AMD Ryzen AI Embedded P164 vs AMD Ryzen AI Max+ 388 Comparison
AMD Ryzen AI Embedded P164
Ryzen AI Max+ 388
PERFORMANCE BENCHMARKS
Analysis: AMD Ryzen AI Embedded P164 vs AMD Ryzen AI Max+ 388
Where Each One Wins
The recorded benchmark data presents a clear, one-sided picture: the AMD Ryzen AI Max+ 388 wins every single head-to-head comparison in the database. Out of 11 shared tests, the Max+ 388 takes all 11, with the P164 recording zero wins. This does not mean the P164 is without merit; rather, its strengths lie in characteristics not captured by these particular workloads, such as its lower power envelope and different platform integration.
The P164, with a 28 watt TDP and an 89.6 GB/s memory bandwidth, is positioned for efficiency-oriented mobile designs. Its smaller cache hierarchy and dual-channel memory bus suggest it targets scenarios where sustained throughput is secondary to power discipline. The Max+ 388, by contrast, operates at 55 watts, uses a quad-channel memory bus with 256.0 GB/s bandwidth, and carries a larger 32 MB shared L3 cache. These specifications align with workloads that are memory-bandwidth sensitive or cache-hungry.
In the benchmark suite, the Max+ 388 demonstrates its dominance most decisively in compute-heavy tasks. The largest margin appears in prime number finding, where the Max+ 388 scores 145 versus the P164's 71, a 51% advantage. Physics simulation also shows a wide gap: 1843 versus 1210, a 34.3% lead. These are workloads that scale with raw compute throughput and memory latency, both areas where the Max+ 388's higher base clock (3.60 GHz versus 2.00 GHz) and larger cache provide structural advantages.
The P164's best relative showing is in single-thread performance. There, the Max+ 388 leads by only 3.7% (4185 versus 4029). This narrow margin indicates that for lightly threaded tasks, the P164's Zen 5 architecture holds its own, and the efficiency benefit of the lower TDP becomes more relevant. In data compression, the Max+ 388 leads by 18.2% (400887 versus 327891), while in floating-point math the gap is 23.3% (72722 versus 55799). The P164's role, based on this data, is not to win raw performance contests but to deliver adequate compute within a power-constrained design.
Architecture Differences
Both processors share the same manufacturer, AMD, and are built on the same 4 nanometer TSMC process node. Both use Zen 5 cores, but the P164 is part of the Gorgon Point generation with a hybrid Zen 5 / Zen 5c configuration, while the Max+ 388 is a Strix Halo part using standard Zen 5 cores. Both have 8 cores and 16 threads, and each core has 80 KB of L1 cache and 1 MB of L2 cache. The L3 cache, however, differs significantly: the P164 has 8 MB, while the Max+ 388 has 32 MB shared.
The sockets differ. The P164 uses AMD Socket FP8, while the Max+ 388 uses AMD Socket FP11. This is not a drop-in upgrade path; the two are separate platforms. The P164's die size is 233 mm², while the Max+ 388 uses a dual-die design with each die at 70.6 mm², totaling roughly 141.2 mm² of silicon. Both support Gen 4 PCIe with 16 CPU lanes, and both support ECC memory. Memory support diverges: the P164 supports DDR5 and LPDDR5X over a dual-channel bus, while the Max+ 388 supports only LPDDR5X over a quad-channel bus, resulting in 256.0 GB/s versus 89.6 GB/s of bandwidth.
The integrated graphics differ as well. The P164 carries a Radeon 880M, while the Max+ 388 has a Radeon 8060S. The latter is positioned as a higher-tier iGPU, consistent with the Max+ 388's larger memory bandwidth and higher TDP. The base clocks are a major distinction: the P164 runs at 2.00 GHz, while the Max+ 388 runs at 3.60 GHz. Both boost to 5.00 GHz. The P164's release date is later (2026-03-08) than the Max+ 388's (2026-01-05), but both are currently active in production.
The P164 sits in the 91st percentile of all CPUs, while the Max+ 388 sits in the 90th. Their average benchmark scores are close: 52901 for the P164 and 49796 for the Max+ 388. This apparent contradiction, where the P164 has a higher average yet loses every head-to-head test, stems from the P164's benchmark list including only PassMark tests, while the Max+ 388 also includes Cinebench R15 and R23 results, which may drag its average down relative to the P164's narrower suite.
