AMD EPYC 9655P vs AMD EPYC 9734 Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive sweep across all 17 head-to-head benchmark comparisons. The AMD EPYC 9655P wins every single test, with no wins recorded for the AMD EPYC 9734. The margin of victory varies significantly by workload, ranging from a modest 12% advantage to a staggering 283.1% lead.

The Cinebench suite reveals a remarkably consistent pattern. Across all six Cinebench tests (R15, R20, and R23, both single-core and multi-core), the EPYC 9655P holds an identical 56.8% advantage over the EPYC 9734. In Cinebench R23 multi-core, the 9655P scores 136354 against 86943 for the 9734. Single-core R23 shows the same proportional gap: 19250 versus 12274. This uniformity across both single-threaded and multi-threaded rendering workloads suggests the performance delta is fundamental to the architecture rather than dependent on core count scaling.

PassMark testing breaks down the picture further. The largest single delta appears in PassMark physics, where the 9655P scores 25847 against 6747 for the 9734, a 283.1% lead. This is the outlier result in the entire dataset. Prime number finding also shows an extreme gap: 1686 versus 829, a 103.4% difference. These two workloads amplify the architectural advantage far beyond the average.

Integer math shows a 48.8% lead (1225251 versus 823150), while floating point math trails at 30.4% (715866 versus 549045). Data encryption shows a 22.7% advantage (220074 versus 179390), and data compression a 20.2% edge (3486158 versus 2900008). Extended instructions come in at 12% (230609 versus 205925), the smallest margin in the entire comparison. Random string sorting shows a 26.3% gap (451824 versus 357638). PassMark multi-thread shows a 56.9% lead (160490 versus 102286), nearly matching the Cinebench multi-core delta. Single-thread PassMark shows 66.6% (3849 versus 2310).

The average benchmark score reinforces the hierarchy. The 9655P records an average of 397773, placing it in the 100th percentile of all CPUs in the database. The 9734 averages 310619, sitting at the 99th percentile. The 9655P's nearest rival, the AMD Ryzen Threadripper PRO 9995WX, averages 412068, which is 3.5% ahead of the 9655P. Meanwhile, the 9734 sits within a tight cluster: the AMD EPYC 9575F is 0.4% behind it, the Intel Xeon 6781P is 1.6% ahead, and the AMD Ryzen Threadripper PRO 9985WX is 3.2% ahead.

Architecture Differences

The two processors come from different generations and use different microarchitectures. The EPYC 9655P belongs to the EPYC 9005 series, built on Zen 5 architecture under the codename Turin. The EPYC 9734 belongs to the EPYC 9004 series, built on Zen 4c architecture under the codename Bergamo. The process nodes differ accordingly: the 9655P uses a 4 nm process from TSMC, while the 9734 uses a 5 nm process, also from TSMC.

Transistor counts and die configurations diverge substantially. The 9655P packs 99,780 million transistors across 12 dies, each measuring 70.6 mm². The 9734 contains 71,000 million transistors across 8 dies, each measuring 73 mm². Despite having fewer cores, the 9655P carries significantly more transistors, reflecting the denser Zen 5 design.

Core and thread counts favor the 9734 numerically. The 9734 offers 112 cores and 224 threads, while the 9655P offers 96 cores and 192 threads. Despite the 16-core deficit, the 9655P wins every benchmark, which points to per-core efficiency advantages in the Zen 5 design.

Clock speeds tell a similar story. The 9655P runs at a 2.60 GHz base clock and boosts to 4.50 GHz. The 9734 runs at a 2.20 GHz base clock and boosts to only 3.00 GHz. The 1.50 GHz boost clock advantage for the 9655P is substantial and directly contributes to its single-threaded dominance.

Cache hierarchies differ as well. The 9655P provides 80 KB of L1 cache per core, 1 MB of L2 cache per core, and 384 MB of shared L3 cache. The 9734 provides 64 KB of L1 per core, 1 MB of L2 per core, and 256 MB of shared L3 cache. Both use the same per-core L2 size, but the 9655P has more L1 per core and 128 MB more shared L3.

Memory support shows both processors using DDR5 with a twelve-channel memory bus. However, memory bandwidth differs: the 9655P reaches 576.0 GB/s, while the 9734 reaches 460.8 GB/s. Both support ECC memory. PCIe connectivity is identical: Gen 5 with 128 lanes (CPU only). Both use AMD Socket SP5, and both have locked multipliers.

Power envelopes differ modestly. The 9655P carries a 400 W TDP, while the 9734 carries a 340 W TDP. The 9655P draws more power but delivers higher performance across the board.

Where Each One Wins

The data shows no benchmark categories where the EPYC 9734 wins. However, the size of the 9655P's advantage varies by workload type, which informs deployment decisions.

For single-threaded and lightly threaded workloads, the 9655P is the clear choice. Cinebench R23 single-core shows a 56.8% lead, and PassMark single-thread shows a 66.6% lead. The boost clock difference of 1.50 GHz drives this advantage. Workloads that depend on per-core performance, such as database query processing or certain simulation tasks, will see outsized gains.

