Procesor AMD EPYC 9365 3,4 GHz 192 MB L3 Socket SP5

The AMD EPYC 9365 is a server processor from the 5th Gen EPYC 9005 family, built for data center, cloud and AI workloads. Operating at a base frequency of 3.4 GHz with a boost up to 4.3 GHz, the chip uses Socket SP5, offers a large 192 MB L3 cache and supports dodeca-channel DDR5 memory with up to 576 GB/s memory bandwidth. It is designed to deliver high per-server throughput and energy-efficient performance for enterprise and cloud deployments.

Advantages

The EPYC 9365 provides strong integer and AI inference performance on CPU-only servers and can be used as a host CPU for GPU-accelerated systems. Its high core count and large cache improve virtualization density and server consolidation, reducing rack footprint, power consumption and total cost of ownership compared with some previous-generation platforms. The processor benefits from AMD’s Zen 5 architecture and chiplet design to balance performance, memory capacity and energy efficiency for demanding workloads.

Suitable for

• Virtualized cloud environments and server consolidation
• CPU-based AI inference for language models up to mid-size (e.g., ≤13B params)
• GPU-host systems for accelerated AI and ML workloads
• High-throughput enterprise and database servers

Key specifications

• Model: AMD EPYC 9365
• Base frequency: 3.4 GHz, Boost frequency: up to 4.3 GHz
• L3 cache: 192 MB
• Cores: 36
• Socket: SP5
• Memory: DDR5, dodeca-channel, up to 576 GB/s bandwidth
• Integrated graphics: none

How to use

Install the EPYC 9365 into a compatible Socket SP5 server motherboard following the server vendor’s hardware installation guide. Populate supported DDR5 DIMMs according to the motherboard’s channel recommendations to achieve the stated dodeca-channel bandwidth. Use appropriate cooling and power delivery for data center operation, and update firmware/BIOS and server firmware to the vendor-recommended versions before deploying workloads. Configure the operating system and hypervisor to recognize available cores and memory for optimal virtualization or containerized deployments.

Recommendations

For best results, pair the EPYC 9365 with server platforms validated for DDR5 dodeca-channel memory and adequate cooling. When running mixed workloads, monitor power and thermal metrics and tune BIOS/firmware settings to balance performance and efficiency. Use the processor as a host CPU in GPU-accelerated nodes when AI models exceed CPU-only performance limits to improve overall throughput and ROI.