AMD EPYC 4564P vs AMD Ryzen Threadripper PRO 9955WX Comparison

The AMD Ryzen Threadripper PRO 9955WX and the AMD EPYC 4564P occupy unusual positions for a head-to-head: both are 16-core, 32-thread processors from AMD, both run a 4.50 GHz base clock, and both sit in the top few percent of the database, at the 97th and 96th percentile versus all CPUs respectively. The recorded data shows the Threadripper PRO 9955WX winning 14 of 17 shared benchmarks, yet the EPYC 4564P holds its ground in specific workloads and does so at a far lower TDP of 170 W versus 350 W. The Threadripper's average benchmark score of 101041 against the EPYC's 95183 frames this as a genuine contest rather than a rout, and the details of where each one wins matter more than the aggregate.

Head-to-Head Benchmarks

The Cinebench suite is a clean sweep for the Threadripper PRO 9955WX, and the margin is remarkably consistent. In Cinebench R15 the scores are 5996 versus 5603 in multi-core and 846 versus 791 in single-core, both a 7 percent gap. R20 repeats the pattern: 24987 versus 23348 multi-core, 3527 versus 3296 single-core, again 7 percent each way. R23 lands at 59494 versus 55592 in multi-core and 8399 versus 7848 in single-core, still 7 percent. That uniformity across three render-heavy test generations indicates a structural advantage in sustained rendering throughput rather than a quirk of any single test.

The Passmark suite is where the picture gets interesting. The Threadripper's biggest wins come in physics at 4156 versus 3392, a 22.5 percent gap, and in extended instructions at 76363 versus 63136, a 21 percent gap. Floating point math goes to the Threadripper by 10.8 percent (156215 versus 141017), data compression by 7.3 percent (920954 versus 858105), multithread by 4.2 percent (67035 versus 64357), integer math by 3.2 percent (236120 versus 228761), and single-thread by 5.5 percent (4530 versus 4292).

The EPYC 4564P's three wins are specific and notable. Data encryption is its strongest result: 50708 versus 44389, a 12.5 percent lead for the EPYC despite its nominally lower single-thread result. Find prime numbers also goes its way at 365 versus 337, a 7.7 percent gap, and random string sorting at 103202 versus 99813, a 3.3 percent margin. These are integer- and branch-sensitive tests, and the EPYC's higher 5.70 GHz boost clock, against the Threadripper's 5.40 GHz, is the plausible explanation the data supports. The EPYC wins precisely where raw per-thread speed dominates and memory bandwidth does not.

Where Each One Wins

For the Threadripper PRO 9955WX, the use-case split is clear: rendering, simulation, and memory-bandwidth-hungry professional workloads. The 7 percent Cinebench sweep covers content creation directly, and the 22.5 percent physics lead points to simulation and game-development toolchains. The 21 percent extended-instructions win and 10.8 percent floating-point win favor scientific compute and code that leans on wider vector throughput. Its eight-channel DDR5 bus delivering 409.6 GB/s of bandwidth, nearly five times the EPYC's 83.2 GB/s on dual-channel, means large dataset workloads will scale well beyond what these synthetic scores fully capture.

The EPYC 4564P wins where per-clock integer performance and efficiency matter. Encryption-heavy server duties, prime-based workloads, and string sorting all favor it, and its 170 W TDP is less than half the Threadripper's 350 W. In the database's aggregate terms, the Threadripper sits just below the AMD EPYC 7513 (average score 102244, a 1.2 percent gap) and the AMD EPYC 8324P (103329, 2.2 percent ahead), while beating the AMD EPYC 4585PX by 1.7 percent and the AMD Ryzen Threadripper PRO 5965WX by 2.6 percent. The EPYC 4564P's neighborhood is tighter: the AMD EPYC 9175F is only 0.5 percent ahead, the AMD EPYC 4565P 0.6 percent ahead, the AMD Ryzen 9 PRO 9945 0.9 percent ahead, and it leads the Intel Xeon 6520P by 1.5 percent. Both CPUs are locked in dense clusters of near-equal rivals, but the Threadripper's cluster sits at a higher absolute level.

