AMD Ryzen Threadripper PRO 3995WX

The AMD Ryzen Threadripper PRO 3995WX is a 64-core, 128-thread desktop processor built on the Zen 2 architecture (Castle Peak) and manufactured on TSMC's 7 nm process. Its benchmark results place it in the 99th percentile of all CPUs, with an average benchmark score of 171,748. The data shows a processor engineered for extreme multi-threaded throughput, while its single-thread performance remains modest relative to its multi-core might; this split defines its character as a workstation part rather than a general-purpose desktop chip.

Single-Thread vs Multi-Thread Behavior

The performance gap between single-thread and multi-thread workloads is the defining feature of this processor. In Cinebench R23, the multi-core score is 71,338, while the single-core score is 10,071 — a ratio of roughly 7:1. In Geekbench, the multi-core score of 14,092 versus a single-core score of 1,577 reveals an even starker divide, approximately 9:1. This indicates that the processor's massive 64-core, 128-thread configuration delivers exceptional scaling in parallel tasks, but each individual core runs at a modest 2.70 GHz base and 4.20 GHz boost clock.

The PassMark suite reinforces this split. The multi-thread score of 83,928 dwarfs the single-thread score of 2,595, while integer math reaches 490,116 and floating-point math hits 277,715. Data compression scores 1,841,292, and data encryption scores 124,433 — all indicating that workloads that can be broken into many parallel threads will see enormous throughput. Conversely, the single-thread score of 2,595 places it far behind typical high-frequency desktop chips, meaning latency-sensitive, lightly-threaded tasks will not benefit from this architecture.

Real-world implications are clear: tasks like video rendering, 3D scene simulation, scientific computing, and large-scale data processing — where the 128 threads can all be utilized — will see near-linear scaling. However, tasks like web browsing, office document editing, or legacy single-threaded applications will run at a level comparable to a mid-range processor from its era, since the boost clock of 4.20 GHz is not exceptional by modern standards. The 256 MB L3 cache (with 64 KB L1 and 512 KB L2 per core) helps mitigate some latency, but the fundamental design prioritizes core count over per-core speed.

Who Should Consider It

Workloads that demand maximum parallel compute are the primary audience. The Cinebench R20 multi-core score of 29,961 and R15 multi-core score of 7,190 demonstrate that rendering engines and physics simulations that scale across 128 threads will achieve results unattainable on lower-core-count parts. The PassMark find prime numbers score of 566, which appears low, is actually a measure of single-threaded integer performance in that specific test — indicating that this chip is not suited for workloads that are inherently sequential.

For content creation, the data supports this as a top-tier choice for 3D rendering, video encoding, and complex compositing. The floating-point math score of 277,715 and extended instructions score of 106,423 suggest strong performance in scientific simulations and cryptographic workloads. The data encryption score of 124,433 further points to suitability for server-like or heavy data-processing tasks.

Conversely, gamers should avoid this processor. The single-thread score of 2,595 in PassMark and 1,577 in Geekbench indicate that most game engines, which rely heavily on one or two primary threads, will be bottlenecked. Office productivity suites that are not heavily multi-threaded will also see no benefit from the 64-core design. The processor's market segment is explicitly "Desktop," but its behavior is that of a server-class workstation chip — it belongs in systems used for rendering farms, data analysis, and heavy compilation workloads, not interactive or latency-sensitive applications.

Benchmark Performance

The average benchmark score of 171,748 places it in the 99th percentile of all CPUs, confirming its top-tier status. Comparing against its nearest rivals, the data shows a tightly contested field. It outperforms the AMD EPYC 7C13 (average score 167,788) by 2.4%, a modest but consistent lead. Against the AMD EPYC 9375F (average score 162,497), it is 5.7% ahead, and versus the AMD EPYC 7663 (average score 161,973), it leads by 6.0%. These deltas indicate that while the Threadripper PRO 3995WX is not the absolute fastest in every metric, it consistently edges out these server-class competitors in aggregate performance.

The only rival it trails is the AMD EPYC 7763 (average score 179,916), which is 4.5% ahead. This suggests that in highly optimized multi-threaded server workloads, the EPYC 7763 has a slight advantage, likely due to architectural refinements. However, the Threadripper's 128 PCIe Gen 4 lanes (CPU only) and eight-channel DDR4 memory support with 204.8 GB/s bandwidth provide a platform advantage for workstation use cases that require extensive I/O and memory throughput.

