AMD Ryzen 7 8840HX vs AMD Ryzen 7 PRO 9745 Comparison

The AMD Ryzen 7 PRO 9745 and the AMD Ryzen 7 8840HX are two very different interpretations of what an 8-to-12-core Ryzen 7 chip should be. One is a socketed Zen 5 desktop part aimed at server and workstation duty, the other a mobile Dragon Range processor with more cores and a lower thermal envelope. The recorded benchmark data shows a clean split: the 9745 dominates in per-core responsiveness, while the 8840HX wins the multithreaded war. Here is how the numbers break down.

The Verdict

The data points in opposite directions depending on workload. The Ryzen 7 PRO 9745 is the pick when single-thread speed matters most: its 4,624 single-thread score beats the 8840HX by 16.8 percent, and its overall average score of 74,761 sits in the 95th percentile of all CPUs in the database. It also offers ECC memory support, a workstation-grade feature the 8840HX lacks entirely. Choose it for latency-sensitive professional applications, platform stability requirements, and any scenario where clock speed per core drives performance.

The Ryzen 7 8840HX is the pick for throughput. It wins 8 of 11 head-to-head benchmarks, including a 38.2 percent victory in prime number finding and a 23.6 percent win in encryption, on the strength of 12 cores and 24 threads versus 8 cores and 16 threads. Its average score of 71,797 lands it in the 94th percentile, effectively tied with rivals like the Intel Core Ultra 7 265KF, which it beats by a hair. If sustained parallel workloads are the job, the 8840HX delivers more work per system despite being a mobile part.

Architecture Differences

These chips come from different generations and different design philosophies. The 9745 is a Granite Ridge processor built on TSMC's 4 nm node, running Zen 5 cores at a 3.80 GHz base and 5.40 GHz boost. The 8840HX is a Dragon Range design on the older 5 nm node using Zen 4 cores, clocking from 2.90 GHz base to 5.10 GHz boost. The node and architectural advantage explains the single-thread gap in the 9745's favor; the core count advantage explains nearly everything else.

Cache configurations diverge sharply. The 9745 pairs 80 KB of L1 and 1 MB of L2 per core with a 32 MB shared L3. The 8840HX has smaller 64 KB L1 slices but doubles the shared L3 to 64 MB, again at 1 MB of L2 per core. That larger L3 feeds the extra cores in long-running parallel tasks.

Physically, the 8840HX is actually the bigger design: 13,140 million transistors across a two-die layout of 2x 71 mm², versus 8,315 million transistors on a single 70.6 mm² die for the 9745. The chiplet-style construction is how AMD fits 12 cores into the Dragon Range package.

Platform features split both ways. The 9745 uses Socket AM5 with 24 PCIe 5.0 lanes and supports DDR5 at 89.6 GB/s of bandwidth in a dual-channel configuration. The 8840HX uses Socket FL1 with 28 PCIe 5.0 lanes, four more than the 9745, but slightly lower memory bandwidth at 83.2 GB/s. ECC support exists only on the 9745. Integrated graphics differ too: Radeon Graphics on the 9745 versus Radeon 610M on the 8840HX. Two more distinctions: the 8840HX has an unlocked multiplier for overclocking, the 9745 does not, and the 8840HX carries a 55 W TDP against 65 W for the 9745. Both remain in active production, with the 8840HX shipping in April 2025 and the 9745 in September 2025.

Head-to-Head Benchmarks

The 8840HX sweeps the throughput tests. Its largest win is passmark_find_prime_numbers, where it scores 304 against 188, a 38.2 percent gap. Data encryption shows the same pattern: 29,919 versus 22,848, a 23.6 percent advantage for the mobile chip. Integer math lands at 146,506 against 124,167, a 15.2 percent win, and string sorting goes to the 8840HX at 57,971 versus 49,219, down 15.1 percent from the 9745's perspective.

The overall multithread picture confirms it. Passmark_multithread reads 41,732 for the 8840HX against 38,201 for the 9745, an 8.5 percent margin. Data compression favors the 8840HX by 7.9 percent (496,427 versus 457,211), physics by 7.3 percent (2,185 versus 2,026), and floating point math by 4.9 percent (86,747 versus 82,461).

The 9745 answers where clock and architecture matter. Its single-thread score of 4,624 versus 3,958 is a 16.8 percent lead, the single biggest margin either chip holds over the other outside of prime finding. Extended instructions also tilt red in its favor, 36,800 versus 36,527, though at just 0.7 percent that result is effectively a tie.

Context matters here. The 9745's average of 74,761 places it in rare company: within 0.2 percent of the Ryzen 9 9950X (74,640), ahead of the Ryzen AI 9 HX PRO 475 (74,496) by 0.4 percent, and just 0.8 to 1 percent behind the EPYC 4545P (75,373) and Intel Core Ultra 9 285 (75,488). The 8840HX's 71,797 sits within a point of an all-Intel rival set: the Core Ultra 7 265KF at 71,910, the Xeon Platinum 8270 at 71,370, and the Xeon 6517P and 6724P at 72,350 and 72,396. Both CPUs occupy the top few percent of the database, just with different specializations.

FAQ

Q: Which CPU is faster overall?

A: It depends on the metric. The 8840HX wins 8 of 11 head-to-head benchmarks and leads multithread performance by 8.5 percent, but the 9745 has the higher average benchmark score (74,761 versus 71,797) thanks to its 16.8 percent single-thread lead.

Q: Which CPU is better for single-threaded workloads?

A: The 9745, clearly. Its 4,624 single-thread score beats the 8840HX's 3,958 by 16.8 percent, driven by the newer Zen 5 architecture on a 4 nm node and a 5.40 GHz boost clock.

Q: Which is better for heavily parallel workloads?

A: The 8840HX. With 12 cores and 24 threads versus 8 cores and 16 threads, it leads multithread by 8.5 percent, prime finding by 38.2 percent, encryption by 23.6 percent, and integer math by 15.2 percent.

Q: Do both CPUs support ECC memory?

A: No. Only the 9745 supports ECC, consistent with its server/workstation market segment. The 8840HX, a mobile-segment part, does not.

Q: Can either CPU be overclocked?

A: Only the 8840HX, which has an unlocked multiplier. The 9745 is multiplier-locked.

Q: Which platform has more PCIe lanes?

A: The 8840HX offers 28 PCIe 5.0 lanes against 24 on the 9745, both CPU-only counts. The 9745 counters with slightly higher memory bandwidth, 89.6 GB/s versus 83.2 GB/s.

Where Each One Wins

The Ryzen 7 8840HX wins wherever parallelism is king. Data compression, encryption, integer math, string sorting, physics simulation, prime search, and floating point all go its way, most by double-digit margins. Compilation jobs, rendering passes, simulation, batch media encoding, and cryptographic workloads are its natural habitat, and it does all this at a 55 W TDP from a mobile-oriented package. Its unlocked multiplier also makes it the only option of the two for tuners. The 28 PCIe 5.0 lanes give it an edge for multi-device storage or expansion configurations.

The Ryzen 7 PRO 9745 wins on speed per core and platform rigor. Its 16.8 percent single-thread advantage makes it the better choice for applications that lean on one fast thread: many CAD tools, game engines, scripting runtimes, and interactive professional software. Its 95th percentile average score, ECC support, and server/workstation positioning make it the safer fit for stability-critical deployments on the AM5 platform. When the workload is bursty, latency-sensitive, or thread-limited, the data says take the 9745; when the workload scales across cores, take the 8840HX.