The platform dimension reinforces the split. The Threadripper's 128 PCIe Gen 5 lanes on Socket sTR5 suit GPU-heavy workstation builds and storage arrays; the EPYC's 28 lanes on Socket AM5 target conventional single-socket server deployments.

Architecture Differences

The Threadripper PRO 9955WX is a Zen 5 design, codenamed Shimada Peak, built on TSMC's 4 nm process with 16,630 million transistors across two dies of 70.6 mm² each. The EPYC 4564P is a Zen 4 design, codenamed Raphael, on TSMC's 5 nm process with 13,140 million transistors across two dies of 71 mm² each. The generational jump from Zen 4 to Zen 5 is the architectural root of the Threadripper's broad benchmark lead, and the newer node carries more transistors in nearly identical die area.

Cache configurations are identical on paper: 64 KB of L1 per core, 1 MB of L2 per core, and 64 MB of L3. The memory subsystems diverge sharply. Both support DDR5 with ECC, but the Threadripper runs an eight-channel bus at 409.6 GB/s while the EPYC runs dual-channel at 83.2 GB/s. That bandwidth disparity is the single largest platform difference between them and the most likely driver of the Threadripper's compression and multithread advantages.

One more divergence: the EPYC 4564P includes integrated Radeon Graphics, while the Threadripper has none. The EPYC's multiplier is locked; the Threadripper's is unlocked. Both remain in active production. The Threadripper launched on 2025-07-22 with a launch MSRP of $1649; the EPYC launched on 2024-05-20 with a launch MSRP of $699.

FAQ

Q: Which CPU is faster overall?

A: The Ryzen Threadripper PRO 9955WX. It wins 14 of 17 head-to-head benchmarks and carries a higher average benchmark score, 101041 versus 95183.

Q: Does the EPYC 4564P win anything?

A: Yes, three tests: data encryption (50708 versus 44389, a 12.5 percent lead), find prime numbers (365 versus 337), and random string sorting (103202 versus 99813).

Q: Both chips have 16 cores and a 4.50 GHz base clock, so why the performance gap?

A: Architecture and platform. The Threadripper uses Zen 5 on a 4 nm node with an eight-channel memory bus, while the EPYC uses Zen 4 on 5 nm with dual-channel memory. The 7 percent Cinebench gap and 22.5 percent physics gap reflect that combination.

Q: Which one boosts higher?

A: The EPYC 4564P, at 5.70 GHz boost versus 5.40 GHz for the Threadripper, which helps explain its wins in clock-sensitive integer tests.

Q: How do they compare in power draw?

A: The EPYC 4564P has a 170 W TDP versus 350 W for the Threadripper, less than half, despite delivering most of the multithreaded performance (64357 versus 67035 in Passmark multithread).

Q: Do both support ECC memory?

A: Yes, both support DDR5 with ECC enabled.

Specification Differences

| Field | Ryzen Threadripper PRO 9955WX | EPYC 4564P |

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

| Architecture | Zen 5 | Zen 4 |

| Codename | Shimada Peak | Raphael |

| Process node | 4 nm | 5 nm |

| Boost clock | 5.40 GHz | 5.70 GHz |

| TDP | 350 W | 170 W |

| Socket | AMD Socket sTR5 | AMD Socket AM5 |

| Transistors | 16,630 million | 13,140 million |

| Die size | 2x 70.6 mm² | 2x 71 mm² |

| Memory bus | Eight-channel | Dual-channel |

| Memory bandwidth | 409.6 GB/s | 83.2 GB/s |

| PCIe | Gen 5, 128 lanes (CPU only) | Gen 5, 28 lanes (CPU only) |

| Integrated graphics | N/A | Radeon Graphics |

| Market segment | Desktop | Server/Workstation |

| Series | 9000 series | EPYC 4004 series |

| Release date | 2025-07-22 | 2024-05-20 |

| Launch MSRP | $1649 | $699 |

| Multiplier unlocked | Yes | No |

| Part number | 100-000000725 | 100-000001476 |

Cores (16), threads (32), base clock (4.50 GHz), cache hierarchy (64 KB L1 per core, 1 MB L2 per core, 64 MB L3), DDR5 support, ECC support, foundry (TSMC), and production status (Active) are identical between the two.