AMD EPYC 9355P vs AMD Ryzen Threadripper PRO 3995WX Comparison

The AMD Ryzen Threadripper PRO 3995WX and AMD EPYC 9355P represent two very different approaches to high-core-count computing. The Threadripper PRO 3995WX is a 2020-era Zen 2 behemoth with 64 cores, while the EPYC 9355P is a 2024 Zen 5 part with half the cores but much newer architecture. The data shows a clear split: the EPYC 9355P wins 12 of 17 head-to-head benchmarks, including every Cinebench test, while the Threadripper PRO 3995WX wins 5 benchmarks, all in PassMark’s integer, floating-point, and data-heavy workloads. The verdict is straightforward: if you need maximum multi-threaded throughput in rendering and general compute, the EPYC 9355P is the stronger choice; if your workload involves data compression, encryption, or integer math at massive scale, the 64-core Threadripper PRO retains a decisive edge.

The Verdict

The data indicates two distinct buyer profiles. For anyone running Cinebench-style rendering workloads, the AMD EPYC 9355P is the clear pick. It outperforms the Threadripper PRO 3995WX in Cinebench R15, R20, and R23 multicore tests by 13.7% across the board – a consistent margin that reflects its superior per-core efficiency. The EPYC 9355P also dominates in single-threaded performance, with a 30.7% lead in PassMark single-thread, which translates to snappier responsiveness in lightly-threaded tasks. For physics simulations, the EPYC 9355P is dramatically ahead, scoring 13515 versus 5498 – a 59.3% advantage.

Conversely, the Threadripper PRO 3995WX is the pick for workloads that scale with raw core count and older instruction set optimizations. It wins data compression by 28.8%, data encryption by 53.7%, integer math by 18.9%, and floating-point math by 8.2%. If your daily driver involves zipping/unzipping large datasets, encrypting files, or heavy integer arithmetic, the 64-core part’s extra cores more than compensate for its older architecture. The Threadripper PRO also holds a slight edge in random string sorting (6.7%). However, the EPYC 9355P wins the overall PassMark multithread score (96603 vs 83928), showing that its newer Zen 5 cores are more efficient even when outnumbered.

Architecture Differences

The two CPUs are separated by four years and two full architecture generations. The Threadripper PRO 3995WX is built on TSMC’s 7 nm process using the Zen 2 architecture (codename Castle Peak), while the EPYC 9355P uses TSMC’s 4 nm process with the Zen 5 architecture (codename Turin). The transistor counts reflect this: the Threadripper PRO has 30,400 million transistors across eight 74 mm² dies, whereas the EPYC 9355P packs 66,520 million transistors into eight 70.6 mm² dies – more than double the transistors on slightly smaller dies.

Core configuration differs significantly. The Threadripper PRO offers 64 cores and 128 threads, while the EPYC 9355P offers 32 cores and 64 threads. Cache layouts also diverge: the Threadripper PRO has 64 KB L1 and 512 KB L2 per core with a 256 MB L3 cache, while the EPYC 9355P has larger 80 KB L1 and 1 MB L2 caches per core, plus a 256 MB shared L3. Both have the same total L3 capacity, but the EPYC’s per-core L2 is double that of the Threadripper.

Memory and I/O show a generational leap. The Threadripper PRO uses DDR4 with an eight-channel memory bus and 204.8 GB/s bandwidth. The EPYC 9355P moves to DDR5 with a twelve-channel bus and 576.0 GB/s bandwidth – nearly three times the memory bandwidth. Both support ECC memory. PCIe connectivity is identical in lane count (128 lanes from the CPU), but the EPYC 9355P runs Gen 5 while the Threadripper PRO runs Gen 4. The EPYC 9355P is classified as Server/Workstation, while the Threadripper PRO is Desktop; both use different sockets (SP5 for EPYC, WRX8 for Threadripper). The EPYC 9355P has a listed integrated graphics of N/A, while the Threadripper PRO has none listed.

FAQ

Q: Which CPU has more cores?

A: The AMD Ryzen Threadripper PRO 3995WX has 64 cores and 128 threads, exactly double the EPYC 9355P’s 32 cores and 64 threads.

Q: Is the newer EPYC 9355P faster in multi-threaded Cinebench despite having fewer cores?

A: Yes. The EPYC 9355P wins Cinebench R15, R20, and R23 multicore tests, all by a 13.7% margin. For example, its R23 multicore score is 82666 versus 71338 for the Threadripper PRO.

Q: Which CPU offers better single-threaded performance?

A: The EPYC 9355P is significantly ahead. It leads by 13.8% in Cinebench R15 single-core (1176 vs 1014) and by 30.7% in PassMark single-thread (3747 vs 2595).

Q: Where does the Threadripper PRO 3995WX win?

A: It wins in PassMark data compression (1841292 vs 1429976, +28.8%), data encryption (124433 vs 80961, +53.7%), integer math (490116 vs 412067, +18.9%), floating-point math (277715 vs 256635, +8.2%), and random string sorting (188574 vs 176697, +6.7%).

