AMD EPYC 9965 vs AMD Ryzen Threadripper PRO 9985WX Comparison

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9965 has 192 cores and 384 threads, while the AMD Ryzen Threadripper PRO 9985WX has 64 cores and 128 threads. The EPYC offers triple the core and thread count.

Q: How do the two processors compare in single-threaded performance?

A: The Ryzen Threadripper PRO 9985WX wins in PassMark single-thread testing with a score of 4482, which is 29.1% ahead of the EPYC 9965's score of 3176. However, in Cinebench R23 single-core, the EPYC 9965 scores 19293 compared to 18756 for the Threadripper, a 2.9% advantage.

Q: What is the memory configuration difference?

A: The EPYC 9965 uses a twelve-channel memory bus with 576.0 GB/s bandwidth, while the Threadripper PRO 9985WX uses an eight-channel bus with 409.6 GB/s bandwidth. Both support DDR5 memory and ECC.

Q: Which CPU has a higher boost clock?

A: The Ryzen Threadripper PRO 9985WX has a boost clock of 5.40 GHz, significantly higher than the EPYC 9965's 3.70 GHz boost clock. The Threadripper also has a higher base clock at 3.20 GHz versus 2.25 GHz.

Q: What is the difference in the manufacturing process?

A: The EPYC 9965 is built on a 3 nm process node, while the Threadripper PRO 9985WX uses a 4 nm node. Both are manufactured by TSMC.

Q: In how many benchmark categories does each processor win?

A: The EPYC 9965 wins 15 of the 17 head-to-head benchmarks. The Threadripper PRO 9985WX wins only 2, both being the PassMark single-thread tests.

Architecture Differences

The AMD EPYC 9965 and AMD Ryzen Threadripper PRO 9985WX both utilize the Zen 5 architecture, but they diverge significantly in implementation. The EPYC 9965 carries the codename Turin and belongs to the EPYC 9005 series, built on a 3 nm TSMC process. The Threadripper PRO 9985WX, codenamed Shimada Peak, uses a 4 nm TSMC process and is part of the 9000 series. This process difference contributes to their distinct power and density profiles.

The EPYC 9965 is a dense, high-core-count part with 192 cores and 384 threads. Its cache hierarchy includes 80 KB of L1 per core, 1 MB of L2 per core, and a massive 384 MB of shared L3 cache. The Threadripper PRO 9985WX offers 64 cores and 128 threads with 64 KB of L1 per core, 1 MB of L2 per core, and 256 MB of L3 cache. The EPYC's L3 cache is 128 MB larger, which benefits workloads that repeatedly access large datasets.

The EPYC 9965 uses AMD Socket SP5, while the Threadripper PRO 9985WX uses AMD Socket sTR5. The EPYC has a TDP of 500, and the Threadripper has a TDP of 350. Both are listed as active production parts with no integrated graphics. The EPYC is not multiplier-unlocked, whereas the Threadripper PRO 9985WX has an unlocked multiplier for overclocking.

Transistor and die data are recorded for the Threadripper PRO 9985WX: 66,520 million transistors across an 8x 70.6 mm² die arrangement. No equivalent data is listed for the EPYC 9965. The EPYC remains a server-focused design with twelve-channel memory, while the Threadripper uses eight-channel memory; both support DDR5 and ECC. PCIe connectivity is identical on paper: Gen 5 with 128 lanes (CPU only) for each.

The release timeline differs as well. The EPYC 9965 launched on 2024-10-09, while the Threadripper PRO 9985WX launched on 2025-07-22. The EPYC's launch MSRP is $14813, and the Threadripper's launch MSRP is $7999.

The Verdict

The recorded data points to a clear split: the AMD EPYC 9965 is the choice for heavily threaded throughput, and the AMD Ryzen Threadripper PRO 9985WX is the choice for single-thread responsiveness and overclocking flexibility.

The EPYC 9965 dominates the head-to-head benchmark list, winning 15 of 17 tests. Its biggest margins come in PassMark integer math (120.7% ahead), data encryption (125.1% ahead), and floating-point math (108.4% ahead). In Cinebench R23 multicore, the EPYC scores 136661 versus 132859 for the Threadripper, a 2.9% edge. For workloads that scale across many cores, such as rendering, scientific computing, or large-scale virtualization, the EPYC's 192 cores and 384 threads provide a substantial throughput advantage.

The Threadripper PRO 9985WX counters with a 29.1% lead in PassMark single-thread performance, scoring 4482 versus 3176. Its higher boost clock of 5.40 GHz and base clock of 3.20 GHz make it better suited for lightly threaded tasks, interactive workloads, and applications where per-core speed matters more than raw core count. Its unlocked multiplier also allows tuning beyond stock settings, which the EPYC does not permit.

The EPYC also posts a higher average benchmark score of 620487, compared to 320749 for the Threadripper. The EPYC sits at the 100th percentile among all CPUs in the database, while the Threadripper is at the 99th. The EPYC's nearest rivals include other EPYC parts, with a 50.6% delta over the Ryzen Threadripper PRO 9995WX. The Threadripper's nearest rival is the Ryzen Threadripper 9980X, with a nearly identical average score.

