CPU Comparison

Both the AMD EPYC 9745 and the AMD EPYC 9535 are members of the AMD EPYC 9005 series, launched on the same date and built for the same AMD Socket SP5 platform. Despite their shared family lineage, the data reveals two distinctly different processors: the 9745 is a 128-core behemoth designed to obliterate multi-threaded workloads, while the 9535 is a 64-core part with a significant clock-speed advantage that dominates single-threaded tasks. The benchmark results show a clear trade-off between core count and frequency, making the choice between them a straightforward question of workload priorities.

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9745 has 128 cores and 256 threads, exactly double the 64 cores and 128 threads of the AMD EPYC 9535.

Q: What is the difference in their boost clocks?

A: The AMD EPYC 9535 has a boost clock of 4.30 GHz, which is significantly higher than the 3.70 GHz boost clock of the AMD EPYC 9745. Both share the same 2.40 GHz base clock.

Q: Which CPU wins in single-threaded performance?

A: The AMD EPYC 9535 wins decisively in the PassMark single-thread test with a score of 3720, compared to the 9745's score of 2806. This represents a 24.6% advantage for the 9535.

Q: Which CPU has the higher TDP?

A: The AMD EPYC 9745 has a TDP of 400, which is 100 higher than the 300 TDP of the AMD EPYC 9535.

Q: Are they built on the same manufacturing process?

A: No. The AMD EPYC 9745 uses a 3 nm process node, while the AMD EPYC 9535 is built on a 4 nm process node. Both are manufactured by TSMC.

Q: What is the average benchmark score difference between them?

A: The AMD EPYC 9745 has an average benchmark score of 425973, while the AMD EPYC 9535 scores 379408. The 9745 leads by 12.3% in this metric.

Architecture Differences

The core architectural difference between these two parts lies in their core density versus clock speed trade-off. The AMD EPYC 9745 is based on the Zen 5c (Turin) generation and packs 128 cores onto a 3 nm TSMC process node. In contrast, the AMD EPYC 9535 uses the standard Zen 5 (Turin) design on a 4 nm process node. This process difference is a key enabler for the 9745's massive core count, allowing more cores to fit within the same thermal and power envelope.

Both processors share identical cache hierarchies: 80 KB of L1 per core, 1 MB of L2 per core, and a substantial 256 MB of shared L3 cache. The memory subsystem is also identical on paper, with DDR5 support, a twelve-channel memory bus, and 576.0 GB/s of memory bandwidth. Both support ECC memory and offer Gen 5 PCIe with 128 lanes. The 9535's datasheet lists 66,520 million transistors across an 8x 70.6 mm² die configuration, whereas the 9745's transistor count and die size are not provided in the data.

The practical consequence of the 9745's denser 3nm process is that it can double the core count while maintaining the same base clock of 2.40 GHz. However, this density comes at a cost: the 9745's boost clock is capped at 3.70 GHz, a full 600 MHz lower than the 9535's 4.30 GHz boost. This clock deficit is the primary reason the 9535 excels in single-threaded workloads, despite having half the cores. The 9745 also carries a higher TDP of 400, reflecting the power demands of its 128 cores, compared to the 9535's 300 TDP.

Where Each One Wins

The AMD EPYC 9745 is the undisputed champion of heavily parallelized, throughput-intensive workloads. Its 128 cores provide a massive advantage in scenarios that can utilize all available threads. The benchmark data shows it winning 9 out of 11 head-to-head comparisons, with particularly large margins in physics calculations and data compression. For workloads like database compression, scientific simulation, or large-scale virtualization, the 9745's core count is a decisive factor.

The AMD EPYC 9535, on the other hand, is the clear winner in single-threaded performance. Its 4.30 GHz boost clock gives it a 24.6% advantage over the 9745 in the PassMark single-thread test. This makes it the better choice for workloads that are latency-sensitive or rely on a few fast cores, such as certain database queries, legacy applications, or lightly-threaded server tasks. While it loses in raw multi-threaded throughput, its higher frequency can provide snappier response times for interactive workloads.