The Cinebench R23 multi-core score of 71,338 is a standout figure, indicating that for pure render workloads, this chip delivers performance near the top of the spectrum. The PassMark random string sorting score of 188,574 demonstrates strong data-manipulation capabilities, while the 490,116 integer math score shows robust general-purpose processing. The single-thread scores across all benchmarks (Cinebench R23: 10,071; Geekbench: 1,577; PassMark: 2,595) are consistent in showing that per-core performance is not this chip's strength, but the aggregate multi-core results more than compensate in parallel environments.

Platform and Compatibility

The processor uses AMD Socket WRX8, a platform designed specifically for the Ryzen Threadripper PRO series. It supports DDR4 memory with an eight-channel memory bus, delivering a memory bandwidth of 204.8 GB/s — a critical specification for feeding 64 cores with data. ECC memory is supported, which is essential for long-running compute tasks and server-like reliability.

PCIe connectivity is Gen 4 with 128 lanes (CPU only), providing exceptional expansion capability for multiple GPUs, NVMe storage arrays, and high-speed networking cards. This makes it suitable for systems that require massive parallel I/O, such as machine learning workstations or multi-GPU render nodes. The processor is not multiplier unlocked, meaning overclocking is not supported; users must rely on the stock 2.70 GHz base and 4.20 GHz boost clocks.

The production status is "Active," and the release date is July 13, 2020. The architecture is Zen 2 (Castle Peak), which is a generation behind more recent releases, but the 7 nm process and 30,400 million transistors indicate a mature, well-optimized design. The die size is 8x 74 mm², showing a multi-chiplet design. The platform's upgrade path is limited to other WRX8-compatible processors, so users should consider this when planning long-term system longevity.

How It Compares

vs. AMD EPYC 7C13: The Threadripper PRO 3995WX holds a 2.4% lead in average benchmark score (171,748 vs. 167,788). This indicates that for workstation workloads, the Threadripper's higher memory bandwidth and dedicated desktop platform provide a slight edge over this EPYC part, despite both having similar core-count capabilities.

vs. AMD EPYC 7763: The EPYC 7763 is 4.5% ahead (179,916 vs. 171,748). The data shows that this EPYC processor offers superior aggregate performance, likely due to architectural improvements. However, the Threadripper's advantage lies in its platform — eight-channel memory and 128 PCIe Gen 4 lanes — which may be more relevant for desktop workstation use than raw compute.

vs. AMD EPYC 9375F: The Threadripper PRO 3995WX is 5.7% ahead (171,748 vs. 162,497). This is a clear win in average benchmark terms, suggesting that for general multi-threaded workloads, the Threadripper delivers higher throughput. The delta is significant enough to recommend the Threadripper over this EPYC model for creation and analysis tasks.

vs. AMD EPYC 7663: The Threadripper PRO 3995WX leads by 6.0% (171,748 vs. 161,973). This is the largest margin among its nearest rivals, confirming that the Threadripper's combination of high core count and desktop-oriented memory subsystem outperforms this EPYC variant in aggregate benchmarks.

FAQ

Q: Does this processor support ECC memory?

A: Yes, ECC memory is supported, which is important for error-free long-running compute workloads.

Q: What is the memory bandwidth and channel configuration?

A: It supports eight-channel DDR4 memory with a bandwidth of 204.8 GB/s.

Q: Is this processor unlocked for overclocking?

A: No, the multiplier is not unlocked, so overclocking is not supported.

Q: What socket does it use?

A: It uses AMD Socket WRX8, which is specific to the Ryzen Threadripper PRO series.

Q: How many PCIe lanes does it provide?

A: It provides 128 PCIe Gen 4 lanes from the CPU only, enabling extensive multi-GPU and NVMe configurations.

Q: What is the launch MSRP?

A: The launch MSRP is $5489.

Power and Thermals

The TDP is 280 watts, which places it in the high-power class of desktop processors. This TDP figure, combined with the 64-core design on a 7 nm process, implies that a robust cooling solution is required. The data does not specify a cooler; however, the 280 W TDP indicates that a high-end air cooler or a substantial liquid cooling solution would be necessary to maintain sustained multi-core performance without thermal throttling.

The power draw is consistent with the processor's performance class — it is not an energy-efficient part for light workloads, but in full multi-threaded operation, the 280 W TDP is justified by the output. The eight-chiplet die design (8x 74 mm²) spreads heat generation across the package, which can aid in thermal management if the cooler is designed for even heat distribution. Users should plan for adequate case airflow and a cooler rated for 280 W or higher to achieve the benchmark scores shown in the data. In idle or single-threaded tasks, power consumption will be lower, but the platform's memory and PCIe controllers still contribute to baseline power draw.