For a single-socket workstation where per-core speed and overclocking matter, the Threadripper PRO 9985WX is the practical pick. For a server or compute node where maximum parallel throughput is the goal, the EPYC 9965 is the clear winner. The data does not support one CPU being universally better; it supports matching the hardware to the workload.

Specification Differences

The two processors differ across nearly every core specification. The EPYC 9965 has 192 cores and 384 threads, while the Threadripper PRO 9985WX has 64 cores and 128 threads. Base clocks are 2.25 GHz for the EPYC and 3.20 GHz for the Threadripper. Boost clocks are 3.70 GHz and 5.40 GHz, respectively.

TDP ratings differ: the EPYC is rated at 500, and the Threadripper at 350. The EPYC uses AMD Socket SP5, while the Threadripper uses AMD Socket sTR5. Process nodes differ, with the EPYC on 3 nm and the Threadripper on 4 nm, both from TSMC. The Threadripper has recorded transistor and die size data (66,520 million transistors, 8x 70.6 mm²), while the EPYC has none listed.

Cache configurations differ in L1 and L3. The EPYC has 80 KB of L1 per core and 384 MB of shared L3. The Threadripper has 64 KB of L1 per core and 256 MB of L3. Both have 1 MB of L2 per core.

Memory channels and bandwidth differ: the EPYC uses twelve-channel memory with 576.0 GB/s bandwidth, and the Threadripper uses eight-channel memory with 409.6 GB/s bandwidth. Both support DDR5 and ECC.

The EPYC is not multiplier-unlocked; the Threadripper is. The EPYC launched on 2024-10-09 with a launch MSRP of $14813. The Threadripper launched on 2025-07-22 with a launch MSRP of $7999. Both have 128 PCIe Gen 5 lanes (CPU only) and no integrated graphics. The EPYC's part number is 100-000000976, and the Threadripper's is 100-000000722.

Head-to-Head Benchmarks

The head-to-head results show the EPYC 9965 winning 15 of 17 tests, with the Threadripper PRO 9985WX winning the remaining 2. The margins vary widely across test categories.

In Cinebench tests, the EPYC leads consistently by 2.9% across all six runs: R15 multicore (13775 vs 13392), R15 single-core (1944 vs 1890), R20 multicore (57397 vs 55800), R20 single-core (8103 vs 7877), R23 multicore (136661 vs 132859), and R23 single-core (19293 vs 18756). These are modest but consistent wins.

The PassMark suite reveals where the EPYC's core advantage becomes decisive. In data compression, the EPYC scores 5679990 versus 2912972, a 95% lead. Data encryption shows a 125.1% lead (348449 vs 154824). Extended instructions favor the EPYC by 70.1% (383298 vs 225340). Floating-point math sees a 108.4% lead (1153453 vs 553348). Integer math is 120.7% ahead (1926069 vs 872710). Random string sorting is 92.4% ahead (633030 vs 329014). Physics tests show a 35.7% lead (18707 vs 13783). Multithread performance is 7% ahead (160542 vs 150071). Find prime numbers shows a smaller 6.2% lead (1208 vs 1138).

The Threadripper PRO 9985WX wins the two PassMark single-thread tests by 29.1% (4482 vs 3176). This is its only area of dominance, but it is a meaningful one for single-threaded application responsiveness.

The biggest wins for the EPYC are in encryption and integer math, both exceeding 120% leads. The smallest EPYC win is in find prime numbers at 6.2%. The Threadripper's win is confined to single-thread performance. The benchmark data shows that the EPYC's advantage grows in workloads that scale across many cores and threads.

Where Each One Wins

The AMD EPYC 9965 wins in every heavily threaded category in the head-to-head data. Its 192 cores and 384 threads translate into large leads in integer math (120.7%), floating-point math (108.4%), data encryption (125.1%), and data compression (95%). These results indicate strong performance for rendering, video encoding, scientific simulation, database workloads, and large-scale data processing. The EPYC also has a 128 MB larger L3 cache and 166.4 GB/s more memory bandwidth, which supports workloads with large working sets.

The EPYC's Cinebench wins, though smaller at 2.9%, still put it ahead in all render-loop tests, from R15 through R23, in both single-core and multicore modes. Its 100th percentile ranking among all CPUs reinforces its position as the top overall performer in the database.

The AMD Ryzen Threadripper PRO 9985WX wins in PassMark single-thread performance with a 29.1% lead (4482 vs 3176). Its higher base and boost clocks (3.20 GHz and 5.40 GHz) make it better suited for interactive applications, legacy software that relies on single-thread speed, and tasks where per-core latency matters more than aggregate throughput. The unlocked multiplier also gives it flexibility for overclocking, which the EPYC lacks.

The Threadripper's 99th percentile ranking and average benchmark score of 320749 place it just below the EPYC's 620487 average. Its nearest rival, the Ryzen Threadripper 9980X, scores nearly identically, indicating that the Threadripper PRO 9985WX is at the top of its own segment but not in the same throughput class as the EPYC 9965.

For a builder assembling a single-socket workstation with an emphasis on frequency, the Threadripper PRO 9985WX is the appropriate choice. For a server or compute node where maximum parallel throughput is the priority, the EPYC 9965 is the stronger option. The data does not show one CPU as universally superior; it shows two different design goals, each optimized for a different workload profile.