AMD Ryzen Threadripper PRO 3995WX vs AMD Ryzen Threadripper PRO 9965WX Comparison

Head-to-Head Benchmarks

The benchmark data presents a clear split: the newer AMD Ryzen Threadripper PRO 9965WX dominates the Cinebench suite and single-threaded workloads, while the older AMD Ryzen Threadripper PRO 3995WX retains decisive advantages in several heavy parallel workloads. Across 17 head-to-head comparisons, the 9965WX wins 12, the 3995WX wins 5.

The most striking margin belongs to single-thread performance. In PassMark single-thread testing, the 9965WX scores 4,551 against 2,595 for the 3995WX, a 43% advantage. This gap reflects the newer architecture's higher boost clock and per-core efficiency, not merely a clock difference. Cinebench single-core results reinforce the pattern: R23 single-core shows 11,431 versus 10,071 (11.9% ahead), and R20 single-core shows 4,801 versus 4,229 (11.9% ahead).

The multicore Cinebench results also favor the 9965WX despite its 24 cores versus 64. In Cinebench R23 multicore, the 9965WX reaches 80,976 versus 71,338, an 11.9% lead. The R20 multicore test shows 34,009 versus 29,961, again an 11.9% margin. R15 multicore follows the same pattern: 8,162 versus 7,190, an 11.9% edge. These consistent deltas across all three Cinebench generations indicate a fundamental per-core throughput advantage that overcomes the 40-core deficit.

The 3995WX strikes back in specific PassMark workloads. Data encryption is its largest win: 124,433 versus 66,155, a massive 88.1% advantage. Integer math also goes decisively to the older chip: 490,116 versus 349,195, a 40.4% lead. Data compression shows a 36.9% margin (1,841,292 versus 1,345,230), and random string sorting delivers a 26% advantage (188,574 versus 149,617). Floating-point math favors the 3995WX by 20.9% (277,715 versus 229,685).

The 9965WX wins the remaining PassMark tests. Physics shows 7,529 versus 5,498, a 27% lead. Prime number finding scores 752 versus 566, a 24.7% edge. Extended instructions are nearly even, with the 9965WX ahead by only 2.1% (108,753 versus 106,423). PassMark multithread also goes to the newer chip: 92,604 versus 83,928, a 9.4% margin.

Where Each One Wins

The data paints a clear picture of workload specialization. The 3995WX is the choice for encryption-heavy tasks, large integer calculations, and data compression pipelines. Its 88.1% encryption advantage and 40.4% integer math lead suggest workloads that scale with raw core count and memory-level parallelism benefit substantially from the 64-core configuration. The 26% random string sorting win and 20.9% floating-point math win further indicate strengths in sorting algorithms, scientific computations, and any workload that can saturate many cores with independent operations.

The 9965WX wins in rendering, simulation, and general productivity. Its consistent 11.9% lead across all Cinebench multicore versions is particularly telling for 3D rendering and animation workloads, where Cinebench is a direct proxy. The 43% single-thread advantage makes it superior for application responsiveness, code compilation with serial segments, and any software that is not perfectly threaded. The 27% physics win suggests better performance in simulation and physics engines, while the 24.7% prime number advantage points to stronger integer throughput per clock.

For mixed workloads, the 9965WX's PassMark multithread score (92,604 versus 83,928) indicates it handles diverse parallel tasks more efficiently overall, despite the 3995WX winning specific narrow tests. The extended instructions result (nearly tied at 2.1% difference) shows both processors are competent for AVX and similar vectorized workloads, but the newer chip edges ahead.

Architecture Differences

The two processors belong to different eras of AMD's Threadripper PRO lineup. The 3995WX is a 3000 series part, built on the Zen 2 architecture with the codename Castle Peak. It uses a 7 nm process at TSMC and packs 64 cores and 128 threads. The 9965WX is a 9000 series part, built on Zen 5 with the codename Shimada Peak, using a 4 nm process. It has 24 cores and 48 threads.

The transistor counts are surprisingly close: the 3995WX has 30,400 million transistors across 8 dies of 74 mm² each, while the 9965WX has 33,260 million across 4 dies of 70.6 mm² each. This means the newer chip concentrates more transistors on fewer, slightly smaller dies, enabling far higher per-core performance.

Cache configurations differ substantially. Both have 64 KB L1 per core, but the 3995WX has 512 KB L2 per core versus 1 MB L2 per core on the 9965WX. The L3 cache is inverted: the 3995WX has 256 MB total, while the 9965WX has 128 MB. The larger L3 on the older chip helps explain its wins in data compression and encryption, where large working sets benefit from more cache capacity per thread. The newer chip's larger L2 per core supports its single-thread dominance.

Memory architecture diverges by generation. The 3995WX supports DDR4 with a memory bandwidth of 204.8 GB/s. The 9965WX supports DDR5 with 409.6 GB/s, exactly double the bandwidth. Both use an eight-channel memory bus, and both support ECC memory. The doubled bandwidth on the 9965WX likely contributes to its physics and multithread wins, though the 3995WX's core count compensates in some memory-intensive tests.