Q: What are the memory differences?

A: The Threadripper PRO uses DDR4 with an eight-channel bus and 204.8 GB/s bandwidth. The EPYC 9355P uses DDR5 with a twelve-channel bus and 576.0 GB/s bandwidth, a 2.8x increase.

Q: Which CPU has a higher average benchmark score?

A: The Threadripper PRO 3995WX has a higher average benchmark score of 171748, compared to 160358 for the EPYC 9355P. Both sit at the 98th percentile of all CPUs.

Specification Differences

The table below highlights only the fields where the two CPUs differ.

| Specification | AMD Ryzen Threadripper PRO 3995WX | AMD EPYC 9355P |

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

| Cores | 64 | 32 |

| Threads | 128 | 64 |

| Base Clock | 2.70 GHz | 3.55 GHz |

| Boost Clock | 4.20 GHz | 4.40 GHz |

| Socket | AMD Socket WRX8 | AMD Socket SP5 |

| Architecture | Zen 2 | Zen 5 |

| Codename | Castle Peak | Turin |

| Generation | Ryzen Threadripper (Zen 2) | EPYC (Zen 5) |

| Process Node | 7 nm | 4 nm |

| Transistors | 30,400 million | 66,520 million |

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

| L1 Cache | 64 KB (per core) | 80 KB (per core) |

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

| L3 Cache | 256 MB | 256 MB (shared) |

| Memory Support | DDR4 | DDR5 |

| Memory Bus | Eight-channel | Twelve-channel |

| Memory Bandwidth | 204.8 GB/s | 576.0 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 | 2024-10-09 |

| Launch MSRP | $5489 | $2998 |

| Part Number | 100-000000087 | 100-000001521 |

Head-to-Head Benchmarks

The benchmark data reveals a consistent pattern: the EPYC 9355P dominates in rendering and single-threaded tasks, while the Threadripper PRO 3995WX holds sway in specific data-processing workloads. Starting with Cinebench, the EPYC 9355P wins all six tests. In Cinebench R15 multicore, it scores 8332 versus 7190 (+13.7% advantage for the EPYC). The single-core R15 shows 1176 versus 1014 (+13.8%). This pattern repeats exactly in R20 (34719 vs 29961 multicore, 4901 vs 4229 single-core) and R23 (82666 vs 71338 multicore, 11670 vs 10071 single-core). The consistent 13.7% delta across all multicore tests suggests the EPYC’s advantage scales uniformly with its 4 nm process and Zen 5 cores.

The PassMark suite tells a more nuanced story. The EPYC 9355P wins the overall multithread score (96603 vs 83928, +13.1%), but its biggest wins come in physics (13515 vs 5498, a staggering 59.3% advantage) and prime number finding (1044 vs 566, +45.8%). It also takes extended instructions (107622 vs 106423, a narrow 1.1% lead) and single-thread (3747 vs 2595, +30.7%).

The Threadripper PRO’s wins are concentrated in data-heavy tasks. Its largest margin is data encryption: 124433 vs 80961, a 53.7% lead. Data compression shows 1841292 vs 1429976 (+28.8%). Integer math favors it at 490116 vs 412067 (+18.9%), and floating-point math at 277715 vs 256635 (+8.2%). Random string sorting is close but still a win: 188574 vs 176697 (+6.7%). The overall benchmark count is 12 wins for the EPYC 9355P and 5 for the Threadripper PRO.

Where Each One Wins

The EPYC 9355P is the clear winner for rendering and simulation workloads. Its Cinebench R23 multicore score of 82666 is 13.7% ahead of the Threadripper PRO, making it the better choice for 3D rendering, video encoding, and any task that scales with multi-threaded CPU performance. The physics benchmark is its most dramatic victory – a 59.3% lead – suggesting it excels in scientific computing and physics simulation. Single-threaded performance is also decisively in its favor, with a 30.7% PassMark single-thread lead, which matters for everyday responsiveness, legacy software, and lightly-threaded applications. The newer memory subsystem (DDR5, twelve-channel, 576.0 GB/s) gives it a massive bandwidth advantage for memory-bound tasks, even if the specific benchmark data doesn’t isolate this.

The Threadripper PRO 3995WX wins where raw core count and older instruction paths matter. Its 53.7% lead in data encryption and 28.8% lead in data compression make it the superior choice for archival, backup, and security workloads. The 18.9% integer math advantage and 8.2% floating-point math advantage indicate it handles mathematical computation better when the problem fits within its 256 MB L3 cache. Random string sorting (+6.7%) is a niche but relevant win for database and text-processing applications. Its higher average benchmark score (171748 vs 160358) and higher percentile ranking among its own nearest rivals (2.4% above the EPYC 7C13) suggest that in mixed workloads, the extra 32 cores can still win out despite older architecture. For users with workloads that hit these specific PassMark categories, the 64-core Threadripper PRO remains a viable, albeit older, workhorse.