AMD Ryzen 9 9900X3D vs AMD Ryzen AI Embedded P164 Comparison

FAQ

Q: What is the average benchmark score difference between the AMD Ryzen 9 9900X3D and the AMD Ryzen AI Embedded P164?

A: The Ryzen 9 9900X3D records an average benchmark score of 54762, while the Ryzen AI Embedded P164 scores 52901. That places the desktop chip roughly 3.5% ahead in aggregate performance, though both sit at the 91st percentile against all CPUs in the database.

Q: Which processor wins in single-threaded performance?

A: The Ryzen 9 9900X3D wins every recorded single-thread test. In PassMark single-thread, it scores 4646 versus 4029 for the embedded part, a 15.3% lead. The gap is narrower than in multi-threaded workloads, but it is still consistent across all measured tests.

Q: How large is the multi-threaded performance gap?

A: The Ryzen 9 9900X3D leads by 116.9% in PassMark multithread (56154 versus 25889). In the most extreme case, PassMark find prime numbers, the desktop chip scores 544 against 71, a 666.2% advantage. These are the largest deltas in the entire head-to-head comparison.

Q: Does the Ryzen AI Embedded P164 win any benchmark in the comparison?

A: No. Out of 11 recorded head-to-head benchmarks, the Ryzen 9 9900X3D wins all 11. The embedded chip's closest result is in single-thread tests, where it trails by 15.3%, but it never takes a single victory.

Q: What are the core and thread counts for each processor?

A: The Ryzen 9 9900X3D has 12 cores and 24 threads. The Ryzen AI Embedded P164 has 8 cores and 16 threads. That 50% core advantage and 50% thread advantage for the desktop chip explains much of the multi-threaded score gap.

Q: How do the two compare in power envelope and socket?

A: The Ryzen 9 9900X3D has a 120 W TDP and uses AMD Socket AM5, while the Ryzen AI Embedded P164 has a 28 W TDP and uses AMD Socket FP8. The embedded chip is designed for a much lower power ceiling, which aligns with its mobile market segment.

Architecture Differences

The architectural gap between these two AMD parts is substantial. The Ryzen 9 9900X3D belongs to the 9000 series and is built on the Zen 5 architecture with the Granite Ridge codename. The Ryzen AI Embedded P164 uses the Gorgon Point codename and combines Zen 5 with Zen 5c cores, as indicated by its generation label. Both are manufactured on a 4 nm process at TSMC, but the physical designs diverge sharply.

The Ryzen 9 9900X3D is a desktop-focused processor with 12 cores and 24 threads, all running Zen 5. The embedded chip halves that to 8 cores and 16 threads, and the mixed Zen 5 / Zen 5c design suggests a blend of performance and efficiency cores. The desktop chip's die is composed of two 70.6 mm² chiplets, totaling 16,630 million transistors. The embedded chip is a monolithic 233 mm² die, and the database does not list a transistor count for it.

Cache allocations differ even more dramatically. Both chips share the same per-core L1 and L2 sizes: 80 KB and 1 MB per core respectively. However, the L3 cache tells a very different story. The Ryzen 9 9900X3D carries 128 MB of L3 cache, a hallmark of the 3D V-Cache approach, while the Ryzen AI Embedded P164 has only 8 MB. That is a 16-fold difference in last-level cache, which directly impacts workloads that reuse large datasets.

Memory support also separates the pair. The Ryzen 9 9900X3D supports DDR5 only, with a dual-channel bus and 89.6 GB/s of memory bandwidth. The embedded chip supports both DDR5 and LPDDR5X, also dual-channel and also rated at 89.6 GB/s. Both support ECC memory. PCIe connectivity differs as well: the desktop chip offers Gen 5 with 24 CPU lanes, while the embedded chip provides Gen 4 with 16 lanes.

Integrated graphics are not equal. The Ryzen 9 9900X3D includes Radeon Graphics, while the Ryzen AI Embedded P164 features Radeon 880M. The multiplier is unlocked on the desktop part, allowing overclocking, whereas the embedded part is locked. The desktop chip uses part number 100-000001368, while the embedded chip's part number is listed as unknown.

Where Each One Wins

The recorded benchmark data shows no scenario where the Ryzen AI Embedded P164 wins. Every single one of the 11 head-to-head tests favors the Ryzen 9 9900X3D. That does not mean the embedded chip is without purpose, but its strengths lie outside the measured performance metrics. The 28 W TDP and mobile socket position it for low-power, embedded applications where thermal and power budgets are constrained, not for peak performance.

The Ryzen 9 9900X3D wins across every category in the PassMark suite. In integer math, it scores 179727 against 87940, a 104.4% lead. In floating-point math, the gap is 114.4% (119622 versus 55799). Data compression shows a 108.9% advantage (685074 versus 327891), and data encryption shows 107.3% (33282 versus 16055). Extended instruction workloads give the desktop chip a 129.1% edge (55426 versus 24193). Random string sorting is 106.5% ahead (71864 versus 34801). Physics simulation is particularly lopsided: 5340 versus 1210, a 341.3% difference.

The single-thread results are the closest. In PassMark single-thread, the Ryzen 9 9900X3D scores 4646 versus 4029, a 15.3% edge. That is the smallest delta in the entire comparison, but it is still a clear victory. The Ryzen AI Embedded P164's best showing is therefore in lightly threaded, short-duration tasks, where its 5.00 GHz boost clock partially compensates for the lower core count and smaller cache.

