We chose Google Cloud (“GCP”) as the foundation for Kinsta’s hosting services so that our customers would have the fastest servers connected to the fastest networks.

When GCP unveiled its high-performance C2 virtual machines in 2019, Kinsta moved quickly to put that technology behind customer websites everywhere it was available. Four years later, the next generation of the C family of virtual machines is rolling out to Google Cloud data centers, and Kinsta has chosen the most powerful — the C3D — to help push performance to the next level.

At the heart of advancements found in Google’s C3D VMs are the 4th Gen AMD EPYC™ CPUs (formerly codenamed “Genoa”) and a custom Infrastructure Processing Unit (IPU) to supercharge networking performance and data I/O.

In this article, we explore in depth these technological advancements and the strategic benefits of migrating to Google’s C3D machines at Kinsta.

The introduction of C3D machines on Google Cloud

The C3D machines on the Google Cloud represent a significant leap in cloud computing capabilities, chiefly due to the integration of 4th Gen AMD EPYC™ CPUs and Google’s custom IPUs. The 4th Gen AMD EPYC™ CPUs are renowned for their high performance, with a maximum frequency of 3.7 GHz.

This allows C3D machines to handle demanding computational workloads efficiently — particularly important for applications like high-traffic web hosting and databases to CPU-based machine learning training and inference.

Google’s custom IPU complements the AMD EPYC™ CPUs to boost the efficiency and performance of C3D machines. These application-specific integrated circuits (ASICs) help accelerate system-level infrastructure resources, including networking and security functions.

By offloading these tasks to the IPU, the C3D machines can offer:

  1. Improved computing performance.
  2. Enhanced security features.
  3. Better isolation capabilities.

As a result, the hardware architecture is optimized and can deliver reliable, consistent performance across workloads.

Google Cloud’s C3D series is optimized for the underlying hardware architecture to ensure customers fully benefit from these technological advancements. With the integration of AMD EPYC™ CPUs and Google’s IPU, C3D machines can support up to 360 vCPUs and 2,880 GB of DDR5 memory. Also, Standard and Premium Tier networking configurations provide up to 100 Gbps and 200 Gbps of bandwidth, respectively.

These qualities make C3D exceptionally well-suited for applications that require high computational power and fast data processing capabilities, such as media streaming, transcoding, and data analytics.

The introduction of C3D machines underscores Google’s commitment to providing state-of-the-art infrastructure solutions. By using the latest advancements in processor technology and custom hardware accelerators, Google offers a platform that meets the current demands of cloud computing and sets a new benchmark for performance, security, and scalability.

At Kinsta, complex WordPress sites heavily benefit from C3D’s power. Your boosted websites will perform faster than ever before while benefiting from security, reliability, uptime, and all the other vital features we already offer, such as:

  1. Friendly UI.
  2. Automatic and manual backups.
  3. Staging environments.
  4. Content delivery networks (CDNs).
  5. Edge caching.
  6. SSL/TLS.
  7. Malware detection.
  8. Distributed denial of service (DDoS) protection and mitigation.
  9. And much more.

Advantages of C3D machines for high-performance computing

With the C3D advancements outlined above, as well as integrated 4th Gen AMD EPYC™ CPUs, Kinsta becomes particularly advantageous for high-performance computing (HPC) applications.

1. Faster individual cores

Boasting a maximum frequency of 3.7 GHz, these processors boost the speed of individual cores. And since core speed significantly impacts the overall completion time for computational tasks, like simulations, modeling, and data analysis, HPC applications using these processors experience improved performance and speed.

2. Compute-intensive tasks are easier to handle

The ability of C3D machines to handle demanding computational workloads efficiently aligns perfectly with the needs of environments that often run complex, compute-intensive tasks. This substantial performance boost over previous generations leads to faster and more efficient applications, quicker insights, and the ability to tackle more sophisticated computational problems.

C3D machines’ IPUs accelerate system-level infrastructure resources (such as networking and security functions), improve compute performance, bolster security, and support isolation. This is crucial for HPC workloads requiring both computational robustness, high data integrity and security, and provides consistent high performance for most general-purpose computing workloads.

