AMD Ryzen Threadripper PRO 9965WX

The AMD Ryzen Threadripper PRO 9965WX is a 24-core, 48-thread workstation processor built on the Zen 5 architecture (codenamed Shimada Peak) and manufactured on TSMC's 4 nm process node. It sits in the 98th percentile of all CPUs benchmarked, with an average benchmark score of 146814, placing it in direct competition with server-class EPYC parts. Its launch MSRP is $2899.

Benchmark Performance

The aggregate benchmark data positions the 9965WX in a tight cluster with several high-end EPYC processors. Its average score of 146814 is virtually identical to the AMD EPYC 8434P, which scores 146881 — a delta of 0%. This means the two processors deliver statistically equivalent overall performance despite their different platform orientations. The 9965WX edges ahead of the AMD EPYC 7643P by 1.4% (144824), a modest but consistent margin that appears across multi-threaded workloads.

The picture shifts more dramatically against other rivals. The 9965WX holds a 10.2% advantage over the AMD EPYC 9275F (133174), a meaningful gap that suggests the Threadripper's larger core count and cache configuration pay off in sustained throughput. However, the AMD EPYC 9355P (160853) is 8.7% faster than the 9965WX, indicating that AMD's newer EPYC generation offers a measurable performance headroom at the top of the stack.

Looking at specific Cinebench runs, the multi-core scores are particularly telling. The R23 multi-core result of 80976 demonstrates exceptional parallel scaling, while the R20 multi-core score of 34009 and R15 multi-core score of 8162 follow the expected progression. These numbers, combined with the PassMark multi-thread score of 95346, paint a picture of a processor that excels in heavily threaded rendering and simulation tasks. The data encryption score of 71873 and extended instructions score of 102589 further confirm strong computational throughput in specialized workloads.

Platform and Compatibility

The 9965WX uses AMD Socket sTR5, a platform designed for professional workstation and server deployments. Memory support is DDR5 with an eight-channel memory bus, delivering 409.6 GB/s of peak bandwidth. This is a critical specification for memory-hungry workloads like large-scale data analysis or complex 3D scenes, where the eight-channel configuration prevents bandwidth from becoming a bottleneck. ECC memory is supported, which is essential for error-critical professional work.

PCIe connectivity is Gen 5 with 128 lanes available from the CPU alone. This generous lane allocation allows for multiple high-speed GPUs, NVMe storage arrays, and network interface cards to be installed simultaneously without requiring a switch chip. The upgrade path is inherently tied to the sTR5 platform, which is designed for long-life workstation cycles. The processor has an unlocked multiplier, enabling overclocking for users who need to extract extra performance. The transistor count is 33,260 million, spread across four chiplets with a die size of 4x 70.6 mm².

Single-Thread vs Multi-Thread Behavior

The benchmark data reveals a clear split between single-thread and multi-thread performance. In Cinebench R23, the single-core score is 11431 against a multi-core score of 80976 — a ratio of roughly 7.1x scaling from 24 cores. The R20 results show 4801 single-core versus 34009 multi-core, and R15 shows 1152 single-core versus 8162 multi-core. These ratios are consistent, indicating that the Zen 5 architecture scales efficiently without significant thermal or power throttling as more cores are engaged.

The PassMark single-thread score of 4555 is competitive for a workstation chip, while the multi-thread score of 95346 shows the massive uplift when all threads are active. Integer math scores 365422, floating point math scores 220036, and physics scores 11278, each demonstrating strong per-thread efficiency. The data compression score of 1314596 and random string sorting score of 164382 are particularly high, suggesting that the processor handles data movement and manipulation tasks with exceptional speed.

Real-world implications are clear: workloads that rely on single-thread responsiveness — such as interactive CAD modeling, spreadsheet recalculation, or scripted automation — will see solid but not class-leading performance. Workloads that scale across all 48 threads, such as video rendering, scientific computing, or batch photo processing, will see near-linear gains. The low single-core-to-multi-core ratio (around 7x for 24 cores) indicates that the processor does not suffer from significant frequency degradation under full load, a common issue in high-core-count parts.

