AMD EPYC 4564P vs AMD EPYC 4585PX Comparison

The AMD EPYC 4585PX and AMD EPYC 4564P are both 16-core, 32-thread server/workstation parts on the same socket, but they are not equals. The 4585PX, built on the newer Zen 5 architecture, wins the majority of benchmark comparisons, while the 4564P, a Zen 4 part, holds clear advantages in a couple of specific workloads. This analysis breaks down the data to show where each chip excels and which buyer profile each one suits.

Head-to-Head Benchmarks

The most striking pattern in the head-to-head results is the consistency of the 4585PX’s lead in CPU-intensive compute tasks. Across all six Cinebench tests—R15, R20, and R23, both single-core and multi-core—the 4585PX wins by exactly 8.7%. For instance, in Cinebench R23 multi-core, the 4585PX scores 60,451 against the 4564P’s 55,592. The single-core gap is equally pronounced, with the 4585PX hitting 8,534 versus 7,848 in R23 single-core. This uniform 8.7% delta suggests a fundamental architectural advantage rather than a workload-specific quirk.

The 4585PX also dominates in PassMark’s integer and floating-point math tests, winning both by 8.7%. In floating-point math, it scores 153,219 versus 141,017, and in integer math, 248,563 versus 228,761. The gap widens dramatically in PassMark’s find prime numbers test, where the 4585PX scores 547 against the 4564P’s 365, a 49.9% difference. Even more extreme is the physics test, where the 4585PX posts 6,612 versus 3,392, a massive 94.9% lead. These are not marginal improvements; the 4585PX is decisively faster in raw computational throughput.

However, the 4564P is not without its wins. It takes the PassMark data encryption test with a score of 50,708 versus 47,224, a 6.9% advantage for the 4564P. It also wins the random string sorting test, scoring 103,202 versus 95,210, a 7.7% lead. These two wins are notable because they highlight that the 4564P’s older architecture is not universally slower. In data compression, the 4585PX wins by a narrower 3.1% (884,774 vs 858,105), showing that even in its losses, the 4564P can keep things close. Overall, the 4585PX wins 15 of the 17 head-to-head benchmarks, with the 4564P taking only two.

Architecture Differences

The core architectural split is the process node and CPU generation. The 4585PX is built on TSMC’s 4 nm process with the Zen 5 architecture, codenamed Grado, while the 4564P uses the older Zen 4 architecture on a 5 nm process, codenamed Raphael. This node shrink is a key reason for the 4585PX’s consistent performance lead, as it allows for higher efficiency and improved instruction handling per clock.

Cache configurations differ significantly. The 4585PX has an L1 cache of 80 KB per core and a large 128 MB L3 cache, while the 4564P has 64 KB of L1 per core and a smaller 64 MB L3 cache (shared). Both have 1 MB of L2 per core. The 4585PX’s doubled L3 cache is a major contributor to its performance in multi-threaded workloads, as it reduces the need to access slower system memory.

Transistor counts and die sizes also reflect the generational leap. The 4585PX packs 16,630 million transistors across a die size of 2x 70.6 mm², while the 4564P has 13,140 million transistors on a 2x 71 mm² die. Despite the similar physical footprint, the 4585PX crams in over 3,000 million more transistors, which aligns with its larger cache and newer core design.

Both chips share the same socket (AMD Socket AM5), base TDP of 170 watts, and support for DDR5 memory with ECC. The memory bus is dual-channel for both, but the 4585PX has a higher memory bandwidth rating of 89.6 GB/s versus 83.2 GB/s for the 4564P. PCIe lanes are another difference: the 4564P offers 28 Gen 5 lanes (CPU only), while the 4585PX provides 24 Gen 5 lanes. Both include integrated Radeon Graphics, and both have a boost clock of 5.70 GHz; the base clock is 4.30 GHz for the 4585PX and 4.50 GHz for the 4564P.

The Verdict

The data paints a clear picture for most workloads: the AMD EPYC 4585PX is the superior processor. Its 8.7% lead across all Cinebench tests and the majority of PassMark tests makes it the default choice for compute-heavy server tasks. The 49.9% advantage in prime number finding and the 94.9% lead in physics simulation are particularly strong indicators for scientific computing, financial modeling, or any workload that stresses integer math and physics engines. With an average benchmark score of 99,324 and a 97th percentile ranking among all CPUs, the 4585PX sits in elite company, trading blows with the AMD Ryzen Threadripper PRO 9955WX (which scores 101,041, a 1.7% gap) and the AMD EPYC 7513 (102,244, a 2.9% gap).

