can be verified from the recorded fields.

Architecture Differences

The two processors come from different AMD embedded families, and the data reveals substantial architectural divergence. The Ryzen AI Embedded P164 uses the Gorgon Point codename and belongs to the Ryzen AI Embedded generation built on Zen 5 / Zen 5c cores. The Ryzen Embedded 9900X3D uses the Granite Ridge codename and belongs to the Ryzen Embedded (Zen 5, Granite Ridge) generation. Both are fabricated on TSMC's 4 nm process, but their physical designs differ sharply.

The P164 is a monolithic die measuring 233 mm², while the 9900X3D uses a chiplet layout with two dies, each measuring 70.6 mm². The 9900X3D also lists a transistor count of 16,630 million, a figure not provided for the P164. Core counts diverge as well: the P164 provides 8 cores and 16 threads, whereas the 9900X3D provides 12 cores and 24 threads. This gives the 9900X3D a 50% advantage in core and thread counts, which typically translates to higher multi-threaded throughput, assuming similar per-core performance.

Clock speeds differ considerably. The P164 has a base clock of 2.00 GHz and a boost clock of 5.00 GHz. The 9900X3D operates at a base clock of 4.40 GHz and a boost clock of 5.50 GHz. The 9900X3D's base clock is more than double that of the P164, indicating a much higher sustained frequency floor. Boost clocks are closer, with the 9900X3D leading by 0.50 GHz.

Cache hierarchy shows a major difference in L3 capacity. Both parts use 80 KB of L1 per core and 1 MB of L2 per core, but the P164 has 8 MB of L3 cache total, while the 9900X3D has 128 MB of L3 cache. The 128 MB figure reflects the 3D V-Cache technology implied by the X3D suffix, though the database does not explicitly label it as such. This 16x difference in L3 capacity is the most significant architectural gap, favoring the 9900X3D for workloads that benefit from large on-chip data residency.

Memory support differs: the P164 supports both DDR5 and LPDDR5X, while the 9900X3D supports DDR5 only. Both use dual-channel memory buses and share the same memory bandwidth of 89.6 GB/s. Both support ECC memory, which is expected for embedded and workstation-class parts.

PCIe connectivity also diverges. The P164 provides Gen 4 with 16 lanes (CPU only), while the 9900X3D provides Gen 5 with 24 lanes (CPU only). The 9900X3D offers both a newer PCIe generation and more lanes, which matters for expansion cards, NVMe storage, and GPU connectivity.

Integrated graphics differ in naming: the P164 carries a Radeon 880M, while the 9900X3D lists a generic Radeon Graphics. The database does not provide core counts or clock speeds for either iGPU, so performance cannot be quantified.

The socket and market segment also differ. The P164 uses AMD Socket FP8 and is classified as a mobile part, while the 9900X3D uses AMD Socket AM5 and is classified as a desktop part. The 9900X3D has an unlocked multiplier, whereas the P164 does not. The 9900X3D has a part number of 100-000001368E; the P164's part number is listed as unknown.

FAQ

Q: Which processor has more cores and threads?

A: The AMD Ryzen Embedded 9900X3D has 12 cores and 24 threads, compared to the AMD Ryzen AI Embedded P164's 8 cores and 16 threads.

Q: How do the base and boost clocks compare?

A: The 9900X3D has a base clock of 4.40 GHz and a boost clock of 5.50 GHz. The P164 has a base clock of 2.00 GHz and a boost clock of 5.00 GHz. The 9900X3D is higher in both metrics.

Q: What is the L3 cache difference?

A: The P164 has 8 MB of L3 cache, while the 9900X3D has 128 MB of L3 cache. This is a 16x difference, with the X3D part holding a substantial advantage for cache-sensitive workloads.

Q: Do both processors support ECC memory?

A: Yes, both the P164 and the 9900X3D support ECC memory. Both also support dual-channel memory, and both have a memory bandwidth of 89.6 GB/s.

Q: What PCIe versions and lane counts are available?

A: The P164 uses PCIe Gen 4 with 16 lanes (CPU only). The 9900X3D uses PCIe Gen 5 with 24 lanes (CPU only). The 9900X3D offers a newer standard and more lanes.

Q: Are there recorded benchmark scores for both processors?

A: No. The P164 has a full set of PassMark scores and an average benchmark score of 52,901. The 9900X3D has no recorded benchmark scores in the database, so its average benchmark score is zero.

