AMD Ryzen 9 8945HX vs Intel Core Ultra 9 386H Comparison

FAQ

Q: Which processor has more cores and threads?

A: Both have 16 cores, but the AMD Ryzen 9 8945HX supports 32 threads while the Intel Core Ultra 9 386H supports 16 threads. The AMD part uses simultaneous multithreading, which is not present on the Intel part.

Q: What is the biggest single benchmark performance gap between the two?

A: In Cinebench R23 single-core, the AMD Ryzen 9 8945HX scores 6030 against 2071.5 for the Intel Core Ultra 9 386H, a delta of 191.1% in favor of AMD. This is the largest percentage difference recorded in the head-to-head data.

Q: Does the Intel processor win any benchmarks at all?

A: Yes. The Intel Core Ultra 9 386H wins 4 out of 17 head-to-head tests: Passmark find prime numbers, Passmark physics, and Passmark single-thread (both listings of the same test). The Intel part also has a higher single-thread score in Passmark, 4218 versus 3907.

Q: What are the average benchmark scores for each CPU?

A: The AMD Ryzen 9 8945HX has an average benchmark score of 76212, while the Intel Core Ultra 9 386H has an average of 43210. The AMD part sits at the 95th percentile of all CPUs, while the Intel part sits at the 88th percentile.

Q: What is the process node difference?

A: The AMD Ryzen 9 8945HX is built on a 5 nm process by TSMC, while the Intel Core Ultra 9 386H uses a 3 nm process from Intel. The Intel chip is also on a newer architecture, Panther Lake, versus Zen 4 (Dragon Range) for AMD.

Q: Which CPU has more L3 cache?

A: The AMD Ryzen 9 8945HX has 64 MB of L3 cache, while the Intel Core Ultra 9 386H has 18 MB of shared L3 cache. The AMD chip also has a larger L2 cache per core at 1 MB per core versus 2.5 MB per core for Intel, but the Intel L1 cache is larger at 192 KB per core versus 64 KB per core.

Architecture Differences

The AMD Ryzen 9 8945HX uses the Zen 4 architecture under the Dragon Range codename, part of the 8000 series. It is fabricated on a 5 nm process by TSMC, with a transistor count of 13,140 million spread across two 71 mm² dies. This is a high-performance mobile design with a 55 W TDP and an unlocked multiplier, allowing overclocking. The chip supports PCIe Gen 5 with 28 lanes from the CPU and uses dual-channel DDR5 memory with 83.2 GB/s bandwidth. Its integrated graphics is the Radeon 610M.

The Intel Core Ultra 9 386H belongs to the Core Ultra Series 3, built on the Panther Lake architecture (Panther Lake-H generation). It uses a 3 nm process from Intel, a more advanced node than the AMD part. The TDP is much lower at 25 W, and the multiplier is locked. It supports dual-channel DDR5 and LPDDR5X memory, with a higher memory bandwidth of 115.2 GB/s. PCIe support is Gen 5 but with only 12 lanes from the CPU. Its integrated graphics is Intel Xe3 Graphics.

A critical architectural difference is threading. The AMD chip has 16 cores and 32 threads, meaning each core can handle two threads. The Intel chip has 16 cores and 16 threads, with no multithreading. This explains why the AMD part dominates heavily in multi-threaded workloads. The cache layout also differs significantly: AMD uses 64 KB L1 per core, 1 MB L2 per core, and 64 MB shared L3; Intel uses 192 KB L1 per core, 2.5 MB L2 per core, and 18 MB shared L3. The Intel part has a smaller total L3 but larger per-core L1 and L2 caches, which may benefit certain latency-sensitive tasks.

The memory bandwidth advantage goes to Intel, with 115.2 GB/s versus 83.2 GB/s, but the AMD part offers more PCIe lanes (28 vs 12), which could matter for expansion in a laptop chassis. Both are mobile parts, and both are in active production. The AMD part was released on 2025-04-22, while the Intel part has a release date of 2026-01-04, making the Intel chip a newer generation.

Where Each One Wins

The AMD Ryzen 9 8945HX wins 13 of the 17 head-to-head benchmarks, and the margin is often substantial. Its clear domain is multi-threaded compute. In Cinebench R23 multi-core, it scores 42713 versus 20547, a 107.9% lead. In Passmark integer math, it scores 195180 versus 87284, a 123.6% lead. Data compression also favors AMD heavily: 677755 versus 352365, a 92.3% delta. For content creation, rendering, compilation, or any task that scales with thread count, the AMD part is the decisive winner.

The Intel Core Ultra 9 386H wins specific niche tests. In Passmark physics, it scores 3028 against 2168, a 28.4% advantage. In Passmark find prime numbers, it scores 341 versus 262, a 23.2% lead. The Intel part also wins Passmark single-thread with 4218 versus 3907, a 7.4% advantage. These wins suggest the Intel chip has a stronger per-core integer execution capability in certain algorithmic patterns, despite having fewer threads. The physics test often depends on single-core integer performance.

For general productivity, the AMD part is the safer choice. The Intel part's wins are isolated to specific workloads that do not translate into broad performance superiority. The average benchmark scores confirm this: AMD at 76212 versus Intel at 43210, which is a 76.4% difference in average score. The AMD part also ranks higher on the percentile scale, 95th versus 88th.