3. Greater capability to scale

Additionally, the support for up to 360 vCPUs, 2,880 GB of DDR5 memory, and high-bandwidth networking options makes C3D machines ideal for the parallel processing requirements inherent in HPC tasks. The capability to scale computational resources effectively ensures that HPC applications can access the computational power necessary to explore new frontiers in research and development.

Networking performance, isolation, and security with IPUs

IPUs are a pivotal innovation in cloud computing. By offloading specific infrastructure services like virtual switching, security, and storage from CPUs, they significantly reduce the CPU cycles spent on these tasks. CPU cores are freed up to improve application performance. Cloud service providers can deploy infrastructure functions with the agility of software applications. As a result, data center use is optimized, workload placement is more flexible, and overall system efficiency is improved.

The technical mechanisms behind these IPUs involve leveraging programmable ASICs or field-programmable gate arrays (FPGAs) to manage and accelerate network infrastructure tasks. These also include the implementation of virtual switches (vSwitches) in hardware, which traditionally consume significant CPU resources when implemented in software. By running vSwitches on IPUs, network traffic routing between virtual machines (VMs) becomes far more efficient, reducing latency and freeing up server CPU for primary application tasks.

This IPU-backed infrastructure:

  • Diminishes operational costs and capital expenditures by reducing the servers needed to support data center workloads.
  • Increases security because IPUs create secure routes to each core, preventing data leakage across multi-tenant environments and helping protect against cyber threats.
  • Directly support ultra-low latency, high-performance microservices-based applications, essential for modern cloud-native applications. Given that server CPU use becomes optimized, it minimizes latency between microservers and significantly reduces the total number of servers required for the workload, considerably reducing server costs and energy consumption.

Migration made easy with Kinsta

Migrating from C2 to C3D machines on Google Cloud brings technical challenges, such as adapting to advanced hardware capabilities, integrating new security protocols, ensuring seamless network performance enhancements, and maintaining operational continuity. With Kinsta, you don’t need to do anything to migrate — we do that for you at no extra cost.

We’re here to ensure you have a smooth and efficient migration experience tailored to your specific needs. We simplified the transition for our clients, ensuring you can quickly experiment with the benefits of C3D machines without handling the migration process on your own.

At Kinsta, we have already migrated all WordPress websites from C2 to C3D in all data centers where C3D machines are available on the Google Cloud:

  1. Jurong West, Singapore (asia-southeast1)
  2. St. Ghislain, Belgium (europe-west1)
  3. Eemshaven, Netherlands (europe-west4)
  4. Council Bluffs, Iowa, USA (us-central1)
  5. Moncks Corner, South Carolina, USA (us-east1)
  6. Ashburn, Virginia, USA (us-east4)

As C3Ds become available in other data centers, we’ll follow the transition.

To start using C3D machines with new sites right now, just choose a data center labeled Boosted in the MyKinsta UI when you create your WordPress site:

Selecting a Boosted data center location in MyKinsta. Selected: South Carolina - US-East1.
Selecting a Boosted data center location.


The introduction of the C3D machines to Google Cloud marks a pivotal advancement in cloud computing, emphasizing enhanced performance, security, and efficiency. Central to this leap is the 4th Gen AMD EPYC™ CPUs and Google’s custom IPUs, which together ensure high computational power and improved network capabilities. The IPUs, in particular, offer significant enhancements in networking performance, isolation, and security, aligning with Google Cloud’s vision for a future-proof cloud infrastructure.

Migrating to C3D machines on Google Cloud provides strategic advantages for high-performance computing and other demanding workloads, thanks to their optimized hardware architecture and the advanced capabilities of IPUs.

For businesses aiming to stay at the forefront of cloud technology without the hassle of direct migration complexities, hosting sites on Kinsta is a smart choice. Kinsta’s expert support and migration services let you seamlessly leverage the benefits of C3D machines, ensuring minimal disruption and maximizing the potential of the latest cloud innovations.

With Kinsta, rest assured that anytime there’s an upgrade, Kinsta has you covered. Enjoy the best Managed WordPress Hosting with C3Ds with no effort demanded from you and your team.

Marcia Ramos Kinsta

I'm the Editorial Team Lead at Kinsta. I'm a open source enthusiast and I love coding. With more than 7 years of technical writing and editing for the tech industry, I love collaborating with people to create clear and concise pieces of content and improve workflows.