AMD Ryzen AI 9 365 vs AMD Ryzen AI Max PRO 385 Comparison

FAQ

Q: Which processor has more cores and threads?

A: The AMD Ryzen AI 9 365 has 10 cores and 20 threads, while the AMD Ryzen AI Max PRO 385 has 8 cores and 16 threads.

Q: How do the two chips compare in single-core Cinebench R23 performance?

A: The AMD Ryzen AI Max PRO 385 scores 4012 in Cinebench R23 single-core, which is 50.3% ahead of the Ryzen AI 9 365's 1992.

Q: What is the memory configuration difference?

A: The Ryzen AI 9 365 uses a dual-channel bus with 89.6 GB/s bandwidth, while the Ryzen AI Max PRO 385 uses a quad-channel bus with 256.0 GB/s bandwidth. The Max PRO also supports ECC memory, which the 365 does not.

Q: Which processor has a higher TDP?

A: The AMD Ryzen AI Max PRO 385 has a TDP of 55, compared to 28 for the AMD Ryzen AI 9 365.

Q: Do both processors use the same architecture?

A: Yes, both use Zen 5 architecture on a 4 nm TSMC process, but the Ryzen AI 9 365 is codenamed Strix Point while the Max PRO 385 is codenamed Strix Halo.

Q: What is the average benchmark score difference?

A: The Ryzen AI Max PRO 385 has an average benchmark score of 43326, which is 8.2% higher than the Ryzen AI 9 365's 40048.

Architecture Differences

Both processors are built on AMD's Zen 5 architecture and manufactured on TSMC's 4 nm process, but they diverge significantly in their overall design. The Ryzen AI 9 365, codenamed Strix Point, belongs to the Ryzen AI 300 generation that combines Zen 5 with Zen 5c cores. It packs 10 cores and 20 threads, with a base clock of 2.00 GHz and a boost clock of 5.00 GHz. The die size is recorded at 233 mm².

The Ryzen AI Max PRO 385, codenamed Strix Halo, sits in the Ryzen AI Max PRO generation and uses pure Zen 5 cores without the efficiency-core variant. It has 8 cores and 16 threads, but runs a much higher base clock of 3.60 GHz while matching the 5.00 GHz boost clock of the 365. The Max PRO's TDP of 55 is nearly double the 365's 28, reflecting a more power-hungry design aimed at sustained performance.

Cache configurations differ notably. Both processors share 80 KB of L1 cache per core and 1 MB of L2 cache per core, but the L3 cache is a major differentiator. The Ryzen AI 9 365 has 16 MB of L3 cache, while the Ryzen AI Max PRO 385 has 32 MB of shared L3 cache, doubling the available last-level cache.

Memory subsystems are also distinct. The Ryzen AI 9 365 supports both DDR5 and LPDDR5X memory over a dual-channel bus, delivering 89.6 GB/s of bandwidth. The Ryzen AI Max PRO 385 supports only LPDDR5X but uses a quad-channel bus, more than doubling bandwidth to 256.0 GB/s. The Max PRO also adds ECC memory support, a feature absent from the 365.

Both processors use AMD Socket FP8 and FP11 respectively, and both run PCIe Gen 4 with 16 lanes from the CPU. Integrated graphics differ as well: the Ryzen AI 9 365 uses the Radeon 880M, while the Ryzen AI Max PRO 385 uses the Radeon 8050S. The Max PRO was released on 2025-01-05, roughly six months after the 365's 2024-06-30 launch. Neither processor has an unlocked multiplier, and both are active in production.

The Verdict

The benchmark data presents a clear hierarchy. The AMD Ryzen AI Max PRO 385 wins all 15 recorded head-to-head comparisons against the Ryzen AI 9 365. Its average benchmark score of 43326 places it in the 88th percentile of all CPUs, while the 365's 40048 sits in the 87th percentile. The Max PRO's nearest rival is the AMD Ryzen AI 9 465, which scores 43431, a difference of just 0.2%. The 365's closest competitor is the AMD Ryzen 7 7700 at 40081, only 0.1% away.

For workloads that stress single-core performance, the Max PRO 385 is the definitive choice. Its Cinebench R23 single-core score of 4012 is more than double the 365's 1992, a 50.3% advantage. The gap persists in Cinebench R15 single-core, where the Max PRO scores 404 against 303, a 25% lead. Even in multi-threaded tests where the 365 has two additional cores and four additional threads, the Max PRO still wins by substantial margins. In Cinebench R23 multi-core, the Max PRO delivers 28424 versus 18698, a 34.2% advantage. The Cinebench R15 multi-core result is closer, with the Max PRO at 2865 versus 2842, only 0.8% ahead.

