Optimize Your Cloud OPEX with Google Cloud C3D Virtual Machines

Choosing the Right VM Matters

When migrating to a new cloud computing provider, the specific virtual machines (VMs) you choose have a major impact on operating costs. Choosing the lowest priced VMs that meet the minimum performance can end up costing more than VMs running on the latest generation. The C3D VM series, built with AMD EPYC™ 9004 processors, provides a 47% performance uplift on average across a selection of popular workloads when compared to the previous generation.1

Higher Performance Can Equal Lower Costs 

When considering modernization, performance and cost can also vary based on the processor you choose. Compared to the Intel-based N2 machine series, C3D can provide significant savings on cloud OPEX along with a notable performance uplift.

Flexible & Optimized Google Cloud VMs with EPYC™

Flexibility is critical when selecting a cloud VM or instance, and Google Cloud offers a variety of VMs powered by AMD EPYC™ processors that are designed for specific use cases, such as memory-intensive, compute-intensive, and HPC applications. See the full list of EPYC Google Cloud VMs.

How Much Can C3D Lower Your Cloud Computing Costs?

Cloud OPEX Savings when choosing C3D Virtual Machines

Choosing C3D VMs can mean significant savings on operating costs. When adopting C3D, users can see a 37% performance uplift and save 31% Cloud OPEX on average across five application benchmarks when compared to the prior generation Intel-based N2 VMs.2

Savings and performance increases can be found across many of your workloads running on Google Cloud C3D VMs.

Web/App Tier
NGINX Server-side Java FFmpeg
~1.2x Requests/Sec ~1.7x Max Ops, multi-instance ~1.4x Frames/Sec
23% Savings 45% Savings 32% Savings
Data Tier
MySQL Redis
~1.2x Transactions/Min ~1.4x Requests/Sec
22% Savings 32% Savings

The same workloads show similar performance and savings advantages when compared to the newest Intel-based C3 VMs. Across many workloads, C3D VMs provide an 16% average performance uplift and 24% average lower Cloud OPEX.3

Web/App Tier
NGINX Server-side Java4 FFmpeg
~1.00x Requests/Sec ~1.16x Max Ops, multi-instance ~1.23x Frames/Sec
13% Savings 25% Savings 30% Savings
Data Tier
MySQL Redis
~1.05 Transactions/Min ~1.36 Requests/Sec
17% Savings 36% Savings

Confidential VMs Powered by AMD Infinity Guard5

With AMD Infinity Guard, AMD EPYC™ processors help safeguard privacy and integrity by encrypting each virtual machine with unique encryption keys known only to the processor.

Learn more about how Google’s Confidential VMs and Confidential GKE Nodes enable AMD Secure Encrypted Virtualization to help deliver confidential computing for the cloud.

Google Cloud VMs Powered by AMD EPYC Processors

Turn your cloud environment into a competitive advantage with AMD and Google Cloud – setting superior standards for performance and scalability for your most demanding workloads.

Find the right instance to fit your workload needs.

General Purpose Computing

C3D brings the enhanced performance and operating efficiency of 4th Gen AMD EPYC processors and improved performance consistency to the Google Cloud general purpose virtual machine family. T2D offers outstanding price-performance for scale-out workloads with single-threaded processing, and N2D offers customizable VM shapes and a more cost-efficient option for customers.

VM Family

VMs

Specifications

Generation

Key Workloads

  • C3D
  • Consistently High Performance
  • C3D standard (4:1 GiB to vCPU)
  • C3D highcpu (2:1 GiB to vCPU)
  • C3D highmem (8:1 GiB to vCPU)
  • Up to 3.7 GHz*
  • Up to 360 vCPUs
  • Up to 200 Gbps network bandwidth

4th Gen EPYC

  • High traffic web, app, and ad servers
  • Medium-to-large databases
  • Game servers
  • Media streaming and transcoding
  • Data analytics
  • CPU-based inferencing
  • Tau T2D
  • Scale-Out Optimized
  • Tau T2D (4:1 GiB to vCPU)
  • Up to 3.5 GHz*
  • Up to 60 vCPUs
  • Up to 32 Gbps network bandwidth

