Intel Core i9-13905H

Intel Core i9-13905H is a 14-core, 20-thread Raptor Lake-H mobile processor built on Intel's 10 nm process, with a base clock of 2.60 GHz and a boost clock of 5.40 GHz. It sits in the 91st percentile of all CPUs benchmarked, with an average benchmark score of 40712, placing it in a tight competitive cluster where its nearest rivals are separated by roughly one percentage point or less.

Single-Thread vs Multi-Thread Behavior

The benchmark data reveals a processor engineered for asymmetric workloads. In Cinebench R23, the i9-13905H scores 3583 points in single-core and 25380 points in multi-core, a ratio that indicates the multi-core result is roughly 7.1 times higher. This gap is substantial, but it does not tell the whole story: the single-core score of 3583 is a strong figure that benefits directly from the 5.40 GHz boost clock, while the multi-core figure draws on the hybrid arrangement of 14 cores and 20 threads.

Looking at the Passmark suite, the split becomes even more pronounced. The single-thread score is 3703, while the multithread score reaches 29779, an 8-to-1 ratio. The data compression test (354330) and integer math (101716) both show heavy scaling from the thread count, whereas the find prime numbers test (122) and physics test (2054) are far more modest, suggesting that certain integer-heavy or latency-sensitive tasks do not benefit equally from the 20-thread configuration. The floating-point math score of 73434 and extended instructions score of 21297 further indicate strong SIMD throughput, but the random string sorting score of 37460 is comparatively lower, hinting at memory-access patterns that limit parallel gains.

Real-world implications: applications that can saturate 20 threads—video encoding, 3D rendering, code compilation—will see near-linear scaling, while lightly threaded tasks like web browsing or legacy office macros will rely almost entirely on the 5.40 GHz single-core ceiling. The 3583-point R23 single-core result is competitive with desktop parts, meaning the i9-13905H does not sacrifice snappy responsiveness for its multi-core muscle. The data suggests a processor that feels fast everywhere, but truly excels when the scheduler can engage all 20 threads.

Platform and Compatibility

The i9-13905H uses the Intel BGA 1744 socket, a soldered mobile package, which means it is not upgradeable in the traditional sense—it comes welded to the motherboard. This is a Raptor Lake-H part, built on the Raptor Lake architecture, and it supports both DDR4 and DDR5 memory in a dual-channel configuration. Notably, it does not support ECC memory, which rules it out for error-correcting workstation builds but is typical for high-end consumer mobile chips.

PCIe support is Gen 5 with 8 lanes available from the CPU only. This is a critical detail for storage and GPU bandwidth: 8 Gen 5 lanes provide ample throughput for a single high-end discrete GPU or a pair of fast NVMe drives, but it is half the lane count of some desktop platforms. For a mobile part, this is a reasonable allocation, as most laptops pair it with one GPU and one or two SSDs. The integrated graphics are Iris Xe Graphics with 96 execution units, which is a competent iGPU for display output and light 3D work, though the presence of a discrete GPU is assumed for gaming or heavy creative tasks.

The upgrade path is effectively nil—no socket upgrades, no multiplier unlocking (the multiplier is locked), and the production status is active, meaning it is a current product. The launch MSRP is $697. The memory bus being dual-channel is a limiting factor for some memory-bandwidth-sensitive workloads, but the 24 MB of shared L3 cache (with 2 MB L2 per core and 80 KB L1 per core) helps mitigate this by keeping frequently accessed data closer to the cores. The die size is 257 mm², which is large for a mobile chip, but that is the cost of 14 cores and 96 EU graphics on a 10 nm node.

Benchmark Performance

The i9-13905H's average benchmark score of 40712 places it in a razor-thin competitive bracket. Against the AMD Ryzen 7 7700, it leads by 0.6 percent (40468 vs 40712). Against the AMD Ryzen 9 7940H, it leads by 0.7 percent (40431 vs 40712). Against the AMD Ryzen 9 PRO 8945HS, it trails by 0.8 percent (41031 vs 40712). And against the AMD Ryzen AI 9 365, it leads by 1.1 percent (40262 vs 40712). These deltas are within noise—single-digit percentage points—meaning the i9-13905H is effectively tied with these four rivals in aggregate performance.

But the aggregate hides the per-test story. In Cinebench R15, the multi-core score is 2558 and single-core is 360, giving a 7.1-to-1 ratio. In R20, multi-core is 10659 and single-core is 1504, a 7.1-to-1 ratio as well. The R23 scores (25380 multi, 3583 single) show the same 7.1-to-1 scaling, indicating that the Cinebench workloads scale consistently with thread count. The Passmark multithread score of 29779 is roughly 8 times the single-thread score of 3703, which is a slightly better scaling ratio, suggesting that Passmark's multi-threaded tests are more effective at keeping all 20 threads busy.

