AMD Ryzen 7 9850X3D vs Intel Core Ultra 5 235H Comparison

The Verdict

The database places the AMD Ryzen 7 9850X3D and Intel Core Ultra 5 235H in different market segments, and the benchmark data confirms that split. The AMD part is a desktop processor with a 120 TDP, 8 cores, and 16 threads, while the Intel part is a mobile processor with a 28 TDP, 14 cores, and 14 threads. These are not direct competitors in the traditional sense, but the head-to-head results show a clear division of labor.

The AMD Ryzen 7 9850X3D wins 10 of the 15 recorded head-to-head benchmarks, including every PassMark workload except floating point math and data encryption. Its average benchmark score is 58386, placing it in the 92nd percentile of all CPUs. The Intel Core Ultra 5 235H wins 5 benchmarks, including both Cinebench R23 tests and the single-core Cinebench R15 test, with an average score of 37522 and an 85th percentile ranking.

The verdict from the data is straightforward. The AMD Ryzen 7 9850X3D is for users who need raw multi-threaded throughput, integer math, compression, encryption-adjacent workloads, and physics calculations. The Intel Core Ultra 5 235H is for users who prioritize single-core Cinebench performance and floating point math, and who operate within a much lower power envelope. The Intel chip also delivers a surprising Cinebench R23 multi-core win despite having fewer threads, which points to its higher boost clock efficiency in that specific workload.

Architecture Differences

The AMD Ryzen 7 9850X3D uses the Zen 5 architecture on the Granite Ridge codename, built on TSMC's 4 nm process. It packs 8,315 million transistors on a 70.6 mm² die. The Intel Core Ultra 5 235H uses the Arrow Lake architecture on the Arrow Lake-H codename, built on TSMC's 3 nm process. The database does not record transistor count or die size for the Intel part.

The core configurations differ significantly. The AMD chip has 8 cores and 16 threads, meaning it uses simultaneous multithreading. The Intel chip has 14 cores and 14 threads, meaning it has no hyperthreading and relies on physical cores alone. The AMD chip's cache layout includes 80 KB of L1 per core, 1 MB of L2 per core, and 96 MB of shared L3 cache. The Intel chip has 192 KB of L1 per core, 3 MB of L2 per core, and 18 MB of shared L3 cache. The AMD chip's 96 MB L3 is a defining feature, and it likely explains the large wins in data compression and extended instructions.

Both processors are manufactured by TSMC, but on different nodes. The AMD part uses 4 nm, the Intel part uses 3 nm. The AMD chip is a desktop part on AMD Socket AM5 with 24 PCIe Gen 5 lanes (CPU only), while the Intel chip is a mobile part on Intel BGA 2049 with 8 PCIe Gen 5 lanes (CPU only). The AMD chip supports ECC memory and has an unlocked multiplier; the Intel chip does not support ECC and has a locked multiplier.

Memory support also differs. Both use dual-channel memory, but the AMD chip supports DDR5 only with 89.6 GB/s bandwidth, while the Intel chip supports DDR5 and LPDDR5X with 102.4 GB/s bandwidth. The Intel chip's integrated graphics is Arc Graphics 140T, while the AMD chip uses Radeon Graphics.

Head-to-Head Benchmarks

The largest win for the AMD Ryzen 7 9850X3D comes in PassMark physics, where it scores 4171 against 1985 for the Intel chip, a 110.1% advantage. This is the single biggest delta in the entire comparison. The AMD chip also dominates PassMark extended instructions with 39272 against 23354, a 68.2% lead, and PassMark integer math with 123140 against 74247, a 65.9% lead. Data compression shows a 57.1% advantage for AMD (474501 against 301979), and random string sorting shows a 37.4% advantage (49764 against 36208). The AMD chip wins PassMark multithread by 37.3% (41318 against 30091), and PassMark find prime numbers by 82% (435 against 239). Single-thread PassMark goes to AMD by 7.9% (4704 against 4359).

The Intel Core Ultra 5 235H wins the Cinebench R23 multi-core test with 25598 against 22807, a 10.9% advantage. This is notable because the Intel chip has 14 threads versus the AMD chip's 16 threads, yet it still wins in this specific workload. The Intel chip also wins Cinebench R23 single-core by a massive 38.3% (3613 against 2228), and Cinebench R15 single-core by 5.8% (364 against 343). In PassMark floating point math, the Intel chip wins with 93509 against 81998, a 12.3% advantage. The data encryption test is close: Intel wins with 23121 against 22856, a 1.1% margin that is nearly a tie.

The Cinebench R15 multi-core test goes to AMD with 3551 against 2580, a 37.6% advantage. This creates an interesting split within the Cinebench suite itself: AMD wins R15 multi-core, Intel wins R23 multi-core, and Intel wins both single-core tests. The database records these results without explaining the discrepancy, but the pattern suggests workload-specific scaling rather than a general multi-core superiority for either chip.

Specification Differences

The two processors differ across nearly every specification field recorded in the database. The AMD Ryzen 7 9850X3D has a base clock of 4.70 GHz and a boost clock of 5.60 GHz. The Intel Core Ultra 5 235H has a base clock of 2.40 GHz and a boost clock of 5.00 GHz. The AMD chip's TDP is 120, while the Intel chip's TDP is 28. The AMD chip uses AMD Socket AM5, the Intel chip uses Intel BGA 2049.