3rd Gen EPYC

  • Containerized microservices
  • Compression/decompression
  • Image processing
  • Data-logging processing
  • Large-scale Java applications
  • N2D
  • Cost Efficient
  • N2D standard (4:1 GiB to vCPU)
  • N2D highcpu (2:1 GiB to vCPU)
  • N2D highmem (8:1 GiB to vCPU)
  • Custom machine types
  • Up to 3.5 GHz*
  • Up to 224 vCPUs
  • Up to 100 Gbps network bandwidth
  • 3rd Gen EPYC
  • 2nd Gen EPYC
  • Low-medium traffic web and app servers
  • Small-medium databases
  • Business intelligence applications
  • Desktop virtualization
  • CRM applications
  • Dev/test environment

Compute Intensive Computing

C2D is Google Cloud’s AMD based VM series for compute intensive and high-performance computing (HPC) workloads. The C2D series offers the highest per-core performance and the largest available last-level cache (LLC) cache per core among AMD based GC instances. 

VM Family

VMs

Specifications

Generation

Key Workloads

  • C2D
  • Performance per Core
  • C2D standard (4:1 GiB to vCPU)
  • C2D highcpu (2:1 GiB to vCPU)
  • C2D highmem (8:1 GiB to vCPU)
  • Up to 3.5 GHz
  • Up to 112 vCPUs
  • Up to 100 Gbps network bandwidth
  • 3rd Gen EPYC
  • High-performance computing (HPC)
  • EDA/FEA/CFD
  • Modeling and simulation
  • Media transcoding
  • High-performance game servers
  • AI/ML

Confidential Computing

Google Cloud Confidential VMs powered by AMD EPYC™ CPUs encrypt code and data in memory during processing to provide encryption in use. Confidential computing is available on Google Compute Engine (GCE) instances and Google Kubernetes Engine (GCE) nodes. Implementing Confidential VMs is seamless—all GCP workloads you run in N2D and C2D VMs today can run as Confidential VMs instances powered by AMD with the simple click of a button.

VM Family

VMs

Specifications

Generation

  • Key Workloads
  • N2D VMs and GKE nodes
  • N2D standard (4:1 GiB to vCPU)
  • N2D highcpu (2:1 GiB to vCPU)
  • N2D highmem (8:1 GiB to vCPU)
  • Up to 3.5 GHz*
  • Up to 224 vCPUs
  • Up to 100 Gbps network bandwidth
  • 3rd Gen EPYC
  • 2nd Gen EPYC
  • Low-medium traffic web and app servers
  • Small-medium databases
  • Business intelligence applications
  • Desktop virtualization
  • CRM applications
  • Dev/test environment
  • C2D VMs and GKE nodes
  • C2D standard (4:1 GiB to vCPU)
  • C2D highcpu (2:1 GiB to vCPU)
  • C2D highmem (8:1 GiB to vCPU)
  • Up to 3.5 GHz*
  • Up to 112 vCPUs
  • Up to 100 Gbps network bandwidth

3rd Gen EPYC

  • High-performance computing (HPC)
  • EDA/FEA/CFD
  • Modeling and simulation
  • Media transcoding
  • High-performance game servers
  • AI/ML

Google Cloud Case Studies

Newsletter and Request Contact

Subscribe to Data Center Insights from AMD

Request Contact from an AMD EPYC and Google Cloud Sales Expert

Footnotes

* Max boost for AMD EPYC processors is the maximum frequency achievable by any single core on the processor under normal operating conditions for server systems. EPYC-18 

