AMD EPYC 4564P

AMD EPYC 4564P is a 16-core, 32-thread Zen 4 processor for the AMD Socket AM5 platform, built on TSMC's 5 nm process with a 170 W TDP. Its benchmark data places it in the 98th percentile of all CPUs tested, with an average benchmark score of 96,263, making it a top-tier part for single-socket server and workstation workloads. The data shows a chip that leads its immediate rivals in multi-threaded throughput while offering single-thread performance that competes with the best desktop and workstation parts.

Benchmark Performance

The EPYC 4564P delivers a commanding multi-core performance profile, as evidenced by a Cinebench R23 multi-core score of 55,592. This result positions the chip 2.8% ahead of the Intel Core Ultra 9 285K, which posts an average benchmark score of 93,672, and 4.2% ahead of the AMD EPYC 9175F with an average score of 92,399. In the Passmark multithread test, the 4564P scores 65,403, reinforcing its strength in heavily threaded workloads such as rendering, simulation, and code compilation. The data indicates a clear advantage over the AMD Ryzen Threadripper PRO 5965WX, which trails by 2.3% in average score (98,504 vs. 96,263), though the Threadripper part remains a close competitor in overall throughput.

Single-core performance is equally robust, with a Cinebench R23 single-core score of 7,848 and a Passmark single-thread score of 4,371. These numbers show that the EPYC 4564P does not sacrifice responsiveness for core count; it outperforms the EPYC 4565P by a marginal 0.5% in average score, indicating nearly identical per-core efficiency. The Cinebench R20 results—23,348 multi-core and 3,296 single-core—further confirm a balanced architecture. In integer math, the chip scores 230,740 on Passmark, while floating-point math reaches 142,391, and data compression hits 869,555, all pointing to strong computational breadth. The 98th percentile ranking underscores that this processor sits above the vast majority of CPUs in the database, with only a handful of higher-end parts (like the Threadripper PRO 5965WX) edging it out in aggregate.

Platform and Compatibility

The EPYC 4564P uses the AMD Socket AM5, a platform traditionally associated with consumer Ryzen parts, but here repurposed for EPYC server and workstation duties. It supports DDR5 memory via a dual-channel interface, delivering a memory bandwidth of 83.2 GB/s. ECC memory is supported, which is critical for data integrity in server environments where bit-flip errors cannot be tolerated. The integrated memory controller handles dual-channel configurations, and while this is narrower than the quad-channel setups found on higher-end EPYC or Threadripper platforms, the 83.2 GB/s figure is sufficient for many enterprise workloads, especially those that rely on high core counts rather than memory bandwidth saturation.

PCIe connectivity is provided by 28 Gen 5 lanes from the CPU, enabling high-speed attachment of GPUs, NVMe storage, and networking cards. This lane count supports a single flagship GPU or multiple storage devices without bottlenecking. The integrated Radeon Graphics means the processor does not require a discrete GPU for basic display output, a useful feature for headless servers or low-power workstations. The part is unlocked for multipliers? No, the multiplier is locked, so overclocking is not supported, but the base clock of 4.50 GHz and boost clock of 5.70 GHz are already high for a server chip, reducing the need for manual tuning. The launch MSRP is $699, and the production status is active, indicating ongoing availability. Upgrade path considerations: since it uses AM5, users can move between EPYC 4004 series parts, but the platform is not compatible with the larger EPYC 9004 series (which uses SP5), so scaling up requires a platform change.

Single-Thread vs Multi-Thread Behavior

The split between single-thread and multi-thread scores reveals a processor that excels at both, but with a tilt toward parallel workloads. The Cinebench R23 single-core score of 7,848 is exceptional, placing it in the same league as the Intel Core Ultra 9 285K, which has a similar average score (93,672) and is only 2.8% behind in aggregate. In contrast, the multi-core score of 55,592 shows a 16-core part that scales well, with a multi-to-single ratio of roughly 7.1x, indicating efficient utilization of all threads. The Passmark data reinforces this: the multithread score of 65,403 is 15x higher than the single-thread score of 4,371, which is expected for a 16-core/32-thread design.

For real workloads, this means the EPYC 4564P handles single-threaded tasks like database queries, legacy software, and lightly threaded gaming engines with the same agility as a high-end desktop CPU. The boost clock of 5.70 GHz is a key contributor here, allowing short bursts of high-frequency operation. Meanwhile, multi-threaded applications—such as video encoding, 3D rendering, scientific computing, and virtualization—see near-linear scaling across the 32 threads. The Passmark physics score of 3,554 and extended instructions score of 63,663 suggest that the chip is well-suited for simulation and cryptographic workloads, where both branch prediction and SIMD efficiency matter. The data indicates no significant single-thread bottleneck, so users can run mixed workloads without the processor becoming a weak link.