The data encryption score of 20464 is notable—it is lower than the integer math score of 101716 by a factor of 5, which reflects the cryptographic workload's dependency on specific instruction paths rather than raw parallel throughput. The extended instructions score of 21297 is also modest compared to floating-point math (73434), indicating that AVX-512 or similar extended instruction sets are not the primary strength here. The find prime numbers score of 122 is the weakest result in the suite, a classic indicator of a workload that is heavily memory-latency bound and does not scale with core count. In contrast, data compression (354330) is the highest absolute score, showing that the 24 MB L3 cache and 20 threads are excellent for compression algorithms that exhibit high data reuse.

The percentile ranking of 91 means this chip outperforms 91 percent of all CPUs in the database, but the nearestRivals data shows that the top tier is crowded—the spread from the slowest rival (40262) to the fastest (41031) is only 1.9 percent. The i9-13905H sits in the middle of that pack, neither a clear winner nor a loser, but a consistent performer across the board.

Who Should Consider It

For content creators running Cinebench-class rendering workloads, the R23 multi-core score of 25380 is the headline figure—it is roughly 7 times the single-core score, meaning that video exports or 3D renders will complete in a fraction of the time they would on a dual-core part. The data compression score of 354330 also makes it a strong fit for file archival, backup tools, or database operations that involve heavy zipping or deduplication.

For gamers, the single-thread performance is the more relevant metric. The R23 single-core score of 3583 and Passmark single-thread score of 3703 are both strong, ensuring that frame rates in CPU-bound titles will not be bottlenecked by the processor. The 8 Gen 5 PCIe lanes are sufficient for a high-end GPU, and the 96 EU Iris Xe graphics can handle light gaming on its own, though a discrete GPU is recommended for modern titles. The find prime numbers score of 122 is a cautionary note for any workload that is pure pointer-chasing or latency-bound—this is not a chip that excels at that class of problem.

For office productivity and general desktop use, the 2.60 GHz base clock and 5.40 GHz boost clock ensure that the chip idles efficiently but snaps to high frequency when needed. The Passmark integer math score of 101716 and floating-point score of 73434 indicate strong throughput for spreadsheet recalculation, data analysis, and financial modeling. The random string sorting score of 37460 is adequate but not exceptional, meaning that heavy text processing or log parsing may not see the full benefit of the 20 threads.

The data suggests three distinct user profiles: (1) mobile workstation users who need maximum multi-threaded throughput in a laptop form factor, (2) gamers who want a CPU that can keep up with a powerful GPU without being the bottleneck, and (3) power users who run a mix of single-threaded and multi-threaded applications and want no weak spots. The 91st percentile ranking confirms that this is a top-tier mobile chip, but the tight rival deltas mean that brand loyalty or platform-specific features (e.g., Thunderbolt, connectivity) may be the deciding factor rather than raw performance.

Power and Thermals

The i9-13905H carries a TDP of 45 watts, which places it in the high-performance mobile class—this is not an ultra-low-power chip for thin-and-light laptops. A 45 W TDP implies that a laptop equipped with this processor needs a robust cooling solution: a capable air cooler with multiple heat pipes and a decent fan, or a vapor chamber in premium designs. The 10 nm process node and 257 mm² die size mean that the chip does generate significant heat under sustained multi-threaded loads, as evidenced by the multi-core scores that push all 20 threads to their limits.

The TDP class dictates the laptop form factor. Expect chassis that are at least 20 mm thick, with substantial thermal vents and possibly liquid-metal thermal compound in the most aggressive designs. The boost clock of 5.40 GHz is only sustainable for short bursts on a single core; sustained all-core loads will cause the clock to drop as power and thermal limits are enforced. The 45 W base TDP can be configured up or down by the laptop manufacturer, meaning some designs may run it at a higher power limit for better performance, while others may cap it lower to fit in a thinner chassis—that is a vendor decision, not a chip limitation.

The integrated Iris Xe Graphics with 96 EU adds another thermal consideration, though for gaming or GPU-accelerated tasks, the discrete GPU will be the primary heat source. The lack of ECC memory support and the locked multiplier are minor drawbacks for enthusiasts, but they are typical for mobile parts. The data shows that the i9-13905H is a high-performance part that demands a high-performance laptop—the 45 W TDP is the price of admission for that 25380-point multi-core Cinebench R23 score. Users coming from lower-TDP parts should be prepared for a louder fan curve and a warmer chassis under load, but the benchmark results justify the thermal overhead.