The AMD EPYC 4564P, with an average score of 95,183 and a 96th percentile ranking, is no slouch. It is competitive with the AMD EPYC 9175F (95,615, a 0.5% gap) and the AMD EPYC 4565P (95,764, a 0.6% gap). However, it only wins in data encryption and random string sorting. If your primary workload is cryptographic operations or heavy string manipulation, the 4564P’s 6.9% and 7.7% leads, respectively, are worth considering. That said, for a general-purpose server CPU, the 4585PX’s dominant win count and higher average score make it the safer, more capable pick.

Specification Differences

The two processors differ in several key specification fields. The 4585PX uses the Zen 5 architecture on a 4 nm process, while the 4564P uses Zen 4 on a 5 nm process. The 4585PX has 80 KB of L1 cache per core and 128 MB of L3 cache; the 4564P has 64 KB of L1 per core and 64 MB of L3. Transistor counts are 16,630 million for the 4585PX versus 13,140 million for the 4564P, with die sizes of 2x 70.6 mm² and 2x 71 mm², respectively. Memory bandwidth is higher on the 4585PX at 89.6 GB/s compared to 83.2 GB/s. The 4564P offers more PCIe lanes (28) than the 4585PX (24). Base clocks differ slightly, with the 4564P at 4.50 GHz and the 4585PX at 4.30 GHz, though both boost to 5.70 GHz. The release dates also differ: the 4585PX launched on 2025-05-12, and the 4564P launched on 2024-05-20. Both have a launch MSRP of $699.

FAQ

Q: Which CPU has more L3 cache?

A: The AMD EPYC 4585PX has 128 MB of L3 cache, while the AMD EPYC 4564P has 64 MB (shared).

Q: Is the AMD EPYC 4585PX always faster than the 4564P?

A: No. The 4564P wins the PassMark data encryption test (50,708 vs 47,224) and the random string sorting test (103,202 vs 95,210).

Q: What is the performance gap in Cinebench R23 multi-core?

A: The 4585PX scores 60,451, which is 8.7% higher than the 4564P’s score of 55,592.

Q: Do both CPUs support ECC memory?

A: Yes, both the AMD EPYC 4585PX and the AMD EPYC 4564P support ECC memory with DDR5.

Q: How do these chips compare to their nearest rivals in average score?

A: The 4585PX has an average score of 99,324, which is 1.7% lower than the AMD Ryzen Threadripper PRO 9955WX (101,041) and 2.9% lower than the AMD EPYC 7513 (102,244). The 4564P’s average score of 95,183 is 0.5% lower than the AMD EPYC 9175F (95,615) and 1.5% higher than the Intel Xeon 6520P (93,786).

Q: Which CPU has a higher base clock?

A: The AMD EPYC 4564P has a higher base clock at 4.50 GHz, compared to the 4585PX’s 4.30 GHz. Both have the same 5.70 GHz boost clock.

Where Each One Wins

AMD EPYC 4585PX: This is the winner for general server and workstation computing. It wins all Cinebench rendering tests, integer and floating-point math, prime number finding, physics simulation, and data compression. If your workload involves compiling code, running virtual machines, 3D rendering, or any CPU-bound scientific calculation, the 4585PX is the clear choice. Its 128 MB L3 cache and higher memory bandwidth (89.6 GB/s) give it an edge in data-heavy multi-threaded tasks. The 49.9% lead in prime numbers and 94.9% lead in physics make it especially strong for financial risk analysis and engineering simulations.

AMD EPYC 4564P: This chip is specifically better for two niche workloads. First, data encryption: the 4564P scores 50,708 versus 47,224, a 6.9% advantage. This makes it a better fit for security appliances, VPN gateways, or database servers that do heavy TLS/SSL termination. Second, random string sorting: the 4564P’s 103,202 score beats the 4585PX’s 95,210 by 7.7%. This could benefit applications that do lots of text processing, log analysis, or key-value store operations. Additionally, the 4564P’s higher base clock of 4.50 GHz and extra PCIe lanes (28 vs 24) could be relevant for lightly threaded tasks or systems with many Gen 5 NVMe drives, though the benchmark data does not directly support a performance claim there. For everything else, the 4585PX is the superior processor.