CPU Comparison

The AMD EPYC 9845 and AMD EPYC 9535 are both members of the EPYC 9005 series, sharing the Turin codename and the AMD Socket SP5 platform, yet they are engineered for fundamentally different roles within the server and workstation market. The 9845 is a massive 160-core density-optimized part built on a 3 nm process, while the 9535 is a 64-core general-purpose processor on a 4 nm node. Benchmark data reveals a clear division of labor: the 9845 dominates in every multi-threaded workload tested, while the 9535 counterattacks with a decisive single-thread advantage. This analysis breaks down the head-to-head results, architectural divergences, and the specific use cases where each processor's strengths dictate the optimal choice.

Head-to-Head Benchmarks

The head-to-head benchmark results are overwhelmingly one-sided, with the AMD EPYC 9845 claiming victory in 9 out of 11 tests. The most extreme margin comes in the PassMark physics test, where the 9845 scores 19,631 versus the 9535's 3,834, a staggering 412% advantage. This indicates a massive disparity in raw simulation and physics-processing throughput, likely stemming from the 9845's vastly higher core count.

In compute-heavy integer workloads, the 9845 continues its dominance. It scores 1,687,531 in PassMark integer math, a 131.1% lead over the 9535's 730,281. Data encryption shows a similar pattern, with the 9845 achieving 296,808 against 127,372, a 133% delta. Floating-point math also heavily favors the larger chip, where the 9845's 978,377 score is 114.1% higher than the 9535's 457,047. These results paint a picture of a processor that scales nearly linearly with its 2.5x core-count advantage in parallelizable arithmetic tasks.

Memory-sensitive workloads further reinforce the 9845's lead. In data compression, the 9845 scores 4,680,013, which is 102.7% higher than the 9535's 2,308,822. Random string sorting follows suit, with the 9845 at 538,060 versus 247,506, a 117.4% advantage. Extended instruction throughput also favors the 9845, which posts 314,798 compared to the 9535's 175,784, a 79.1% lead. Even the prime number finding test, which can be sensitive to clock speed, shows the 9845 ahead by 43.6% (1,255 vs 874).

The multithreaded PassMark score, which aggregates overall parallel performance, gives the 9845 a 33.6% win (152,985 vs 114,528). While this is a significant victory, it is notably smaller than the core-count ratio might suggest, hinting that the 9535's higher clocks help mitigate the gap in certain mixed workloads.

The single bright spot for the 9535 is in the two single-thread tests. It scores 3,720 in both PassMark single-thread and single-thread, versus the 9845's 3,144. This 15.5% advantage is critical for applications that cannot utilize many cores. The 9535's higher base clock of 2.40 GHz and boost clock of 4.30 GHz, compared to the 9845's 2.10 GHz and 3.70 GHz, directly explain this result.

Architecture Differences

The two processors share the same Zen 5 architecture and Turin codename but diverge significantly in their implementation. The EPYC 9845 is designated as a Zen 5c (Turin) part, manufactured on a 3 nm process at TSMC, while the EPYC 9535 is a standard Zen 5 (Turin) chip built on a 4 nm process. This process node difference is foundational to their design goals.

The 9845 packs 160 cores and 320 threads, while the 9535 houses 64 cores and 128 threads. This 2.5x core and thread ratio is the primary driver of the 9845's multi-threaded supremacy. The 9845's cache hierarchy scales accordingly: it features 80 KB of L1 and 1 MB of L2 per core, identical to the 9535, but its shared L3 cache is 320 MB compared to the 9535's 256 MB. This larger pool of shared cache benefits workloads with large working sets.

Clock speeds tell the opposite story. The 9535 operates at a 2.40 GHz base clock and 4.30 GHz boost clock, while the 9845 is locked to a 2.10 GHz base and 3.70 GHz boost. This 600 MHz boost clock deficit for the 9845 explains its 15.5% single-thread performance loss. The 9535 also carries a higher TDP of 300 W, but the 9845 draws more power at 390 W, reflecting the sheer scale of its core count.

Both processors support DDR5 memory through a twelve-channel bus, delivering an identical 576.0 GB/s of memory bandwidth. Both also feature ECC memory support and provide Gen 5 PCIe with 128 lanes for CPU-only connectivity. The 9535 has a published transistor count of 66,520 million across an 8x 70.6 mm² die configuration, while the 9845's transistor count and die size are not listed in the data.