Head-to-Head Benchmarks
The database records 11 shared PassMark tests, and the Max+ 388 wins all of them. The largest win is in prime number finding, where the Max+ 388 scores 145 against the P164's 71, a 51% advantage. This is an extreme outlier, suggesting the Max+ 388's higher base clock and larger cache disproportionately benefit this algorithm.
The second-largest margin is in physics simulation, where the Max+ 388 scores 1843 versus 1210, a 34.3% lead. Extended instructions show a 26.1% gap (32719 versus 24193). Floating-point math follows at 23.3% (72722 versus 55799). The Max+ 388 also leads by 22.7% in multithread performance (33486 versus 25889), by 20.1% in data encryption (20092 versus 16055), and by 19.8% in integer math (109588 versus 87940).
Random string sorting shows a 19.4% advantage (43196 versus 34801), and data compression shows an 18.2% lead (400887 versus 327891). The closest contest is single-thread performance, where the Max+ 388 leads by only 3.7% (4185 versus 4029). This single-thread result is the most informative for understanding the P164's character: its Zen 5 cores are nearly as fast per thread as the Max+ 388's, but the Max+ 388's higher base clock gives it a slight edge.
The Cinebench results, available only for the Max+ 388, provide additional context for its performance envelope. In Cinebench R23, it scores 18759 multicore and 1960 singlecore. In R15, it scores 2872 multicore and 298 singlecore. These are not comparative against the P164, but they confirm the Max+ 388's strong multithreaded capability relative to its own single-thread scores.
FAQ
Q: Does the AMD Ryzen AI Embedded P164 ever win any benchmark in the database?
A: No. In the 11 shared PassMark tests, the Max+ 388 wins all 11. The P164 records zero wins.
Q: What is the closest benchmark result between the two?
A: Single-thread performance. The Max+ 388 scores 4185, while the P164 scores 4029, a 3.7% gap. This is the only test where the P164 comes within single-digit percentage points.
Q: How much larger is the Max+ 388's L3 cache compared to the P164?
A: The Max+ 388 has 32 MB shared L3 cache, while the P164 has 8 MB. This is a 4x difference.
Q: What are the memory bandwidth specifications for each processor?
A: The P164 uses a dual-channel bus with 89.6 GB/s, while the Max+ 388 uses a quad-channel bus with 256.0 GB/s.
Q: Do both processors have the same number of cores and threads?
A: Yes, both have 8 cores and 16 threads. Both also have 80 KB L1 per core and 1 MB L2 per core.
Q: Which processor has the higher base clock?
A: The Max+ 388 has a 3.60 GHz base clock, while the P164 has a 2.00 GHz base clock. Both boost to 5.00 GHz.
Specification Differences
The following fields differ between the AMD Ryzen AI Embedded P164 and the AMD Ryzen AI Max+ 388:
- Codename: Gorgon Point versus Strix Halo
- Generation: Ryzen AI Embedded (Zen 5 / Zen 5c) versus Ryzen AI Max (Zen 5, Strix Halo)
- Socket: AMD Socket FP8 versus AMD Socket FP11
- Base Clock: 2.00 GHz versus 3.60 GHz
- TDP: 28 watts versus 55 watts
- Die Size: 233 mm² versus 2x 70.6 mm²
- L3 Cache: 8 MB versus 32 MB (shared)
- Memory Support: DDR5, LPDDR5X versus LPDDR5X only
- Memory Bus: Dual-channel versus Quad-channel
- Memory Bandwidth: 89.6 GB/s versus 256.0 GB/s
- Integrated Graphics: Radeon 880M versus Radeon 8060S
- Release Date: 2026-03-08 versus 2026-01-05
- Part Number: unknown versus 100-000001980
- Percentile vs All CPUs: 91 versus 90
- Average Benchmark Score: 52901 versus 49796
- Nearest Rivals: The P164's closest rivals include the AMD Ryzen 5 9500F (0.1% delta), Intel Xeon 634 (-0.1%), AMD EPYC 7313P (-0.6%), and AMD Ryzen 9 7900X (-0.7%). The Max+ 388's closest rivals include the Intel Core 9 273PE (-0.1%), Intel Core i5-14600KF (0.8%), AMD Ryzen 9 7900 (1.2%), and AMD Ryzen 7 PRO 5755G (1.2%).
The P164 and Max+ 388 share the same process node (4 nm, TSMC), the same core count (8), thread count (16), L1 and L2 cache per core (80 KB and 1 MB), boost clock (5.00 GHz), PCIe configuration (Gen 4, 16 lanes), ECC support, and market segment (Mobile). Both are active in production and have locked multipliers.