The Max PRO's advantage extends to every PassMark subtest in the database. Data compression shows 379448 versus 354510, a 6.6% lead. Data encryption shows 18978 versus 18297, a 3.6% lead. Extended instructions score 31442 versus 25113, a 20.1% lead. Floating point math results in 69580 versus 62802, a 9.7% lead. Integer math delivers 105056 versus 101831, a 3.1% lead. The multithread PassMark score is 32075 versus 29467, an 8.1% lead. Physics shows a narrow 1711 versus 1704, just 0.4% ahead. Random string sorting results in 40784 versus 39447, a 3.3% lead. Single-thread PassMark shows 3995 versus 3841, a 3.9% lead. Prime number finding shows 157 versus 117, a 25.5% lead.

The Ryzen AI 9 365 does retain advantages in core count, thread count, and a lower TDP of 28, which may suit power-constrained designs. It also supports DDR5 memory in addition to LPDDR5X, offering more flexible memory options. However, the recorded performance data gives the Max PRO 385 the win in every measured category. Users seeking maximum compute performance from the database should select the Max PRO 385. The 365 remains viable for systems where the lower TDP and dual-channel memory support take priority over raw benchmark scores.

Specification Differences

The two processors differ across several key specifications. The Ryzen AI 9 365 has 10 cores and 20 threads, while the Ryzen AI Max PRO 385 has 8 cores and 16 threads. Base clocks differ substantially: the 365 runs at 2.00 GHz, the Max PRO at 3.60 GHz. Both boost to 5.00 GHz.

TDP ratings diverge sharply, with the 365 at 28 and the Max PRO at 55. The 365 uses AMD Socket FP8, while the Max PRO uses AMD Socket FP11. Cache totals differ, with the 365 offering 16 MB of L3 cache and the Max PRO offering 32 MB of shared L3 cache. Memory support varies: the 365 accepts DDR5 and LPDDR5X over a dual-channel bus with 89.6 GB/s bandwidth, while the Max PRO accepts only LPDDR5X over a quad-channel bus with 256.0 GB/s bandwidth. ECC memory is supported only on the Max PRO.

The integrated graphics differ, with the 365 featuring the Radeon 880M and the Max PRO featuring the Radeon 8050S. Release dates differ as well: the 365 launched on 2024-06-30, the Max PRO on 2025-01-05. The 365 has a die size of 233 mm², while no die size is recorded for the Max PRO. Part numbers also differ, with the 365 listed as 100-000001530 and the Max PRO as 100-000001422.

Head-to-Head Benchmarks

The largest single-core gap appears in Cinebench R23 single-core, where the Max PRO 385 scores 4012 against the 365's 1992, a 50.3% advantage. This is the most decisive result in the entire comparison. Cinebench R15 single-core shows a 25% gap, with the Max PRO at 404 and the 365 at 303. PassMark prime number finding shows a 25.5% lead for the Max PRO, scoring 157 versus 117.

Multi-threaded results also favor the Max PRO heavily. Cinebench R23 multi-core shows the Max PRO at 28424 versus 18698, a 34.2% lead. Extended instructions in PassMark show a 20.1% advantage, with scores of 31442 versus 25113. Floating point math shows a 9.7% lead, 69580 versus 62802. PassMark multithread shows an 8.1% lead, 32075 versus 29467.

Some results are much closer. Cinebench R15 multi-core shows the Max PRO at 2865 versus 2842, just 0.8% ahead. PassMark physics shows 1711 versus 1704, a 0.4% difference, the narrowest margin in the set. Data encryption shows 18978 versus 18297, a 3.6% lead. Integer math shows 105056 versus 101831, a 3.1% lead. Random string sorting shows 40784 versus 39447, a 3.3% lead. Single-thread PassMark shows 3995 versus 3841, a 3.9% lead, matching the singlethread result exactly.

The Max PRO 385 wins every recorded benchmark, but the margin varies by workload. It dominates in single-core and cache-sensitive tests, while remaining ahead in memory-heavy and integer workloads by smaller margins. The 365's additional cores and threads do not translate into a win in any measured test, despite its higher core count.