AMD EPYC 9115 vs AMD Ryzen Threadripper PRO 5955WX Comparison

The AMD EPYC 9115 and AMD Ryzen Threadripper PRO 5955WX are both 16-core, 32-thread workstation processors, yet they represent two distinct generations of AMD silicon. The EPYC 9115 is a Zen 5 part from the EPYC 9005 series, built for the modern SP5 platform, while the Threadripper PRO 5955WX is a Zen 3 part from the 5000 series on the WRX8 platform. Benchmark data shows a near-total sweep for the newer EPYC in rendering workloads, but the older Threadripper fights back decisively in specific data-processing tasks. The results paint a picture of generational efficiency gains meeting mature platform advantages.

Where Each One Wins

The EPYC 9115 is the clear winner in CPU-bound compute and physics simulations. Across all Cinebench iterations, the EPYC 9115 holds a consistent edge, though the margins are slim. In Cinebench R23 multi-core, the EPYC scores 42,003 against the Threadripper’s 41,837, a 0.4% lead. The same 0.4% delta appears in Cinebench R20 multi-core (17,641 vs 17,571) and R15 multi-core (4,233 vs 4,217). Single-core results follow the same pattern, with the EPYC winning every Cinebench single-core test by 0.3% to 0.4%. This indicates that the Zen 5 architecture provides a modest but consistent IPC uplift over Zen 3 in these workloads.

The EPYC’s dominance extends to mathematical workloads. In PassMark’s find prime numbers test, the EPYC wins by a massive 16.5%, scoring 289 versus 248. Floating point math also favors the EPYC, with a 9.1% advantage (113,853 vs 104,377). The most dramatic gap is in PassMark physics, where the EPYC scores 4,188 against the Threadripper’s 2,913 — a 43.8% lead. This suggests the EPYC’s newer architecture handles complex physical simulations far more efficiently.

However, the Threadripper PRO 5955WX is not without its strongholds. It wins the PassMark data compression test decisively, scoring 690,994 versus the EPYC’s 600,099 — a 13.2% difference. Data encryption is another major victory, with the Threadripper scoring 42,524 against 33,489, a 21.2% lead. These are substantial margins in workloads that rely on memory bandwidth and specific instruction paths. The Threadripper also edges out the EPYC in integer math (185,168 vs 181,807, a 1.8% lead), extended instructions (46,952 vs 45,477, a 3.1% lead), and the overall PassMark multi-thread score (49,220 vs 48,936, a 0.6% lead). The final tally shows the EPYC winning 12 of 17 head-to-head benchmarks, but the Threadripper’s five wins are in areas where it dominates by double-digit percentages.

Architecture Differences

The two processors are built on fundamentally different nodes and microarchitectures. The EPYC 9115 uses the Zen 5 architecture on a 4 nm TSMC process, while the Threadripper PRO 5955WX uses the older Zen 3 architecture on a 7 nm TSMC process. This generational gap explains the EPYC’s efficiency in many benchmarks, despite both chips having the same core and thread counts.

The cache configurations differ notably. The EPYC 9115 features 80 KB of L1 cache per core and 1 MB of L2 per core, while the Threadripper has 64 KB of L1 per core and 512 KB of L2 per core. Both share 64 MB of L3 cache, but the EPYC’s larger per-core L1 and L2 caches likely contribute to its superior single-threaded performance. The physical layout also differs: the EPYC uses a dual-die design with 2x 70.6 mm² dies, while the Threadripper uses a quad-die configuration with 4x 81 mm² dies. Transistor counts are nearly identical (16,630 million for the EPYC, 16,600 million for the Threadripper), but the denser 4 nm process allows the EPYC to pack more performance into a smaller footprint.

Memory support represents a significant platform divide. The EPYC 9115 supports DDR5 memory across a twelve-channel bus, delivering 576.0 GB/s of bandwidth. The Threadripper PRO 5955WX is limited to DDR4 across an eight-channel bus, offering only 204.8 GB/s. This 2.8x difference in memory bandwidth is crucial — it likely explains why the Threadripper wins in data compression and encryption, where large data sets must be fed through the memory subsystem quickly. The EPYC’s higher bandwidth should theoretically benefit these workloads, but the benchmark results suggest the Threadripper’s implementation is more effective for these specific tasks.

PCIe capabilities also differ. The EPYC 9115 supports PCIe Gen 5 with 128 lanes, while the Threadripper PRO 5955WX supports PCIe Gen 4 with the same 128 lanes. The newer standard offers double the bandwidth per lane, which matters for future expansion. Both processors support ECC memory, and neither has a multiplier unlocked.

FAQ

Q: Which processor is faster in multi-core rendering?

