AMD Ryzen AI Max 390 vs Intel Core Ultra 5 245K Comparison

Head-to-Head Benchmarks

The benchmark data reveals a fascinating split between these two processors, with each claiming victories in distinctly different workloads. The AMD Ryzen AI Max 390 wins 6 of the 17 recorded tests, while the Intel Core Ultra 5 245K takes 11. Yet the margin of victory tells a more nuanced story than the raw win count suggests.

The most dramatic single result comes from Cinebench R23 single-core, where the AMD Ryzen AI Max 390 scores 5091 against the Intel Core Ultra 5 245K's 2132, a massive 138.8% advantage. This is not a small gap; it is a generational chasm in single-threaded rendering performance. The same pattern appears in Cinebench R15 single-core, where AMD leads 513 to 322, a 59.3% difference. These results indicate that the Zen 5 architecture in the Ryzen AI Max 390 delivers exceptional per-thread efficiency that the Arrow Lake design cannot match in these specific tests.

However, the Intel part strikes back in floating-point mathematics, scoring 130678 versus AMD's 90594, a 30.7% lead. This is one of the largest Intel victories in the entire dataset. The data also shows Intel winning the data encryption test by a substantial margin: 33286 versus 25097, a 24.6% advantage. Prime number finding also favors Intel, with 414 versus 316, a 23.7% gap.

The multi-core results are more mixed. Cinebench R15 multi-core favors Intel narrowly, 3850 versus 3635, a 5.6% difference. Cinebench R20 multi-core is nearly tied, with Intel at 15368 and AMD at 15146, a mere 1.4% gap. But Cinebench R23 multi-core flips decisively to AMD: 36064 versus 25066, a 43.9% lead. This inconsistency between R15, R20, and R23 suggests that workload scaling behavior differs significantly between the two architectures.

PassMark multi-thread shows Intel ahead by just 3% (43047 versus 41737), while PassMark integer math heavily favors AMD at 146519 versus 98524, a 48.7% advantage. Data compression also goes to AMD, 487145 versus 458817, a 6.2% edge. Extended instructions favor AMD by 2.9% (38716 versus 37617). Single-thread PassMark tests go to Intel, 4714 versus 4028, a 14.6% difference, which stands in contrast to the Cinebench single-core results.

Architecture Differences

The underlying designs explain much of this benchmark divergence. The AMD Ryzen AI Max 390 uses the Zen 5 architecture on a 4 nm TSMC process, with a codename of Strix Halo. It is a mobile-part designation, built for AMD Socket FP11. The processor contains 12 cores and 24 threads, with a base clock of 3.20 GHz and a boost clock of 5.00 GHz. Its cache hierarchy includes 80 KB of L1 per core, 1 MB of L2 per core, and a substantial 64 MB of shared L3 cache. Memory support is LPDDR5X with a quad-channel bus delivering 256.0 GB/s of bandwidth. The integrated graphics are Radeon 8050S, and the chip supports ECC memory. The die size is listed as 2x 70.6 mm².

The Intel Core Ultra 5 245K belongs to the Core Ultra Series 2, built on the Arrow Lake architecture with an Arrow Lake-S codename. It uses a 3 nm TSMC process and is a desktop part for Intel Socket 1851. This processor has 14 cores but only 14 threads, indicating no hyperthreading on any core. Base clock is 4.20 GHz, boost clock is 5.20 GHz. Cache is configured as 192 KB L1 per core, 3 MB L2 per core, and 24 MB shared L3. Memory support is DDR5 with a dual-channel bus providing 102.4 GB/s, less than half of AMD's bandwidth. The integrated graphics are Arc Xe-LPG Graphics 64EU. The transistor count is 17,800 million, and the die size is 243 mm². PCIe support is Gen 5 with 20 lanes, versus AMD's Gen 4 with 16 lanes. Intel's multiplier is unlocked, while AMD's is not.

The thread count difference is critical: AMD offers 12 cores with simultaneous multithreading for 24 threads, while Intel provides 14 physical cores with no multithreading, giving 14 threads. This explains why AMD can win heavily in integer math and Cinebench R23 multi-core, where thread-level parallelism and cache efficiency matter greatly. The 64 MB L3 cache on AMD versus 24 MB on Intel is a substantial difference that likely contributes to AMD's data compression victory.

Where Each One Wins

The data points toward a clear workload differentiation. The AMD Ryzen AI Max 390 excels in scenarios that stress single-core integer performance and large shared cache. Its Cinebench R23 single-core score of 5091, which is 138.8% above Intel, suggests exceptional performance in lightly threaded rendering tasks, legacy software, and any application that relies on a single fast core. The 64 MB L3 cache likely boosts data compression, where AMD wins by 6.2%, and integer math, where it wins by 48.7%. For users running heavily threaded workloads that scale with many threads and large caches, such as video encoding or complex simulations, the Cinebench R23 multi-core result of 36064 versus 25066 indicates a 43.9% advantage.