For multi-threaded rendering and compute, the 9655P also wins despite having 16 fewer cores. Cinebench R23 multi-core shows a 56.8% lead, and PassMark multi-thread shows a 56.9% lead. The 9655P's architectural efficiency and higher clocks overcome the core count deficit. Rendering farms and batch compute jobs benefit from the 9655P's superior throughput.

The most extreme advantage appears in physics simulation. The 283.1% lead in PassMark physics is the largest gap in the dataset. Prime number finding also shows an extreme 103.4% advantage. These workloads amplify the architectural differences between Zen 5 and Zen 4c.

For integer-heavy workloads, the 9655P holds a 48.8% lead. Floating point workloads show a 30.4% advantage. Encryption workloads show 22.7%, and compression shows 20.2%. Extended instruction workloads show the smallest gap at 12%, suggesting the 9734's Zen 4c core handles these tasks relatively better, though still losing.

The 9734's only theoretical advantage lies in core count. With 112 cores versus 96, it provides 16 additional cores and 32 additional threads for workloads that scale purely by thread count. However, the benchmark data indicates that even in multi-threaded tests, the 9655P's per-core strength dominates. The 9734 also carries a lower 340 W TDP, which may matter for power-constrained deployments.

FAQ

Q: Which processor has more cores?

A: The AMD EPYC 9734 has 112 cores and 224 threads, while the AMD EPYC 9655P has 96 cores and 192 threads.

Q: How much faster is the EPYC 9655P in single-threaded performance?

A: The 9655P leads by 56.8% in Cinebench R23 single-core and 66.6% in PassMark single-thread. Its boost clock reaches 4.50 GHz versus 3.00 GHz for the 9734.

Q: Does the EPYC 9734 win any benchmarks?

A: No. The recorded data shows 17 head-to-head benchmark comparisons, and the EPYC 9655P wins all 17.

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

A: The largest gap is in PassMark physics, where the 9655P scores 25847 versus 6747, a 283.1% advantage. Prime number finding shows the second-largest gap at 103.4%.

Q: How do the two compare in memory bandwidth?

A: The 9655P provides 576.0 GB/s of memory bandwidth, while the 9734 provides 460.8 GB/s. Both use DDR5 with a twelve-channel memory bus.

Q: What are the TDP ratings for each processor?

A: The 9655P has a 400 W TDP, and the 9734 has a 340 W TDP.

Specification Differences

| Specification | AMD EPYC 9655P | AMD EPYC 9734 |

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

| Series | EPYC 9005 series | EPYC 9004 series |

| Architecture | Zen 5 | Zen 4 |

| Codename | Turin | Bergamo |

| Process node | 4 nm | 5 nm |

| Transistors | 99,780 million | 71,000 million |

| Die size | 12x 70.6 mm² | 8x 73 mm² |

| Cores | 96 | 112 |

| Threads | 192 | 224 |

| Base clock | 2.60 GHz | 2.20 GHz |

| Boost clock | 4.50 GHz | 3.00 GHz |

| TDP | 400 W | 340 W |

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

| L3 cache | 384 MB (shared) | 256 MB (shared) |

| Memory bandwidth | 576.0 GB/s | 460.8 GB/s |

| Release date | 2024-10-09 | 2023-06-12 |

| Launch MSRP | $10811 | $9600 |

Both processors share the following specifications: AMD Socket SP5, 1 MB L2 cache per core, DDR5 memory support, twelve-channel memory bus, ECC memory support, PCIe Gen 5 with 128 lanes (CPU only), locked multipliers, and server/workstation market segment. Both are actively in production and made by TSMC.

The Verdict

The benchmark data makes the choice straightforward for nearly every workload. The AMD EPYC 9655P outperforms the AMD EPYC 9734 in all 17 recorded comparisons, with margins from 12% to 283.1%. It achieves this despite having 16 fewer cores and 32 fewer threads, relying instead on the Zen 5 architecture, a 4 nm process, a 1.50 GHz boost clock advantage, and 128 MB more L3 cache.

The 9655P is the correct selection for single-threaded performance, multi-threaded rendering, physics simulation, integer math, floating point math, encryption, compression, and sorting workloads. Its 100th percentile ranking versus the 9734's 99th percentile confirms its position at the top of the database.

The 9734 retains relevance only in specific scenarios. Its 112 cores provide more raw thread count, and its 340 W TDP draws less power than the 9655P's 400 W. For deployments where thread count is the sole consideration and power draw must be minimized, the 9734 remains a viable option. However, the recorded benchmarks show that even in multi-threaded tests, the 9655P wins by nearly 57%. The 9734's nearest rivals include the AMD EPYC 9575F at 0.4% behind and the Intel Xeon 6781P at 1.6% ahead, placing it in a competitive but not leading position.

For buyers choosing between these two specific processors, the data points overwhelmingly to the 9655P. The launch MSRP of $10811 for the 9655P compares to $9600 for the 9734, but the performance differential across the benchmark suite justifies the higher price for most workloads. The 9655P represents the current generation, released in October 2024, while the 9734 dates to June 2023. The two-year architectural leap from Zen 4c to Zen 5, combined with the process shrink from 5 nm to 4 nm, delivers performance gains that the 9734 cannot counter despite its additional cores.