Compute-optimized machine family for

Compute Engine
bookmark_border
View Release Notes

Compute-optimized virtual machine (VMs) instances are ideal for compute-intensive and
high performance computing (HPC) workloads. Compute-optimized VMs offer the highest
performance per core and are built on architecture that utilizes features like non-uniform
memory access (NUMA) for optimal reliable uniform performance.

Note: For the C3 machine series, see General-purpose machine family.

Machine Workloads

 HPC workloads

H3 machine series (Preview)  Computational fluid dynamics, crash safety, genomics,

 General scientific and engineering computing

 Memory-bound workloads

 Gaming (AAA game servers)

 High-performance computing (HPC)

C2D machine series
 High-performance databases

 Electronic Design Automation (EDA)

 Media transcoding
 Compute-bound workloads

 High-performance web serving

 Gaming (AAA game servers)

C2 machine series  Ad serving

 High-performance computing (HPC)

 Media transcoding

 AI/ML

In this machine family:

 H3 VMs (Preview) are powered by two 4th generation Intel Xeon Scalable processors
(code-named Sapphire Rapids), with a total of 88 cores, up to 352 GB of DDR5
memory, and an all-core frequency of 3.0 GHz.
  C2D VMs run on the 3rd generation AMD EPYC Milan processor and offer up to 3.5
GHz max boost frequency. C2D VMs have flexible sizing between 2 to 112 vCPUs
and 2 to 8 GB of memory per vCPU.
 C2 VMs run on the 2nd Generation Intel Xeon Scalable processor (Cascade Lake)
which offers up to 3.9 GHz sustained single-core max turbo frequency. C2 offers
VMs with 4 to 60 vCPUs and 4 GB of memory per vCPU.

H3 machine series
Preview

This feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section
of the Service Specific Terms. Pre-GA features are available "as is" and might have limited
support. For more information, see the launch stage descriptions.

H3 VMs are powered by the 4th generation Intel Xeon Scalable processors (code-named
Sapphire Rapids), DDR5 memory, and Google's custom Intel Infrastructure Processing
Engine (IPU).

H3 VMs offer the best price performance for compute-intensive high performance computing
(HPC) workloads in Compute Engine. H3 VMs are single-threaded and are ideal for a variety
of modeling and simulation workloads including computational fluid dynamics, crash safety,
genomics, financial modeling, and general scientific and engineering computing. H3 VMs
support compact placement, which is optimized for tightly-coupled applications that scale
across multiple nodes.

The H3 series is available in one size, comprising an entire host server. To save on licensing
costs, you can customize the number of visible cores, but you are charged the same price for
the VM. H3 VMs can use the entire host network bandwidth and come with a default network
bandwidth rate of up to 200 Gbps. However, the VM to internet bandwidth is limited to 1
Gbps.

Simultaneous multithreading (SMT) is disabled for H3 VMs and can't be enabled. There is
also no overcommitting to ensure optimal performance consistency.

H3 VMs are available on-demand, or with one and three-year committed use discounts
(CUDs). H3 VMs can be used with Google Kubernetes Engine.

H3 Limitations

The H3 machine series has the following restrictions:

 The H3 machine series is only available in a predefined machine type. Custom

machine shapes aren't available.
 You can't use GPUs with H3 VMs.
 H3 VMs don't currently support Local SSD.
 Internet egress limited to 1 Gbps.
  Persistent Disk performance capped at 15,000 IOPS and 240 MB/s throughput.

H3 machine types

Machine types vCPUs* Memory (GB) Max network egress bandwidth†

h3-standard-88 88 352 Up to 200 Gbps

*
A vCPU represents an entire core—no simultaneous multithreading (SMT).
†
Default egress bandwidth cannot exceed the number given. Actual egress bandwidth depends on the
destination IP address and other factors. See Network bandwidth.

Supported disk types for H3

H3 VMs can use the following block storage types:

 Balanced Persistent Disk (pd-balanced)

Max number of Persistent Disk Max total disk size (TB)

Machine types Local SSD
volumes per VM* across all disks
h3-standard-88 128 257 No

Network support for H3 VMs

H3 VMs require gVNIC network interfaces. H3 supports up to 200 Gbps network bandwidth
for standard networking.

H3 VMs require an updated gVNIC driver for the gVNIC network interface. Before
migrating to H3 or creating new H3 VMs, ensure that the operating system image that you
use is fully supported. These images include the updated gVNIC driver, even if the guest OS
shows the gve driver version as 1.0.0. If your H3 VM is using an operating system with
limited support (one that uses an older version of the gVNIC driver), the network bandwidth
can't scale to 200 Gbps.

If you use a custom image with the H3 machine series, you can manually install the most
recent gVNIC driver. Google recommends using the gVNIC driver v1.3.0 or later to benefit
from additional features and bug fixes.

C2D machine series

The C2D machine series provides the largest VM sizes and are best-suited for high-
performance computing (HPC). The C2D series also has the largest available last-level cache
(LLC) cache per core.

The C2D machine series comes in different machine types ranging from 2 to 112 vCPUs, and
offer up to 896 GB of memory. You can attach up to 3TB of local storage to these machine
types for applications that require higher storage performance.
  C2D standard and C2D high-cpu machines serve existing compute-bound workloads
including high-performance web servers, media transcoding, and gaming.
 C2D high-memory machines serve specialized workloads such as HPC and EDA,
which need more memory.

The C2D series supports these compute-bound workloads by using the third generation AMD
EPYC Milan platform.

The C2D series supports Confidential VM.

C2D Limitations

The C2D machine series has the following restrictions:

 You cannot use regional persistent disks.

 The C2D machine series is subject to different disk limits than the general-purpose
and memory-optimized machine families.
 The C2D machine series is available only in select zones and regions on specific CPU
processors.
 The C2D machine series doesn't support GPUs.
 The C2D machine series doesn't support sole-tenant nodes.

C2D machine types

C2D standardC2D high-cpuC2D high-mem

Machine types vCPUs* Memory (GB) Default egress bandwidth (Gbps)‡ Tier_1 egress bandw

c2d-standard-2 2 8 10 N/A
c2d-standard-4 4 16 10 N/A
c2d-standard-8 8 32 16 N/A
c2d-standard-16 16 64 32 N/A
c2d-standard-32 32 128 32 50
c2d-standard-56 56 224 32 50
c2d-standard-112 112 448 32 100

*
A vCPU represents a single logical CPU thread. See CPU platforms.
†
Default egress bandwidth cannot exceed the number given. Actual egress bandwidth depends on the destination
IP address and other factors. See Network bandwidth.
‡
Supports high-bandwidth networking for larger machine types.

Supported disk types for C2D

C2D VMs can use the following block storage types:

 Standard Persistent Disk (pd-standard)

 Balanced Persistent Disk (pd-balanced)
 SSD (performance) Persistent Disk (pd-ssd)
 Local SSD