The 9845's process advantage (3 nm vs 4 nm) allows AMD to fit 160 cores within a 390 W TDP envelope, whereas the 9535's larger, denser cores on the older node require a 300 W budget for just 64 cores. This architectural split is a classic density-versus-frequency tradeoff: the 9845 maximizes throughput per socket, while the 9535 optimizes per-core performance and clock headroom.

Where Each One Wins

The EPYC 9845 is the unequivocal choice for throughput-bound workloads that can scale across hundreds of threads. Its 412% lead in physics simulations makes it ideal for scientific computing, engineering analysis, and any physics-based modeling that runs in parallel. The 131.1% advantage in integer math and 114.1% lead in floating-point math position it as a powerhouse for financial modeling, cryptography, and high-performance computing clusters. Data-intensive tasks like compression and sorting, where it holds leads of 102.7% and 117.4% respectively, benefit from its massive L3 cache and core count.

In contrast, the EPYC 9535 wins where single-thread latency and clock speed are paramount. Its 15.5% single-thread advantage makes it the better fit for legacy applications, database transaction processing with low concurrency, or workloads with strict serial dependencies that cannot be parallelized. The 9535's higher boost clock of 4.30 GHz also benefits lightly threaded virtual machines or containerized environments where individual processes are latency-sensitive. For organizations running a mix of moderately threaded workloads, the 9535's 33.6% deficit in multithreaded performance is a more palatable tradeoff when weighed against its 15.5% single-thread edge.

The 9535's lower TDP of 300 W versus 390 W also suggests it may be a more practical fit for rack densities where thermal headroom is limited, though the data does not specify cooling requirements. The 9845's 160 cores in a single socket can replace multiple smaller processors, reducing per-socket overhead in large-scale deployments, but only if the software stack can saturate its threads.

The Verdict

The data is unambiguous: the AMD EPYC 9845 is superior for any workload that is primarily multi-threaded. Its 102.7% lead in data compression and 133% lead in encryption demonstrate that it nearly doubles the 9535's output in these areas. The 412% physics score gap is a generational leap in simulation capability. For server environments running high-core-count virtual machines, large-scale data analytics, or rendering farms, the 9845 is the correct choice based on benchmark performance alone.

The AMD EPYC 9535 is the better pick when single-thread performance is the binding constraint. Its 15.5% single-thread advantage over the 9845 is a decisive factor for applications that cannot use more than a handful of cores, such as certain real-time processing, front-end web servers, or developer workstations. The 9535 still delivers respectable multi-threaded performance, sitting within 33.6% of a 160-core CPU, which speaks to its efficiency per core. In a mixed environment where some workloads are serial and others are parallel, the 9535 offers a more balanced profile, though it sacrifices raw throughput for that flexibility.

Both processors rank in the 100th percentile of all CPUs, indicating top-tier performance in their respective domains. The 9845's average benchmark score of 523,613 is 38% higher than the 9535's 379,408. However, the 9535's nearest rival data shows it is only 1.6% behind the AMD EPYC 9655 and 3.9% ahead of the Intel Xeon 6960P, placing it in a competitive mid-range tier. The 9845, by contrast, is 3.5% ahead of the EPYC 9755 and 22.9% ahead of the EPYC 9745, cementing its position as a halo product for maximum core density.

FAQ

Q: Which processor has a higher core count?

A: The AMD EPYC 9845 has 160 cores and 320 threads, while the AMD EPYC 9535 has 64 cores and 128 threads.

Q: How much faster is the EPYC 9845 in multi-threaded performance?

A: The EPYC 9845 leads by 33.6% in the PassMark multithread test, scoring 152,985 versus the 9535's 114,528.

Q: What is the single-thread performance difference?

A: The EPYC 9535 is 15.5% faster in single-thread tests, scoring 3,720 versus the 9845's 3,144 in PassMark single-thread.

Q: Do both processors support the same memory and PCIe?

A: Yes, both support DDR5 with a twelve-channel bus delivering 576.0 GB/s, and both provide Gen 5 PCIe with 128 lanes.

Q: What is the L3 cache difference between the two?

A: The EPYC 9845 has 320 MB of shared L3 cache, while the EPYC 9535 has 256 MB of shared L3 cache.

Q: Which processor has a higher boost clock?

A: The EPYC 9535 has a boost clock of 4.30 GHz, which is higher than the EPYC 9845's boost clock of 3.70 GHz.