How It Compares

vs AMD EPYC 8434P: This is the closest match, with a delta of just 0%. Both processors average around 146800-146900 in overall benchmark scores. The 9965WX matches the EPYC 8434P in aggregate performance, meaning users choosing between them should base decisions on platform features like memory channels or PCIe lanes rather than raw speed. The Threadripper offers the same performance class with workstation-oriented features.

vs AMD EPYC 7643P: The 9965WX is 1.4% faster than the EPYC 7643P (144824 vs 146814). This is a narrow margin that falls within typical run-to-run variance, but the Threadripper's consistent edge across Cinebench tests suggests a real architectural advantage. The 7643P is an older generation, so this gap likely reflects Zen 5's improved instructions-per-clock.

vs AMD EPYC 9275F: The 9965WX beats the EPYC 9275F by 10.2% (133174 vs 146814). This is a decisive margin, showing that the Threadripper's 24-core configuration with 128 MB of L3 cache outperforms a rival that presumably targets higher clock speeds with fewer cores. The data suggests the 9965WX delivers superior multi-threaded throughput without sacrificing too much single-thread performance.

vs AMD EPYC 9355P: The EPYC 9355P is 8.7% faster (160853 vs 146814), making it the only rival in this group that clearly outperforms the 9965WX. This gap indicates that AMD's newest EPYC generation holds a genuine performance lead. However, the 9355P is designed for dual-socket servers, so the Threadripper's single-socket simplicity and workstation features may still be preferable for desktop-style deployments.

Power and Thermals

The 9965WX has a TDP of 350 watts, placing it in the high-power tier of desktop workstation processors. This TDP class requires substantial cooling — a high-end air cooler with a large heatsink or a robust liquid cooling solution is necessary to maintain sustained performance without thermal throttling. The 4 nm process node from TSMC helps manage power efficiency, but 24 cores running at a 5.40 GHz boost clock will generate significant heat under full load.

The base clock of 4.20 GHz with a boost of 5.40 GHz indicates that the processor can reach high frequencies when thermals allow, but the 350 W envelope means that power delivery and cooling must be carefully planned. The eight-channel memory controller also draws power, so a motherboard with robust VRM phases is recommended. Users moving from lower-TDP platforms should expect higher system-level power draw and correspondingly larger cooling requirements.

FAQ

Q: What is the average benchmark score of the AMD Ryzen Threadripper PRO 9965WX?

A: The average benchmark score is 146814, which places it in the 98th percentile of all CPUs.

Q: How does the 9965WX compare to the AMD EPYC 9355P?

A: The EPYC 9355P is 8.7% faster, scoring 160853 compared to 146814 for the 9965WX.

Q: What memory configuration does this processor support?

A: It supports DDR5 memory with an eight-channel bus, providing 409.6 GB/s of bandwidth, and supports ECC memory.

Q: What is the multi-core performance in Cinebench R23?

A: The multi-core score in Cinebench R23 is 80976, while the single-core score is 11431.

Q: How many PCIe lanes does the CPU provide?

A: The CPU provides 128 PCIe Gen 5 lanes directly, without needing a separate switch chip.

Q: Is the multiplier unlocked for overclocking?

A: Yes, the multiplier is unlocked, allowing users to adjust clock speeds beyond the standard base and boost values.

Who Should Consider It

The 9965WX is best suited for professionals running heavily multi-threaded workloads where the 24 cores/48 threads can be fully utilized. Video editors working with 8K timelines, 3D artists rendering complex scenes, and engineers running finite element analysis will benefit from the 80976 Cinebench R23 multi-core score and the 95346 PassMark multi-thread score. The 128 MB of L3 cache and eight-channel DDR5 memory also make it strong for large dataset processing, as evidenced by the high data compression score of 1314596.

For gaming, the 4555 PassMark single-thread score is respectable but not top-tier; gamers who occasionally render or stream may find it adequate, but dedicated gaming builds would be better served by a higher-clocked consumer chip. Office productivity and light multitasking are well within its capabilities, though the 350 W TDP makes it overkill for such tasks unless paired with the workstation platform features.

Data scientists and researchers using the ECC memory support and 128 PCIe Gen 5 lanes for GPU compute clusters will find the platform compelling. The 10.2% advantage over the EPYC 9275F and the 1.4% edge over the EPYC 7643P show that it holds its own against server parts, while the sTR5 socket offers a more accessible single-socket workstation form factor. The 8.7% deficit against the EPYC 9355P means that users needing absolute maximum performance should look elsewhere, but for most workstation applications, the 9965WX delivers a balanced mix of throughput, memory bandwidth, and I/O capability.