The Intel part does have a lower TDP (25 W versus 55 W), which could be relevant for thin-and-light designs, but the performance data shows that the AMD part is in a different class for heavy workloads. For users whose primary tasks are single-threaded or specific physics/prime calculations, the Intel part has a narrow edge. For everything else, the AMD part is ahead.

Specification Differences

The two CPUs differ in several key specification fields. The AMD Ryzen 9 8945HX has a base clock of 2.50 GHz and a boost clock of 5.40 GHz. The Intel Core Ultra 9 386H has a base clock of 2.10 GHz and a boost clock of 4.90 GHz. The AMD part runs at a 55 W TDP, while the Intel part runs at 25 W.

Socket differences are fundamental: AMD uses AMD Socket FL1, Intel uses Intel BGA 2540. The AMD part is based on Zen 4 (Dragon Range), the Intel part on Panther Lake. The process node differs: AMD uses 5 nm from TSMC, Intel uses 3 nm from Intel.

Cache layouts are different. AMD has 64 KB L1 per core, 1 MB L2 per core, and 64 MB L3. Intel has 192 KB L1 per core, 2.5 MB L2 per core, and 18 MB shared L3. Memory support: AMD only supports DDR5, while Intel supports DDR5 and LPDDR5X. Memory bandwidth: AMD is 83.2 GB/s, Intel is 115.2 GB/s.

PCIe lanes differ: AMD offers Gen 5 with 28 lanes (CPU only), Intel offers Gen 5 with 12 lanes (CPU only). Integrated graphics differ: AMD has Radeon 610M, Intel has Intel Xe3 Graphics. The multiplier is unlocked on AMD, locked on Intel. The transistor count and die size are listed for AMD (13,140 million transistors, 2x 71 mm²), but not available for Intel.

The part numbers differ: AMD is 100-000001848, Intel is SA4R5Q9EH. The release dates also differ, with AMD listed as 2025-04-22 and Intel as 2026-01-04.

Head-to-Head Benchmarks

The Cinebench suite shows a massive AMD advantage. In Cinebench R15 multi-core, AMD scores 4305 versus Intel's 3223, a 33.6% lead. In single-core R15, AMD scores 607 versus Intel's 303.5, a 100% lead. R20 multi-core: AMD 17939 versus Intel 12820, a 39.9% lead. R20 single-core: AMD 2532 versus Intel 1809, a 40% lead. R23 multi-core: AMD 42713 versus Intel 20547, a 107.9% lead. R23 single-core: AMD 6030 versus Intel 2071.5, a 191.1% lead.

The Passmark suite tells a similar story, with a few exceptions. Data compression: AMD 677755 versus Intel 352365, a 92.3% lead. Data encryption: AMD 40836 versus Intel 27150, a 50.4% lead. Extended instructions: AMD 49823 versus Intel 29138, a 71% lead. Floating point math: AMD 117453 versus Intel 108527, an 8.2% lead. Integer math: AMD 195180 versus Intel 87284, a 123.6% lead. Multithread: AMD 51405 versus Intel 35399, a 45.2% lead. Random string sorting: AMD 78781 versus Intel 42135, an 87% lead.

Intel wins in four tests. Find prime numbers: Intel 341 versus AMD 262, a 23.2% lead for Intel. Physics: Intel 3028 versus AMD 2168, a 28.4% lead. Single-thread (listed twice in the data): Intel 4218 versus AMD 3907, a 7.4% lead each time.

The largest single delta is the Cinebench R23 single-core test, where AMD leads by 191.1%. The smallest delta is the floating point math test, where AMD leads by 8.2%. The Intel wins are consistent in direction but smaller in magnitude compared to the AMD wins.

The Verdict

The data shows a clear overall winner. The AMD Ryzen 9 8945HX dominates in average benchmark score, 76212 versus 43210, and holds a higher percentile rank at 95 versus 88. It wins 13 of 17 head-to-head tests, including every Cinebench test and every Passmark test except the four specific ones noted.

For users who prioritize multi-threaded performance, rendering, compilation, data compression, or encryption, the AMD part is the only choice. Its 32 threads versus 16 threads gives it a structural advantage that no single-core win can overcome. The Cinebench R23 multi-core result, 42713 versus 20547, more than doubles the Intel score. The integer math test shows a 123.6% lead, indicating strong sustained throughput.

The Intel Core Ultra 9 386H has a narrow appeal. It wins in Passmark physics, find prime numbers, and single-thread performance. Those wins are real but isolated. The physics score of 3028 versus 2168 is a 28.4% advantage, and the single-thread score of 4218 versus 3907 is a 7.4% advantage. For a workload that is heavily dependent on a single thread and uses prime number algorithms or physics simulation, the Intel part is faster.

The Intel part also has a lower TDP of 25 W versus 55 W, which could matter in a power-constrained design, and it supports LPDDR5X memory with higher bandwidth. But the benchmark results indicate that the AMD part delivers vastly more performance in most scenarios, despite the higher power draw. The AMD part also offers an unlocked multiplier and more PCIe lanes.

The verdict from the recorded data is straightforward. The AMD Ryzen 9 8945HX is the superior processor for compute-heavy tasks, with a 76.4% higher average score and wins in the vast majority of tests. The Intel Core Ultra 9 386H is the better option only for specific single-threaded workloads, particularly physics and prime number calculations, where it shows a measurable but modest edge. For any general-purpose or multi-threaded use case, the database points decisively to the AMD part.