[Features] Overview of new generation CPU “AMD EPYC” for cloud/server [Semiconductor]
This is Ohara from the technical sales department.
we will describe the characteristics of
AMD (Advanced Micro Devices) CPUs used in various clouds / servers such as AWS and GCP *Information as of September 2021.
AMD EPYC Overview
● World's highest performance x86 CPU
● Improved performance with the latest "Zen3" core
● AMD EPYC™ 7003 PROCESSORS (development code name "MILAN")
● Drop-in compatibility with the Rome platform (BIOS update required)
● Security features Expansion
● 4ch / 6ch / 8ch channel memory support
● Improved memory performance (x86 maximum capacity L3 cache up to 32MB per Core)
Characteristics of each cloud
cloud name | Features | overview | Main target instance |
● AWS | - Pricing up to 10% to 45% lower than comparable EC2 instances. - Seamless workload migration and flexibility. |
・EC2 G4ad instances equipped with AMD Radeon Pro V520 GPU and 2nd generation AMDEPYC processors ・Up to 45% lower cost than Nvidia G4dn instances for graphics-intensive applications such as virtual workstations/game streaming/graphics rendering. |
・M5a / M5ad (general purpose) ・R5a / R5ad (memory optimization) ・T3a (burstable general purpose) ・C5a (computing optimization) ・G4ad (graphic intensive) ・3rd generation EPYC core Amazon EC2 instance * Coming soon |
● Microsoft Azure | ・Supports a wide range of applications/workloads. ・Up to 80K cores equipped with HBv2 to realize HPC in the cloud. |
- Azure D32as v4 virtual machines with 2nd generation AMDEPYC processors cost up to 53% less than Azure D32ds v4. | ・Dav4 (general purpose) ・Eav4 (memory optimization) ・HBv2 (HPC optimization) ・HBv3 (HPC optimization) *Coming soon ・Lsv2 (storage optimization) ・NVv4 (GPU + CPU optimization) |
● Google Cloud | - AMD-based N2D instances deliver up to 13% savings compared to comparable Intel N-series instances. - Up to 39% performance improvement on Coremark benchmarks compared to equivalent Intel N1 instances. |
AMD-based N2D VMs improve performance for data management workloads that take advantage of AMD's higher memory bandwidth and higher per-system throughput (available on larger VM selections), with up to 224 vCPUs Google's largest general-purpose VM. - Benchmarks benefit from AMD N2D's improved memory bandwidth, such as STREAM, with up to 2.5x improvements in performance of 20-30% across many workload types. |
・N2D Standard (general purpose) ・N2D HighMEM (memory optimization) ・N2D HighCPU (HPC optimization) ・Confidential VMs ・N2D Milan * Coming soon ・C2D for HPC * Coming soon ・Tau T2D * Coming soon |
● Oracle Cloud | - | - | ・E2 Standard (general purpose) ・E3 Standard (high performance) ・Elastic Block Storage V2 ・Standard E4 Flex Sizing (general purpose) *Coming soon |
● Alibaba Cloud | - | - | ・g6a (general purpose) ・c6a (computing optimization) ・r6a (memory optimization) |
Reference materials
● AWS: AMD EPYC™ & AWS EC2 instances
● Microsoft Azure: AMD EPYC™ & Microsoft Azure Virtual Machine
● Google Cloud: AMD EPYC™ & Google Cloud Platform instances
summary
Recently, AMD CPU instances have been released in the cloud, so when you develop services/content in the cloud, why not consider choosing one depending on the specifications of the application running on the cloud?