Specification Differences

The table below shows only the fields where the two processors differ, based on the recorded data.

| Specification | AMD Ryzen AI Embedded P164 | AMD Ryzen Embedded 9900X3D |

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

| Series | null | 9000 series |

| Cores | 8 | 12 |

| Threads | 16 | 24 |

| Base Clock | 2.00 GHz | 4.40 GHz |

| Boost Clock | 5.00 GHz | 5.50 GHz |

| TDP | 28 W | 120 W |

| Socket | AMD Socket FP8 | AMD Socket AM5 |

| Codename | Gorgon Point | Granite Ridge |

| Generation | Ryzen AI Embedded (Zen 5 / Zen 5c) | Ryzen Embedded (Zen 5, Granite Ridge) |

| Transistors | null | 16,630 million |

| Die Size | 233 mm² | 2x 70.6 mm² |

| L3 Cache | 8 MB | 128 MB |

| Memory Support | DDR5, LPDDR5X | DDR5 |

| PCIe | Gen 4, 16 Lanes (CPU only) | Gen 5, 24 Lanes (CPU only) |

| Integrated Graphics | Radeon 880M | Radeon Graphics |

| Market Segment | Mobile | Desktop |

| Release Date | 2026-03-08 | 2025-10-06 |

| Multiplier Unlocked | false | true |

| Part Number | unknown | 100-000001368E |

| Benchmark Scores | Full PassMark set | None |

| Percentile vs All CPUs | 91 | 50 |

| Average Benchmark Score | 52,901 | 0 |

The Verdict

The data presents a clear split in roles. The AMD Ryzen AI Embedded P164 is a mobile-oriented, low-power part with a TDP of 28 W. It has 8 cores, a 2.00 GHz base clock, 8 MB of L3 cache, and PCIe Gen 4. Its benchmark performance, as measured by the average score of 52,901, places it at the 91st percentile, within 0.7% of the AMD Ryzen 9 7900X. This indicates strong multi-threaded capability for its power envelope.

The AMD Ryzen Embedded 9900X3D is a desktop-class part with a TDP of 120 W. It has 12 cores, a 4.40 GHz base clock, 128 MB of L3 cache, and PCIe Gen 5. It has no recorded benchmark scores, so its measured performance cannot be verified. However, its specifications point to a processor designed for maximum throughput and cache capacity, not for power efficiency.

For an embedded deployment where power draw is the primary constraint, the P164 is the logical choice. Its 28 W TDP is dramatically lower than the 9900X3D's 120 W, and its benchmark scores confirm it can deliver competitive performance. The 91st percentile ranking and an average score nearly identical to a Ryzen 9 7900X show that the P164 punches well above its power class.

For a workload that demands the largest possible L3 cache and the highest core count, the 9900X3D is the specification winner. The 128 MB L3 cache is an enormous advantage for data-intensive tasks, and the 12-core, 24-thread configuration provides more parallel processing headroom. The higher base clock, PCIe Gen 5 connectivity, and unlocked multiplier also favor this part for maximum performance.

The database currently offers no head-to-head scores, so a definitive performance verdict cannot be issued. The user should decide based on the architectural priorities: efficiency and verified benchmark results point to the P164, while raw specifications and cache capacity point to the 9900X3D.

Where Each One Wins

AMD Ryzen AI Embedded P164:

• Power efficiency: The 28 W TDP is far below the 9900X3D's 120 W, making it suitable for thermally constrained or battery-powered systems.
• Verified performance: The P164 has a complete set of PassMark scores, including a multi-thread score of 25,889, a single-thread score of 4,029, and an average benchmark score of 52,901. The 9900X3D has no measured scores.
• Memory flexibility: Support for both DDR5 and LPDDR5X gives the P164 broader memory options, which is useful in mobile and compact embedded designs.
• Market positioning: As a mobile part on Socket FP8, the P164 fits into systems where a smaller footprint and lower power draw are mandatory.

AMD Ryzen Embedded 9900X3D:

• Core and thread count: 12 cores and 24 threads, a 50% advantage over the P164's 8 cores and 16 threads.
• Cache capacity: 128 MB of L3 cache versus 8 MB, a 16x difference that strongly favors workloads with large working sets.
• Clock speeds: A 4.40 GHz base clock and 5.50 GHz boost clock exceed the P164's 2.00 GHz base and 5.00 GHz boost.
• Connectivity: PCIe Gen 5 with 24 lanes provides newer and more abundant I/O bandwidth compared to the P164's Gen 4 with 16 lanes.
• Overclocking: The unlocked multiplier on the 9900X3D allows manual tuning, which the locked P164 does not permit.
• Socket and ecosystem: AM5 is a desktop platform with a broad range of motherboards and cooling solutions, while FP8 is a mobile socket with fewer options.
• Production timeline: The 9900X3D has an earlier release date of 2025-10-06, meaning it has been available longer than the P164, which lists a release date of 2026-03-08.