AMD Ryzen 9 9950X vs Intel Core Ultra X7 368H Comparison

Where Each One Wins

The AMD Ryzen 9 9950X dominates the shared benchmark suite with 13 wins out of 15 head-to-head comparisons. Its victories span every category of multi-threaded workload, from rendering and compression to encryption and physics simulation. The Intel Core Ultra X7 368H claims exactly two wins, both in single-core Cinebench tests, which points to a clear division: AMD for throughput, Intel for raw per-thread speed.

The data shows the Ryzen 9 9950X is the clear choice for heavily parallel workloads. In Cinebench R23 multi-core, it scores 40,924 against the Intel chip's 28,221, a 45% advantage. That gap widens dramatically in PassMark integer math, where AMD leads by 174.9%, and in data compression, where the lead reaches 186.5%. These are the workloads that scale with core count, thread count, and cache capacity, all areas where the AMD part is built to excel.

The Intel Core Ultra X7 368H wins in Cinebench R15 single-core (401 vs. 344, a 14.2% advantage) and Cinebench R23 single-core (3,984 vs. 2,202, a 44.7% advantage). That R23 single-core result is striking: Intel's per-thread performance is nearly double AMD's in that specific test, which suggests a fundamentally different design philosophy. For lightly threaded tasks like legacy applications or certain simulation tools that refuse to use more than one core, the Intel chip has a strong argument.

However, even in PassMark single-thread, the AMD processor wins at 4,737 vs. 4,005, an 18.3% margin. The Intel single-core wins are confined to Cinebench's rendering engine, not general single-thread performance. This is a nuanced picture: Intel's architecture shows exceptional efficiency in one benchmark, but AMD's broader single-thread score remains higher.

For mixed workloads, the Ryzen 9 9950X wins PassMark multi-thread by 100.4% (66,030 vs. 32,956) and PassMark physics by 14.8% (3,279 vs. 2,857). The Intel chip's two Cinebench single-core wins do not offset the massive multi-core deficits in any practical sense. The AMD part also wins PassMark floating point math by 50.5%, data encryption by 75.9%, and random string sorting by 147.7%, covering an unusually broad range of computational patterns.

Architecture Differences

The two processors come from opposite ends of the computing spectrum. The AMD Ryzen 9 9950X is a desktop part built on Zen 5 architecture with the Granite Ridge codename, manufactured on a 4 nm process at TSMC. It has 16 cores and 32 threads, meaning each core supports two threads simultaneously. The Intel Core Ultra X7 368H is a mobile part using Panther Lake architecture, also codenamed Panther Lake, built on Intel's own 3 nm process. It also has 16 cores, but only 16 threads, indicating no simultaneous multithreading.

The cache hierarchy tells a different story. AMD uses 80 KB of L1 cache per core, 1 MB of L2 per core, and 64 MB of shared L3 cache. Intel uses 192 KB of L1 per core, 2.5 MB of L2 per core, and only 18 MB of shared L3. The AMD chip has far more total cache, especially at the L3 level, which explains its dominance in data-heavy workloads like compression and sorting. Intel's larger per-core L1 and L2 caches help with single-thread latency, which aligns with its Cinebench single-core wins.

Clock speeds also differ sharply. The AMD part runs at a 4.30 GHz base clock and boosts to 5.70 GHz. The Intel chip has a 2.00 GHz base clock and a 5.00 GHz boost clock. The lower base clock on the Intel part is typical for mobile processors, but the boost clock is still 700 MHz lower than AMD's. The thermal design power (TDP) figures are even more divergent: AMD is rated at 170 watts, while Intel is rated at just 25 watts. That is a 145-watt gap, which explains why the AMD part can sustain higher clocks and feed its 32 threads, while the Intel chip is designed for battery-powered systems.

Memory support also differs. AMD uses DDR5 with dual-channel support and 89.6 GB/s bandwidth. Intel uses LPDDR5X, also dual-channel, but with 153.6 GB/s bandwidth, a 71.5% higher figure. The Intel chip also integrates Arc B390 graphics, while AMD includes Radeon Graphics. Intel does not support ECC memory; AMD does. PCIe connectivity differs too: AMD offers Gen 5 with 24 lanes, Intel offers Gen 5 with only 4 lanes, a reflection of the mobile socket's constraints.

The AMD processor is unlocked for overclocking and uses the AMD Socket AM5. The Intel chip is locked and uses Intel BGA 2540, a soldered mobile package. The AMD part has a production status of Active, as does the Intel part, but the AMD release date is August 2024 while the Intel release date is January 2026.

Head-to-Head Benchmarks

The largest single win for AMD comes in PassMark data compression, where the Ryzen 9 9950X scores 896,544 against the Intel chip's 312,927, a 186.5% advantage. This workload benefits from the AMD part's 64 MB of L3 cache and 32 threads. The second-largest win is in PassMark extended instructions, at 178.8% (71,564 vs. 25,669), followed by integer math at 174.9% (242,097 vs. 88,083). These three tests show the AMD part is over 2.7 times faster in the most compute-dense integer operations.

The Cinebench R15 multi-core test shows a 122.7% lead for AMD (6,335 vs. 2,844), while Cinebench R23 multi-core shows a 45% lead (40,924 vs. 28,221). The difference between these two margins suggests that the Intel chip scales better as the workload gets larger, but still cannot close the gap. PassMark multi-thread shows a 100.4% lead (66,030 vs. 32,956), and random string sorting shows a 147.7% lead (94,368 vs. 38,093).

