AMD EPYC 9745 (Turin Dense, 128-core Zen 5c)

The AMD EPYC 9745 is a 128-core, 256-thread processor belonging to the fifth-generation EPYC 'Turin Dense' family, built on the Zen 5c compact core architecture using TSMC's 3nm process node. The 'c' suffix denotes a physically smaller Zen 5c die optimized for maximum core density within a constrained thermal design power envelope, making it ideal for operators who need to maximize active threads per rack unit or per watt rather than prioritizing single-threaded peak frequency. Launched in October 2024, the 9745 is manufactured on the SP5 platform, maintaining full socket compatibility with existing EPYC Genoa and Bergamo SP5 motherboards, protecting infrastructure investments while delivering next-generation IPC improvements.

The Zen 5c core architecture retains full 64-bit x86 capability including AVX-512 support and improved branch prediction, delivering meaningful IPC gains over the prior Zen 4c generation. With 256MB of L3 cache distributed across the chiplet complex dies and an eight-channel DDR5-6400 memory controller supporting up to 6TB of registered ECC memory, the 9745 provides substantial memory bandwidth and capacity headroom for in-memory analytics, large-scale containerized microservices, and cloud-native virtualization stacks. The processor supports PCIe 5.0 across 128 lanes, enabling high-throughput connectivity to NVMe storage arrays and high-speed networking adapters simultaneously.

The EPYC 9745 is positioned as a direct answer to the growing demand for energy-efficient scale-out server infrastructure. Its 320W TDP places it firmly in the efficiency-first segment, achieving exceptionally high thread counts at comparatively modest power draw, making it attractive for hyperscale colocation environments, cloud service providers, and enterprises managing strict PUE targets. The dense core count combined with NUMA topology awareness and hardware-accelerated virtualization extensions makes it equally adept at hosting hundreds of lightweight virtual machines or containerized workloads on a single two-socket system.

Ideal for

• Hyperscale cloud and colocation operators deploying high-density virtual machine hosting requiring maximum vCPU count per kilowatt of consumed power
• Web-scale content delivery and application tier servers running thousands of concurrent microservices or containerized workloads within strict rack power budgets
• Large-scale Kubernetes and OpenShift cluster nodes demanding high core counts with efficient context switching across hundreds of simultaneous pods
• Distributed big data processing clusters using Apache Spark or Hadoop where parallel thread throughput directly accelerates job completion rates
• HPC environments running embarrassingly parallel workloads such as genomics pipelines, weather modeling ensembles, and Monte Carlo financial simulations
• Software-defined networking and telecom NFV platforms requiring dense compute to handle packet processing, session management, and real-time analytics concurrently

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.