For the embedded part, the practical win is power efficiency. Its 28 W TDP versus 120 W for the desktop chip means it can operate in fanless or thermally constrained systems. That is a qualitative advantage not captured in the benchmark scores. The database does not record any wins for it, but its existence is justified by the mobile and embedded market segment, not by raw throughput.

Specification Differences

The two processors differ across nearly every major specification category. Core count is 12 versus 8, and thread count is 24 versus 16. Base clocks diverge sharply: 4.40 GHz for the Ryzen 9 9900X3D versus 2.00 GHz for the embedded chip. Boost clocks are closer, with 5.50 GHz versus 5.00 GHz. TDP is 120 W versus 28 W, a 4.3x difference in thermal design power.

Sockets are incompatible: AMD Socket AM5 for the desktop chip, AMD Socket FP8 for the embedded chip. Architecture is Zen 5 for the former, while the latter uses a Zen 5 / Zen 5c combination. The desktop chip's codename is Granite Ridge; the embedded chip's codename is Gorgon Point. The desktop part belongs to the Ryzen 9 generation, while the embedded part is labeled Ryzen AI Embedded.

Process node is identical at 4 nm TSMC, but die size and transistor counts differ. The desktop chip uses two 70.6 mm² dies, while the embedded chip is a single 233 mm² die. Transistors are listed at 16,630 million for the desktop chip, with no figure for the embedded part.

Cache differences are the most extreme. L1 and L2 are identical per core, but L3 is 128 MB versus 8 MB. Memory support adds LPDDR5X to the embedded chip, while the desktop chip supports DDR5 only. Memory bandwidth is the same at 89.6 GB/s, and both are dual-channel. PCIe generation differs: Gen 5 with 24 lanes versus Gen 4 with 16 lanes.

Integrated graphics change from Radeon Graphics to Radeon 880M. The multiplier is unlocked only on the desktop chip. Release dates differ: the Ryzen 9 9900X3D launched on 2025-01-05, while the embedded part is dated 2026-03-08. The database lists a launch MSRP of $599 for the desktop chip, while no MSRP is recorded for the embedded part. Market segments are Desktop versus Mobile.

Head-to-Head Benchmarks

The head-to-head data is unambiguous: the Ryzen 9 9900X3D wins all 11 recorded benchmarks. The largest margin comes in PassMark find prime numbers, where the desktop chip scores 544 against 71. That is a 666.2% delta, the single most dominant result in the comparison. Prime number finding is heavily dependent on integer throughput and memory latency, both of which favor the desktop chip's larger cache and higher clocks.

PassMark physics shows a 341.3% advantage (5340 versus 1210). Physics simulation often scales with core count and cache size, and the 12-core/128 MB combination overwhelms the 8-core/8 MB embedded part. Extended instructions yield a 129.1% lead (55426 versus 24193), reflecting the desktop chip's superior instruction throughput. Multithread is 116.9% ahead (56154 versus 25889), and floating-point math is 114.4% ahead (119622 versus 55799).

Data compression and encryption results are nearly identical in margin: 108.9% and 107.3% respectively. Compression scores 685074 versus 327891, while encryption scores 33282 versus 16055. Random string sorting is 106.5% ahead (71864 versus 34801), and integer math is 104.4% ahead (179727 versus 87940). These workloads all benefit from the desktop chip's higher core count and much larger L3 cache.

The single-thread tests are the only close calls. PassMark single-thread and singlethread both record 4646 for the Ryzen 9 9900X3D and 4029 for the embedded chip, a 15.3% delta. Even here, the desktop chip wins, but the margin is modest compared to the 100%+ gaps seen in multi-threaded and cache-sensitive tests. The embedded chip's 5.00 GHz boost clock helps it stay within striking distance in lightly threaded scenarios.

The Verdict

The data is one-sided. The AMD Ryzen 9 9900X3D wins every benchmark in the head-to-head comparison, with margins ranging from 15.3% in single-thread to 666.2% in prime number finding. Its 12 cores, 24 threads, 128 MB L3 cache, and 5.50 GHz boost clock create a performance profile that the 8-core, 16-thread, 8 MB L3 embedded chip cannot match. Anyone selecting purely on throughput should choose the Ryzen 9 9900X3D.

The Ryzen AI Embedded P164 is not a performance competitor. It is a 28 W mobile part with a locked multiplier and a 2.00 GHz base clock. Its 5.00 GHz boost clock and 8 MB L3 cache are sufficient for embedded workloads that prioritize power efficiency and compact thermal solutions. The database shows no benchmark win for it, but its low TDP and mobile socket make it suitable for systems where the 120 W desktop chip cannot operate.

For desktop users, content creators, or anyone running multi-threaded workloads, the Ryzen 9 9900X3D is the clear choice. Its average benchmark score of 54762 sits at the 91st percentile, and it compares favorably to nearest rivals like the Intel Core Ultra 5 245KF (within 0.6%) and the Intel Core i7-14700F (2.1% ahead). The embedded chip's average score of 52901 also lands at the 91st percentile, but its nearest rivals include the AMD Ryzen 5 9500F (0.1% ahead) and the AMD Ryzen 9 7900X (0.7% ahead), showing it trades blows with lower-tier desktop parts.

The verdict depends on the use case. For raw performance, the Ryzen 9 9900X3D is the only choice. For embedded, low-power systems where 28 W is a hard limit, the Ryzen AI Embedded P164 serves a different purpose, but the recorded data cannot justify it on performance grounds. The 11-0 win count for the desktop chip is the final word.