The data also shows that the 9535 is more closely matched with other rivals in its own performance tier. Its nearest rival, the AMD EPYC 9655, is only 1.6% behind in average score, while the Intel Xeon 6960P trails by 3.9%. The 9745, by contrast, holds a 3.4% lead over the AMD Ryzen Threadripper PRO 9995WX and a 7.1% lead over the AMD EPYC 9655P. This indicates that the 9745 sits at the very top of the performance pyramid, while the 9535 competes in a tighter, but still elite, bracket.

Specification Differences

| Specification | AMD EPYC 9745 | AMD EPYC 9535 |

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

| Cores | 128 | 64 |

| Threads | 256 | 128 |

| Boost Clock | 3.70 GHz | 4.30 GHz |

| TDP | 400 | 300 |

| Architecture | Zen 5 | Zen 5 |

| Codename | Turin | Turin |

| Generation | EPYC (Zen 5c (Turin)) | EPYC (Zen 5 (Turin)) |

| Process Node | 3 nm | 4 nm |

| Transistors | Not specified | 66,520 million |

| Die Size | Not specified | 8x 70.6 mm² |

| Launch MSRP | $12141 | $8992 |

| Part Number | 100-000001460 | 100-000001147 |

The table above highlights the key differences. The core count and TDP are the most significant physical differences, with the 9745 doubling the cores and increasing power draw by 100 watts. The process node difference (3 nm vs 4 nm) is the underlying enabler for this core-density increase. The boost clock difference of 600 MHz is the primary driver of the single-threaded performance gap. The launch MSRP is also distinct, but the decision should be based on performance needs rather than cost. All other specifications, including base clock, cache sizes, memory support, socket, and PCIe lanes, are identical between the two.

Head-to-Head Benchmarks

The head-to-head benchmark results paint a stark picture of dominance for the 9745 in multi-threaded tests. In the PassMark physics test, the 9745 scores 17122 versus the 9535's 3834, a staggering 346.6% advantage. This is the largest margin in the entire comparison and highlights the 9745's ability to handle complex physical simulations with ease. Data compression also shows a massive gap, with the 9745 scoring 3929890 against 2308822, a 70.2% lead.

The 9745 continues its dominance in other compute-heavy tasks. Integer math sees the 9745 at 1224315 versus 730281, a 67.6% advantage. Floating-point math follows a similar pattern, with the 9745 scoring 761219 compared to the 9535's 457047, a 66.6% lead. Data encryption shows an 80.1% advantage for the 9745 (229447 vs 127372), and random string sorting sees a 89.5% lead (468975 vs 247506). Extended instruction sets also favor the 9745, with a 59.6% margin (280477 vs 175784). Even in the find prime numbers test, the 9745 wins, albeit by a smaller 12% margin (979 vs 874). The multi-threaded PassMark score reflects this trend, with the 9745 achieving 130698 versus 114528, a 14.1% lead.

The single-threaded test is the lone bright spot for the 9535. Here, the 9535 scores 3720, while the 9745 manages only 2806. This 24.6% advantage for the 9535 is the only category where it beats the 9745, and it does so by a substantial margin. This single win is dwarfed by the 9745's victories in the other nine benchmarks, but it is a critical data point for specific workloads. The overall average benchmark score reflects the 9745's broader dominance, with a 12.3% lead in that aggregate metric.

The Verdict

The data supports a clear verdict: choose the AMD EPYC 9745 for maximum multi-threaded throughput. It wins 9 out of 11 benchmark comparisons, with decisive leads in physics, compression, and math workloads. Its 128 cores provide a level of parallel processing power that the 64-core 9535 cannot match, making it the superior choice for heavily virtualized environments, large-scale data processing, and high-performance computing tasks. The 9745's 12.3% lead in average benchmark score over the 9535, and its 100th percentile ranking among all CPUs, confirms its position as a top-tier performer.

Choose the AMD EPYC 9535 only if single-threaded performance is your absolute priority. Its 4.30 GHz boost clock delivers a 24.6% advantage in single-threaded tasks, which is a significant edge for software that cannot leverage many cores. This could be relevant for specific database applications, high-frequency trading, or workloads with strict latency requirements. However, it is important to note that the 9535 still loses in overall multi-threaded performance and average score. Its lower TDP of 300 might make it easier to cool, but the data shows that for any workload that can utilize more than 64 cores, the 9745 is the superior processor without exception.