A: The EPYC 9115 wins all Cinebench multi-core tests, but by very slim margins. It leads by 0.4% in Cinebench R23 (42,003 vs 41,837), R20 (17,641 vs 17,571), and R15 (4,233 vs 4,217). The difference is consistent but minimal.

Q: Why does the Threadripper PRO 5955WX win in data compression and encryption despite being older?

A: The Threadripper wins PassMark data compression by 13.2% (690,994 vs 600,099) and data encryption by 21.2% (42,524 vs 33,489). These workloads appear to favor its specific memory and instruction architecture, despite its lower theoretical memory bandwidth of 204.8 GB/s versus the EPYC’s 576.0 GB/s.

Q: What is the biggest performance gap between the two?

A: The largest margin is in PassMark physics, where the EPYC 9115 scores 4,188 versus the Threadripper’s 2,913 — a 43.8% advantage. The second-largest is in data encryption, where the Threadripper leads by 21.2%.

Q: Are these processors comparable in overall average performance?

A: The EPYC 9115 has an average benchmark score of 69,288, while the Threadripper PRO 5955WX averages 67,868. Both rank at the 94th percentile among all CPUs. The EPYC’s nearest rivals include the Intel Core i7-14700K (-0.1%) and AMD Ryzen 9 7950X (-0.3%), while the Threadripper sits near the AMD EPYC 4484PX (+0.1%) and Intel Xeon w5-3525 (+0.3%).

Q: Which processor has a higher boost clock?

A: The Threadripper PRO 5955WX has a higher boost clock of 4.50 GHz compared to the EPYC 9115’s 4.10 GHz. The Threadripper also has a higher base clock at 4.00 GHz versus 2.60 GHz.

Q: What are the memory and PCIe differences?

A: The EPYC 9115 uses DDR5 memory with a twelve-channel bus and 576.0 GB/s bandwidth, plus PCIe Gen 5 with 128 lanes. The Threadripper PRO 5955WX uses DDR4 with an eight-channel bus and 204.8 GB/s bandwidth, plus PCIe Gen 4 with 128 lanes.

Specification Differences

| Specification | AMD EPYC 9115 | AMD Ryzen Threadripper PRO 5955WX |

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

| Architecture | Zen 5 | Zen 3 |

| Codename | Turin | Chagall PRO |

| Process Node | 4 nm | 7 nm |

| Base Clock | 2.60 GHz | 4.00 GHz |

| Boost Clock | 4.10 GHz | 4.50 GHz |

| TDP | 125 W | 280 W |

| Socket | AMD Socket SP5 | AMD Socket WRX8 |

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

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

| Die Size | 2x 70.6 mm² | 4x 81 mm² |

| Memory Support | DDR5 | DDR4 |

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

| Memory Bandwidth | 576.0 GB/s | 204.8 GB/s |

| PCIe | Gen 5, 128 Lanes | Gen 4, 128 Lanes |

| Release Date | 2024-10-09 | 2022-03-07 |

| Part Number | 100-000001552 | 100-000000447 |

Head-to-Head Benchmarks

The Cinebench suite shows a uniform, if narrow, victory for the EPYC 9115. Across R15, R20, and R23, the multi-core margins are all exactly 0.4%, while single-core margins are 0.3% to 0.4%. The consistency suggests a fixed architectural advantage rather than workload-specific luck. In Cinebench R23 multi-core, the EPYC’s 42,003 score is just 166 points ahead of the Threadripper’s 41,837. These are effectively tied results for most practical purposes, but the EPYC takes the win every time.

PassMark results reveal the true character of each chip. The EPYC’s 16.5% win in prime number finding (289 vs 248) and 9.1% win in floating point math (113,853 vs 104,377) show its strength in pure computation. The 43.8% physics win (4,188 vs 2,913) is the single largest margin in the comparison, indicating a substantial generational leap in handling physics calculations. The EPYC also wins random string sorting by a narrow 0.4% (70,151 vs 69,879) and single-thread performance by 1.1% (3,360 vs 3,324).

The Threadripper’s counterattacks are concentrated and powerful. Its 13.2% win in data compression (690,994 vs 600,099) represents a 90,895-point gap. Data encryption shows an even larger 21.2% margin (42,524 vs 33,489). These are the workloads where the Threadripper’s older platform, with its higher base and boost clocks, proves superior. The Threadripper also wins integer math by 1.8% (185,168 vs 181,807), extended instructions by 3.1% (46,952 vs 45,477), and the overall multi-thread score by 0.6% (49,220 vs 48,936). The data suggests that for data-heavy server tasks, the Threadripper’s higher clock speeds and mature platform still deliver, even against a newer architecture with vastly superior memory bandwidth.