  1. SP5C-005: MySQL, Redis, NGINX, server-side Java multi-instances, and FFmpeg comparison of Google Cloud C3D-standard 16 vCPU to N2D-standard 16 vCPU based on AMD testing on 10/5/23. Configurations both with 64GB running Ubuntu 22.04.3  LTS. Uplifts: MySQL 8.0.28 HammerDB 4.2 TPROC-C (+39% avg), Redis 7.2 get/set rps (+62% avg), NGINX 1.1.9-2 WRK 4.2 ops/sec (+73% avg), server-side Java® multi instances max-OPS (+41% avg) and FFmpeg 4.4.2.0 Ubuntu 22.04.1 h264-vp9, raw_h264, raw_vp9, vp9_h264 at 1080p frames/hr (+21% avg).
    Cloud performance results presented are based on the test date in the configuration. Results may vary due to changes to the underlying configuration, and other conditions such as the placement of the VM and its resources, optimizations by the cloud service provider, accessed cloud regions, co-tenants, and the types of other workloads exercised at the same time on the system.
  2. SP5C-006: MySQL™, Redis®, NGINX®, server-side Java multi-instances, and FFmpeg™ comparison of Google Cloud C3D-standard 16 vCPU to N2-standard 16 vCPU based on AMD testing on 11/02/23. Cloud OPEX savings calculated based on on-demand pricing at https://cloud.google.com/compute/vm-instance-pricing for us-central1 (Iowa) as of 11/01/2023. Configurations both with 64GB running Ubuntu 22.04.3 LTS.
    Comparisons
    MySQL 8.0.28 HammerDB 4.2 TPROC-C (~1.2x tpm, 22% Cloud OPEX savings),
    Redis 7.2 get/set: (~1.4x rps, 32% Cloud OPEX savings),
    NGINX 1.1.9-2 WRK 4.2: (~1.2x ops/sec, 23% Cloud OPEX savings),
    server-side Java® multi instances max-OPS (~1.7x OPS, 45% Cloud OPEX savings)
    FFmpeg 4.4.2.0 Ubuntu 22.04.1 h264-vp9, raw_h264, raw_vp9, vp9_h264 at 1080p (~1.4x frames/hr, 32% Cloud OPEX savings).
    Cloud performance results presented are based on the test date in the configuration. Results may vary due to changes to the underlying configuration, and other conditions such as the placement of the VM and its resources, optimizations by the cloud service provider, accessed cloud regions, co-tenants, and the types of other workloads exercised at the same time on the system.
  3. SP5C-007: MySQL™, Redis®, NGINX®, and FFmpeg™ comparison of Google Cloud C3D-standard 8 vCPU to C3-standard 8 vCPU based on AMD testing on 10/05/23. OpEx savings calculated based on on-demand pricing at https://cloud.google.com/compute/vm-instance-pricing for us-central1 (Iowa) as of 11/01/2023. Configurations all with 4GB/vCPU running Ubuntu 22.04.3 LTS.
    Comparisons:
    MySQL 8.0.28 HammerDB 4.2 TPROC-C (~1.05x tpm, 17% Cloud OpEx savings),
    Redis 7.2 get/set: (~1.36x rps, 36% Cloud OpEx savings),
    NGINX 1.1.9-2 WRK 4.2: (Comparable ops/sec, 13% Cloud OpEx savings),
    FFmpeg 4.4.2.0 Ubuntu 22.04.1 h264-vp9, raw_h264, raw_vp9, vp9_h264 at 1080p (~1.23x frames/hr, 30% Cloud OpEx savings).
    Pricing source: https://cloud.google.com/compute/all-pricing
    Cloud performance results presented are based on the test date in the configuration. Results may vary due to changes to the underlying configuration, and other conditions such as the placement of the VM and its resources, optimizations by the cloud service provider, accessed cloud regions, co-tenants, and the types of other workloads exercised at the same time on the system.
  4. SP5C-008A: Server-side Java® multi-instance comparison of Google Cloud C3D-standard 16 vCPU to C3-standard 22 vCPU (on a per-vCPU basis) based on AMD testing on 11/02/23. Cloud OpEx savings calculated based on on-demand pricing at https://cloud.google.com/compute/vm-instance-pricing for us-central1 (Iowa) as of 11/01/2023. Configurations all with 4GB/vCPU running Ubuntu 22.04.3 LTS.
    https://cloud.google.com/compute/all-pricing
    Cloud performance results presented are based on the test date in the configuration. Results may vary due to changes to the underlying configuration, and other conditions such as the placement of the VM and its resources, optimizations by the cloud service provider, accessed cloud regions, co-tenants, and the types of other workloads exercised at the same time on the system.
  5. GD-183: AMD Infinity Guard features vary by EPYC™ Processor generations. Infinity Guard security features must be enabled by server OEMs and/or Cloud Service Providers to operate. Check with your OEM or provider to confirm support of these features. Learn more about Infinity Guard at https://www.amd.com/en/technologies/infinity-guard.