Who Should Consider It

The EPYC 4564P targets server and workstation users who need high core counts without stepping up to a more expensive HEDT or dual-socket platform. For content creators, the Cinebench R23 multi-core score of 55,592 means 4K video exports, complex After Effects comps, and Blender renders will complete quickly, while the single-core score of 7,848 ensures UI responsiveness and fast preview scrubbing. The Passmark integer math score of 230,740 is relevant for software developers compiling large codebases, where the 32 threads can be fully utilized by build systems like Make or Ninja.

For office and enterprise use, the chip is overkill for basic productivity, but it shines in virtualized environments—running multiple VMs or containers—where the 32 threads provide ample headroom for concurrent workloads. The ECC memory support and 83.2 GB/s bandwidth make it viable for database servers, file servers, and financial modeling. Gamers, however, should look elsewhere: while the single-thread performance is strong, the 170 W TDP and server-oriented design are not optimized for gaming workloads, and the integrated Radeon Graphics is not a substitute for a discrete GPU in modern titles. The 98th percentile ranking means it outperforms 98% of CPUs in the database, so any workload that is CPU-bound will benefit, but the platform cost (AM5 server boards) is justified only for professional use.

Power and Thermals

With a TDP of 170 W, the EPYC 4564P sits in a power class that requires a substantial cooling solution. This is not a chip for a stock low-profile cooler; the data suggests it needs a high-end air cooler or a liquid cooler to maintain boost clocks under sustained load. The base clock of 4.50 GHz and boost clock of 5.70 GHz are aggressive for a 16-core part, and the 5 nm process from TSMC helps efficiency, but the 170 W envelope is comparable to many desktop flagship CPUs. In practice, users should plan for a cooling solution capable of dissipating 170 W continuously, especially in a dense server chassis where airflow is limited. The dual-die design (2x 71 mm²) with 13,140 million transistors means heat is spread across two chiplets, which can help with thermal density but still requires adequate heatsink contact. The data does not include thermal throttling behavior, but the high boost clock suggests that cooling directly impacts sustained performance—a well-cooled system will hold 5.70 GHz on all cores, while a marginal cooler may see clock reductions. For server deployments, a 1U or 2U chassis with high-static-pressure fans is recommended, while workstation users can opt for a tower cooler with a 140mm fan.

FAQ

Q: What is the socket type for the AMD EPYC 4564P?

A: It uses AMD Socket AM5, which is unusual for a server chip but enables compatibility with AM5 boards that support the EPYC 4004 series.

Q: Does the EPYC 4564P support ECC memory?

A: Yes, ECC memory is supported, which is essential for error-free operation in server and workstation environments.

Q: How many PCIe lanes does the CPU provide?

A: It provides 28 lanes of PCIe Gen 5 from the CPU, sufficient for a GPU, NVMe drives, and network adapters.

Q: What is the boost clock speed?

A: The boost clock is 5.70 GHz, while the base clock is 4.50 GHz, giving it high single-thread performance.

Q: Does the processor have integrated graphics?

A: Yes, it includes Radeon Graphics, so a discrete GPU is not required for basic display output.

Q: How does it compare to the AMD EPYC 4565P?

A: The EPYC 4564P is 0.5% faster in average benchmark score (96,263 vs. 95,820), indicating nearly identical performance.

How It Compares

AMD EPYC 4565P: The closest rival, the 4565P, trails by just 0.5% in average score (95,820 vs. 96,263). The data shows these two parts are near-identical, with the 4564P holding a slight edge in aggregate benchmarks. Users choosing between them should base the decision on price and availability rather than performance, as the difference is within noise.

AMD Ryzen Threadripper PRO 5965WX: This Threadripper part leads the 4564P by 2.3% in average score (98,504 vs. 96,263). The Threadripper PRO likely benefits from more cores or higher memory bandwidth, but the 4564P offers a lower TDP and AM5 platform compatibility, making it a more accessible option for single-socket builds.

Intel Core Ultra 9 285K: The Intel flagship is 2.8% behind the 4564P in average score (93,672 vs. 96,263). While the Core Ultra 9 is a strong desktop part, the EPYC 4564P outperforms it in multi-threaded workloads like Cinebench R23 (55,592), making it the better choice for server or workstation tasks that exploit all threads.

AMD EPYC 9175F: The 9175F trails by 4.2% in average score (92,399 vs. 96,263). This gap is more significant, and the 4564P’s higher boost clock (5.70 GHz) likely contributes to its advantage in both single-thread and multi-thread benchmarks. The 4564P is the stronger performer, though the 9175F may offer other platform features not captured in these metrics.