AMD also wins PassMark floating point math by 50.5% (159,063 vs. 105,681), data encryption by 75.9% (44,366 vs. 25,228), and PassMark physics by 14.8% (3,279 vs. 2,857). The smallest AMD win is PassMark find prime numbers, at just 7.5% (345 vs. 321), which shows that even in a workload where the Intel chip is relatively competitive, AMD still edges ahead.

The Intel wins are concentrated in Cinebench R15 single-core, where it beats AMD by 14.2% (401 vs. 344), and Cinebench R23 single-core, where it beats AMD by 44.7% (3,984 vs. 2,202). The R23 single-core result is particularly anomalous because AMD wins PassMark single-thread by 18.3% (4,737 vs. 4,005). This means the Intel chip's single-core advantage is not universal; it is specific to the Cinebench rendering engine, which may favor Intel's larger per-core L1 and L2 caches.

FAQ

Q: Which processor has more threads?

A: The AMD Ryzen 9 9950X has 32 threads from 16 cores. The Intel Core Ultra X7 368H has 16 threads from 16 cores, meaning it does not use simultaneous multithreading.

Q: Does the Intel chip have better single-core performance?

A: In Cinebench R23 single-core, yes: Intel scores 3,984 vs. AMD's 2,202, a 44.7% advantage. However, in PassMark single-thread, AMD wins at 4,737 vs. 4,005, an 18.3% margin. The Intel advantage is limited to Cinebench's single-core tests.

Q: Which processor has more L3 cache?

A: The AMD Ryzen 9 9950X has 64 MB of shared L3 cache. The Intel Core Ultra X7 368H has 18 MB of shared L3 cache. AMD also has 1 MB of L2 per core vs. Intel's 2.5 MB per core, and 80 KB of L1 per core vs. Intel's 192 KB per core.

Q: What is the memory bandwidth difference?

A: The Intel chip uses LPDDR5X with 153.6 GB/s bandwidth. The AMD chip uses DDR5 with 89.6 GB/s bandwidth. Intel's memory bandwidth is 71.5% higher, but this does not translate to benchmark wins in the recorded data.

Q: Which processor supports ECC memory?

A: The AMD Ryzen 9 9950X supports ECC memory. The Intel Core Ultra X7 368H does not.

Q: Are these processors in the same market segment?

A: No. The AMD Ryzen 9 9950X is a desktop processor on AMD Socket AM5. The Intel Core Ultra X7 368H is a mobile processor on Intel BGA 2540. The AMD part has a 170 watt TDP, while the Intel part has a 25 watt TDP.

Specification Differences

| Specification | AMD Ryzen 9 9950X | Intel Core Ultra X7 368H |

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

| Cores | 16 | 16 |

| Threads | 32 | 16 |

| Base clock | 4.30 GHz | 2.00 GHz |

| Boost clock | 5.70 GHz | 5.00 GHz |

| TDP | 170 W | 25 W |

| Socket | AMD Socket AM5 | Intel BGA 2540 |

| Architecture | Zen 5 | Panther Lake |

| Process node | 4 nm (TSMC) | 3 nm (Intel) |

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

| L2 cache | 1 MB (per core) | 2.5 MB (per core) |

| L3 cache | 64 MB | 18 MB (shared) |

| Memory support | DDR5 | LPDDR5X |

| Memory bandwidth | 89.6 GB/s | 153.6 GB/s |

| ECC support | Yes | No |

| PCIe lanes | Gen 5, 24 lanes | Gen 5, 4 lanes |

| Integrated graphics | Radeon Graphics | Arc B390 |

| Market segment | Desktop | Mobile |

| Release date | August 2024 | January 2026 |

| Unlocked multiplier | Yes | No |

The Verdict

The recorded data strongly favors the AMD Ryzen 9 9950X for any workload that can use more than a few threads. It wins 13 of 15 benchmarks, with an average benchmark score of 74,640 compared to the Intel chip's 40,518, a difference of about 84%. Its percentile rank among all CPUs is 94, while the Intel chip sits at 87. The AMD part also ranks close to the AMD Ryzen 7 PRO 9745 (0.2% lower average score) and the AMD Ryzen AI 9 HX PRO 475 (0.2% higher), placing it in the company of workstation-class processors.

The Intel Core Ultra X7 368H is competitive in specific single-threaded Cinebench tests, where it beats the AMD part by 14.2% and 44.7% respectively. It also delivers higher memory bandwidth (153.6 GB/s vs. 89.6 GB/s) and uses a more advanced 3 nm process. However, its 25 watt TDP and 16 threads limit its overall throughput. The data shows it sits among rivals like the Intel Core 7 253PE (0.1% lower) and Intel Core 5 223PE (0.2% lower), which are similarly constrained mobile parts.

For desktop users with access to power and cooling, the AMD Ryzen 9 9950X is the clear choice based on the benchmark data. For mobile users who prioritize single-thread rendering performance and low power consumption, the Intel chip has a narrow but measurable edge in Cinebench single-core tests. The AMD part's 64 MB of L3 cache and 32 threads make it the better option for the vast majority of recorded workloads, while the Intel chip's strengths are confined to a narrow slice of the benchmark suite.