AMD Ryzen Threadripper PRO 5975WX vs AMD Ryzen Threadripper PRO 9965WX Comparison

The AMD Ryzen Threadripper PRO 9965WX and the AMD Ryzen Threadripper PRO 5975WX represent two different generations of AMD's workstation platform, and the recorded data shows a generational shift that is unusually one-sided. The 9965WX, a 24-core Zen 5 part on the sTR5 socket, takes 15 of 17 head-to-head benchmark wins against the 32-core Zen 3 5975WX, despite having fewer cores. The older chip's advantages are confined to two integer-leaning PassMark tests, while the newer part sweeps every Cinebench render, every single-thread test, and the platform-level metrics that matter most for workstation builds. Both sit in the 97th and 98th percentiles against all CPUs in the database, so this is a comparison between two elite performers, not a lopsided mismatch between classes.

FAQ

Q: Which CPU is faster overall?

A: The 9965WX. It wins 15 of 17 recorded head-to-head benchmarks, with an average benchmark score of 147,009 versus 124,171 for the 5975WX, an advantage of roughly 18 percent on the database's aggregate metric.

Q: Can the older 32-core chip beat the newer 24-core chip anywhere?

A: Yes, in two PassMark tests. The 5975WX wins data encryption at 80,453 versus 66,155, a 17.8 percent edge, and integer math at 360,156 versus 349,195, a 3 percent edge. Its extra eight cores and sixteen threads keep it competitive in those specific integer and encryption workloads.

Q: How large is the single-thread gap?

A: Large and consistent. The 9965WX scores 11,431 in Cinebench R23 single-core against 9,038, a 26.5 percent lead, and 4,551 in PassMark single-thread against 3,323, a 37 percent lead. Boost clocks of 5.40 GHz versus 4.50 GHz tell much of that story.

Q: How do the memory platforms compare?

A: The 9965WX supports DDR5 across an eight-channel bus with 409.6 GB/s of bandwidth. The 5975WX supports DDR4 on the same eight-channel layout but with 204.8 GB/s of bandwidth, exactly half the figure of the newer part.

Q: Are both CPUs multiplier unlocked?

A: No. The 9965WX has an unlocked multiplier; the 5975WX does not, which removes manual overclocking as an option on the older chip.

Q: How do they rank against the database at large?

A: The 9965WX sits in the 98th percentile versus all recorded CPUs, and the 5975WX sits in the 97th. Both are near the top of the overall distribution.

The Verdict

The data points overwhelmingly toward the 9965WX for new workstation builds. It wins every rendering test by a uniform 26.5 percent margin, every single-thread test, and the platform fundamentals: double the memory bandwidth, PCIe Gen 5 instead of Gen 4 across the same 128 CPU lanes, and DDR5 instead of DDR4 support. Its nearest rivals in the database, the AMD EPYC 8434P, AMD Ryzen 9 PRO 9965, AMD EPYC 7643P, and Intel Xeon w9-3575X, all sit within 1.9 percent of its average score, confirming it competes at the very top of the recorded field.

The 5975WX still makes sense in a narrower set of cases. The recorded data shows it winning data encryption by 17.8 percent and integer math by 3 percent, so deployments built specifically around those workloads can still favor it. Its lower TDP of 280 watts against 350 watts may also matter for thermal planning. Its own rival cluster, including the AMD EPYC 7642, Intel Xeon 6730P, AMD EPYC 9354, and AMD EPYC 9384X, sits within a few percent of its average score, so it remains competitive within its tier. For everyone else, the newer part is the clear pick on pure performance.

Head-to-Head Benchmarks

Rendering is where the gap is most consistent. The 9965WX posts 8,162 in Cinebench R15 multi-core against 6,453, 34,009 in R20 multi-core against 26,888, and 80,976 in R23 multi-core against 64,021. Every one of those wins lands at 26.5 percent. Single-core rendering follows identically: 1,152 versus 910 in R15, 4,801 versus 3,795 in R20, and 11,431 versus 9,038 in R23, margins of 26.6, 26.5, and 26.5 percent. A 24-core chip beating a 32-core chip by more than a quarter in heavily threaded rendering is the clearest evidence of the per-clock and per-core gains in the newer generation.

The PassMark suite adds nuance. The 9965WX dominates prime number finding at 752 versus 438, a 71.7 percent win, and physics at 7,529 versus 4,360, a 72.7 percent win. Extended instructions go to the newer chip 108,753 to 82,850, up 31.3 percent, and single-thread goes to it 4,551 to 3,323, up 37 percent. Multithread favors it 92,604 to 75,319, a 22.9 percent margin, with floating point math at 229,685 versus 202,363 up 13.5 percent, random string sorting at 149,617 versus 123,527 up 21.1 percent, and data compression at 1,345,230 versus 1,293,784 up 4 percent.

The 5975WX's two wins deserve attention. Its data encryption score of 80,453 beats the 9965WX's 66,155 by 17.8 percent, the largest win for either chip in the entire recorded set outside the 9965WX's physics and primes results. Its integer math score of 360,156 versus 349,195 adds a 3 percent edge. These are the only recorded tests where the older part's additional cores translate into a victory, and they define the narrow use cases where it remains the stronger choice.

Specification Differences

The core and thread counts invert expectations: the 9965WX carries 24 cores and 48 threads, while the older 5975WX carries 32 cores and 64 threads. Clocks favor the new part at 4.20 GHz base and 5.40 GHz boost, against 3.60 GHz base and 4.50 GHz boost. TDP is higher on the 9965WX at 350 watts versus 280 watts. The sockets differ completely, AMD Socket sTR5 for the 9965WX and AMD Socket WRX8 for the 5975WX, so the two cannot share a platform.

Cache diverges at L2: the 9965WX provides 1 MB per core against 512 KB per core on the 5975WX, while both share the same 128 MB of L3 and the same 64 KB of L1 per core. Memory support moves from DDR4 to DDR5, and bandwidth doubles from 204.8 GB/s to 409.6 GB/s on the same eight-channel bus. PCIe moves from Gen 4 to Gen 5, both at 128 CPU-only lanes. The 9965WX has an unlocked multiplier, the 5975WX does not. Both support ECC memory and both remain in active production. Launch MSRP was $2899 for the 9965WX and $3299 for the 5975WX. Release dates were 2025 for the newer part and 2022 for the older one.

Architecture Differences

These are two fundamentally different Zen generations. The 9965WX is built on the Zen 5 architecture under the Shimada Peak codename, fabricated on TSMC's 4 nm process. The 5975WX uses Zen 3 under the Chagall PRO codename on TSMC's 7 nm process. The transistor count reflects the generational jump: 33,260 million on the 9965WX against 16,600 million on the 5975WX, exactly double. Die configuration also differs, with the 9965WX using four dies of 70.6 mm² each and the 5975WX using four dies of 81 mm² each, meaning the newer part packs far more circuitry into a smaller per-die area.

The doubled L2 cache per core, the doubled memory bandwidth, and the PCIe generational leap all flow from this architectural gap, and the benchmark data shows the result: a chip with 25 percent fewer cores delivering 26.5 percent more rendering performance. The database places the 9965WX in the 98th percentile of all CPUs with an average score of 147,009, ahead of rivals like the EPYC 8434P and Xeon w9-3575X by margins of 0.1 to 1.9 percent. The 5975WX, at 124,171 average and the 97th percentile, remains a strong performer in its own right, but the recorded data shows the architectural advance of Zen 5 on 4 nm fully overwhelming the core-count advantage of Zen 3 on 7 nm.