The Intel Core Ultra 5 245K dominates floating-point math, winning by 30.7%, and also excels in encryption (24.6% lead), prime number calculations (23.7% lead), and physics simulations (10.4% lead). The PassMark single-thread score of 4714 versus 4028, a 14.6% advantage, indicates Intel's faster per-core clock speed in certain instruction streams. Data encryption workloads, which often rely on AES-NI and similar instruction sets, clearly favor Intel. The higher boost clock of 5.20 GHz versus 5.00 GHz may explain part of this, but the architecture itself appears better suited to cryptographic and floating-point intensive tasks.

Random string sorting, a memory-latency-sensitive task, goes to Intel by 3.6% (55098 versus 53113). The PassMark multi-thread test narrowly favors Intel by 3%, suggesting that in balanced workloads with moderate thread counts, Intel's 14 physical cores can compete effectively despite fewer threads.

The Verdict

The recorded data supports a clear conclusion: choose the AMD Ryzen AI Max 390 for single-threaded rendering, integer-heavy computation, and cache-sensitive workloads. Its 138.8% lead in Cinebench R23 single-core and 48.7% lead in integer math make it the superior choice for developers, 3D artists, and anyone running applications that depend on per-core speed and large L3 caches. The 64 MB L3 cache and 24 threads provide a strong foundation for multi-threaded integer workloads.

Choose the Intel Core Ultra 5 245K for floating-point mathematics, encryption, and physics-based simulations. Its 30.7% lead in floating-point math and 24.6% lead in data encryption are substantial. The 5.20 GHz boost clock and 14 physical cores without hyperthreading deliver consistent performance in tasks that do not benefit from extra threads. The unlocked multiplier also allows for tuning, though the benchmark data does not include overclocked results.

For general-purpose use, the PassMark multi-thread results (Intel 43047 versus AMD 41737) and single-thread results (Intel 4714 versus AMD 4028) suggest Intel holds a slight edge in aggregate PassMark workloads. However, the Cinebench suite, which is more representative of rendering and content creation, splits almost evenly between the two: AMD wins R23 multi-core and single-core, Intel wins R15 and R20 multi-core and R20 single-core. The average benchmark score in the database places AMD at 56273 and Intel at 54053, a 4.1% difference favoring AMD, though both processors sit at the 91st percentile of all CPUs.

FAQ

Q: Which processor has a higher single-core score in Cinebench R23?

A: The AMD Ryzen AI Max 390 scores 5091, which is 138.8% higher than the Intel Core Ultra 5 245K's 2132.

Q: Does the Intel Core Ultra 5 245K have more cores?

A: Yes, Intel has 14 cores, but only 14 threads. AMD has 12 cores and 24 threads due to simultaneous multithreading.

Q: Which chip wins in floating-point math?

A: The Intel Core Ultra 5 245K leads with a PassMark floating-point math score of 130678, versus AMD's 90594, a 30.7% advantage.

Q: What is the cache difference between these two?

A: AMD has 64 MB of shared L3 cache, while Intel has 24 MB of shared L3 cache. AMD also has 1 MB L2 per core versus Intel's 3 MB L2 per core.

Q: Which processor has higher memory bandwidth?

A: AMD supports quad-channel LPDDR5X with 256.0 GB/s, while Intel supports dual-channel DDR5 with 102.4 GB/s.

Q: How many benchmark tests does each processor win?

A: The AMD Ryzen AI Max 390 wins 6 tests, and the Intel Core Ultra 5 245K wins 11 tests out of the 17 recorded head-to-head benchmarks.

Specification Differences

| Specification | AMD Ryzen AI Max 390 | Intel Core Ultra 5 245K |

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

| Cores | 12 | 14 |

| Threads | 24 | 14 |

| Base Clock | 3.20 GHz | 4.20 GHz |

| Boost Clock | 5.00 GHz | 5.20 GHz |

| TDP | 55 W | 125 W |

| Socket | AMD Socket FP11 | Intel Socket 1851 |

| Architecture | Zen 5 | Arrow Lake |

| Codename | Strix Halo | Arrow Lake-S |

| Process Node | 4 nm | 3 nm |

| Transistors | Not listed | 17,800 million |

| Die Size | 2x 70.6 mm² | 243 mm² |

| L1 Cache | 80 KB (per core) | 192 KB (per core) |

| L2 Cache | 1 MB (per core) | 3 MB (per core) |

| L3 Cache | 64 MB (shared) | 24 MB (shared) |

| Memory Support | LPDDR5X | DDR5 |

| Memory Bus | Quad-channel | Dual-channel |

| Memory Bandwidth | 256.0 GB/s | 102.4 GB/s |

| PCIe | Gen 4, 16 Lanes | Gen 5, 20 Lanes |

| Integrated Graphics | Radeon 8050S | Arc Xe-LPG Graphics 64EU |

| Market Segment | Mobile | Desktop |

| Release Date | 2025-01-05 | 2024-10-23 |

| Launch MSRP | Not listed | $319 |

| Multiplier Unlocked | No | Yes |