Core and thread counts differ as noted: 8 cores and 16 threads for AMD, 14 cores and 14 threads for Intel. The AMD chip's process node is 4 nm, the Intel chip's is 3 nm. The AMD chip has 8,315 million transistors on a 70.6 mm² die, while no transistor or die size data is recorded for the Intel chip.

Cache configurations differ in both per-core and shared amounts. AMD uses 80 KB L1 per core, 1 MB L2 per core, and 96 MB shared L3. Intel uses 192 KB L1 per core, 3 MB L2 per core, and 18 MB shared L3. Memory support: AMD supports DDR5 only with 89.6 GB/s bandwidth; Intel supports DDR5 and LPDDR5X with 102.4 GB/s bandwidth. ECC memory support is true for AMD, false for Intel.

PCIe lane counts differ: AMD has 24 Gen 5 lanes (CPU only), Intel has 8 Gen 5 lanes (CPU only). Integrated graphics differ: AMD uses Radeon Graphics, Intel uses Arc Graphics 140T. Market segment differs: AMD is Desktop, Intel is Mobile. The AMD chip has an unlocked multiplier, the Intel chip does not. The AMD chip's launch MSRP is $499; no launch MSRP is recorded for the Intel chip.

The AMD chip was released on 2026-01-28, while the Intel chip was released on 2025-01-12. The AMD part number is 100-000001973, the Intel part number is SRQAP.

FAQ

Q: Which processor has more cores and threads?

A: The Intel Core Ultra 5 235H has 14 cores, but the AMD Ryzen 7 9850X3D has more threads at 16. The Intel chip has 14 threads total, one per core. The AMD chip has 8 cores and 16 threads.

Q: Which chip wins the Cinebench R23 multi-core test?

A: The Intel Core Ultra 5 235H wins with a score of 25598 against 22807 for the AMD Ryzen 7 9850X3D, a 10.9% advantage. This is despite the AMD chip having more threads.

Q: What is the biggest single benchmark win in this comparison?

A: The AMD Ryzen 7 9850X3D wins PassMark physics with 4171 against 1985 for the Intel chip, a 110.1% advantage. This is the largest delta in the recorded head-to-head data.

Q: Does the AMD chip have more cache than the Intel chip?

A: Yes. The AMD Ryzen 7 9850X3D has 96 MB of shared L3 cache, while the Intel Core Ultra 5 235H has 18 MB of shared L3. The Intel chip has larger per-core L1 and L2 caches (192 KB and 3 MB per core) compared to AMD's 80 KB and 1 MB per core.

Q: Which chip has a higher memory bandwidth?

A: The Intel Core Ultra 5 235H has higher recorded memory bandwidth at 102.4 GB/s, compared to 89.6 GB/s for the AMD Ryzen 7 9850X3D. The Intel chip also supports LPDDR5X in addition to DDR5.

Q: Which chip has a higher PassMark single-thread score?

A: The AMD Ryzen 7 9850X3D wins PassMark single-thread with 4704 against 4359 for the Intel chip, a 7.9% advantage. However, the Intel chip wins Cinebench R23 single-core by 38.3% (3613 against 2228) and Cinebench R15 single-core by 5.8% (364 against 343).

Where Each One Wins

The AMD Ryzen 7 9850X3D wins in workloads that stress multi-threaded throughput and large cache reuse. Its PassMark multithread score of 41318 is 37.3% ahead of the Intel chip's 30091. Its physics score of 4171 is more than double the Intel chip's 1985, a 110.1% lead that suggests heavy compute loops benefit from the AMD architecture. Integer math shows a 65.9% advantage, and extended instructions show a 68.2% advantage. Data compression and random string sorting both favor AMD by roughly 57% and 37% respectively. These results point to a processor that handles data transformation, sorting, and compute-heavy parallel workloads with a clear edge.

The Intel Core Ultra 5 235H wins in the Cinebench family and floating point math. Its Cinebench R23 multi-core score of 25598 beats the AMD chip by 10.9%, which is the most surprising result given the thread count disadvantage. Its Cinebench R23 single-core score of 3613 is 38.3% ahead of the AMD chip's 2228, and its Cinebench R15 single-core score of 364 beats AMD's 343 by 5.8%. Floating point math favors Intel by 12.3% (93509 against 81998). Data encryption is effectively a tie, with Intel winning by 1.1%.

The use-case split is clear. The AMD Ryzen 7 9850X3D is for desktop users who run rendering, physics simulation, compression, sorting, and integer-heavy tasks. The Intel Core Ultra 5 235H is for mobile users who need strong single-core Cinebench performance, floating point throughput, and a 28 TDP envelope. The AMD chip sits in the 92nd percentile of all CPUs, while the Intel chip sits in the 85th percentile. The AMD chip's nearest rivals include the Intel Core i9-14900 with a 0.5% delta and the AMD Ryzen AI 9 HX 470 with a -0.7% delta, placing it among high-end desktop parts. The Intel chip's nearest rivals include the Intel Core i9-13900HK and Intel Core i5-13600K, both within a 0.4% delta, placing it in the upper mid-range of mobile and older desktop parts. The data does not support treating these as interchangeable; they serve different sockets, different power budgets, and different workload priorities.