AMD EPYC 9575F processor – 3,3 GHz, 64 / 128 cores / threads, 256 MB L3, Socket SP5 (100-000001554)

The AMD EPYC 9575F is a server processor from the 5th Gen AMD EPYC 9005 family, operating at a base frequency of 3.3 GHz with boost up to 5 GHz and featuring a 256 MB L3 cache. Designed for Socket SP5 platforms, it supports dodeca-channel DDR5 memory with up to 576 GB/s memory bandwidth and provides 64 "Zen 5" cores for high-density, data-center and cloud workloads where integer performance, energy efficiency, and GPU-hosting capability matter.

Advantages

The EPYC 9575F delivers high integer and AI inference performance on CPU-only servers and can also serve as a high-performance host CPU for GPU-accelerated systems. Its large L3 cache, high core count and wide DDR5 memory interface enable increased per-server throughput, improved virtual CPU density and competitive energy efficiency and total cost of ownership for enterprise and cloud deployments.

Suitable for

• Cloud infrastructure and virtualized servers (high vCPU density)
• AI inference on CPU-only or GPU-accelerated nodes (up to medium-sized models)
• High-performance computing and data analytics
• GPU-host systems for model training and inference acceleration
• Enterprise databases and large-scale transactional workloads

Key specifications

• Processor family: AMD EPYC 9005 (5th Gen)
• Model: EPYC 9575F
• Base frequency: 3.3 GHz; Boost frequency: up to 5 GHz
• Cores: 64 (Zen 5 cores)
• L3 cache: 256 MB
• Memory: Dodeca-channel DDR5-SDRAM, up to 576 GB/s memory bandwidth
• Socket: SP5
• On-board graphics: Not available

How to use

Install the EPYC 9575F into a compatible Socket SP5 server motherboard following the motherboard and chassis manufacturer's instructions. Populate DDR5 memory modules according to the platform's dodeca-channel recommendations to achieve optimal bandwidth. For GPU-accelerated setups, pair the processor with supported server GPUs and configure the system BIOS/firmware to enable required PCIe lanes and IOMMU settings. Use validated server operating systems and drivers from your hardware vendor to ensure stability and performance.

Recommendations

For best performance in cloud and AI workloads, deploy the EPYC 9575F in systems with fully populated DDR5 channels, adequate cooling and power delivery, and firmware tuned for large-core-count CPUs. When using GPUs, balance GPU count and PCIe/CPU resources to avoid bottlenecks and consult vendor guidance for optimized ML frameworks and inference/throughput tuning.