PCIe connectivity also advances. The 3995WX offers PCIe Gen 4 with 128 CPU lanes. The 9965WX offers PCIe Gen 5 with the same 128 lanes. This doubles the per-lane bandwidth for GPUs and NVMe storage on the newer platform.

The sockets differ: the 3995WX uses AMD Socket WRX8, while the 9965WX uses AMD Socket sTR5. The 9965WX has an unlocked multiplier, while the 3995WX does not. The 9965WX's base clock of 4.20 GHz and boost clock of 5.40 GHz dwarf the 3995WX's 2.70 GHz base and 4.20 GHz boost. The 9965WX carries a 350 W TDP versus 280 W for the 3995WX, reflecting the higher clocks. The 9965WX is classified as Server/Workstation in the database, while the 3995WX is marked Desktop.

The Verdict

The data supports a clear verdict for most users: the AMD Ryzen Threadripper PRO 9965WX is the superior processor for general workstation use. It wins the majority of benchmarks, including every Cinebench test and the PassMark multithread score. Its 43% single-thread advantage and 11.9% multicore Cinebench lead mean it will feel faster in everyday applications and render scenes faster, despite having 40 fewer cores.

The 3995WX remains the correct choice only for narrowly defined workloads that match its specific strengths. If the primary application is data encryption, integer-heavy calculations, or large-scale data compression, the 3995WX's 64 cores deliver 88.1%, 40.4%, and 36.9% advantages respectively. These are substantial margins that no amount of per-core efficiency can overcome in those specific tasks.

For mixed workloads, the 9965WX's 9.4% PassMark multithread lead indicates it is the better all-rounder. The 3995WX's wins are concentrated in a few narrow tests, while the 9965WX wins across rendering, physics, prime number calculations, and single-threaded responsiveness. The newer chip also offers double the memory bandwidth (409.6 GB/s versus 204.8 GB/s) and PCIe Gen 5 connectivity, which future-proofs systems for faster storage and GPUs.

Users upgrading from a 3995WX should note the socket change from WRX8 to sTR5, requiring a new motherboard. The 9965WX's unlock multiplier adds overclocking flexibility absent from the older part. The 3995WX's larger 256 MB L3 cache remains its hidden asset for certain data-heavy workloads, but the benchmark data shows this advantage only manifests in specific tests.

The verdict is straightforward: choose the 9965WX unless the workload is dominated by encryption, integer math, or data compression, in which case the 3995WX's core count wins decisively.

FAQ

Q: Which processor has more cores and threads?

A: The AMD Ryzen Threadripper PRO 3995WX has 64 cores and 128 threads, while the AMD Ryzen Threadripper PRO 9965WX has 24 cores and 48 threads.

Q: Why does the 24-core 9965WX beat the 64-core 3995WX in Cinebench multicore?

A: The 9965WX leads by 11.9% in all Cinebench multicore tests due to its Zen 5 architecture, 4 nm process, and much higher clocks (4.20 GHz base, 5.40 GHz boost versus 2.70 GHz base, 4.20 GHz boost). Per-core throughput overcomes the core count deficit.

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

A: The largest margin is in PassMark data encryption, where the 3995WX scores 124,433 versus 66,155, an 88.1% advantage. The second largest is PassMark single-thread, where the 9965WX leads by 43% (4,551 versus 2,595).

Q: Which processor supports faster memory?

A: The 9965WX supports DDR5 with 409.6 GB/s bandwidth, while the 3995WX supports DDR4 with 204.8 GB/s. Both use eight-channel memory buses and support ECC.

Q: Do these processors use the same motherboard socket?

A: No. The 3995WX uses AMD Socket WRX8, while the 9965WX uses AMD Socket sTR5. They are not interchangeable.

Q: Which chip has a larger L3 cache?

A: The 3995WX has 256 MB L3 cache, double the 128 MB on the 9965WX. The 9965WX has a larger L2 cache per core (1 MB versus 512 KB).

Specification Differences

| Specification | AMD Ryzen Threadripper PRO 3995WX | AMD Ryzen Threadripper PRO 9965WX |

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

| Series | 3000 series | 9000 series |

| Cores | 64 | 24 |

| Threads | 128 | 48 |

| Base Clock | 2.70 GHz | 4.20 GHz |

| Boost Clock | 4.20 GHz | 5.40 GHz |

| TDP | 280 W | 350 W |

| Socket | AMD Socket WRX8 | AMD Socket sTR5 |

| Architecture | Zen 2 | Zen 5 |

| Codename | Castle Peak | Shimada Peak |

| Process Node | 7 nm | 4 nm |

| Transistors | 30,400 million | 33,260 million |

| Die Size | 8x 74 mm² | 4x 70.6 mm² |

| L2 Cache | 512 KB (per core) | 1 MB (per core) |

| L3 Cache | 256 MB | 128 MB |

| Memory Support | DDR4 | DDR5 |

| Memory Bandwidth | 204.8 GB/s | 409.6 GB/s |

| PCIe | Gen 4, 128 Lanes | Gen 5, 128 Lanes |

| Integrated Graphics | None | N/A |

| Market Segment | Desktop | Server/Workstation |

| Release Date | 2020-07-13 | 2025-07-22 |

| Launch MSRP | $5489 | $2899 |

| Multiplier Unlocked | No | Yes |