[Yandex Cloud documentation](../../index.md) > [Yandex Compute Cloud](../index.md) > [Concepts](index.md) > Virtual machines > vCPU performance levels

# vCPU performance levels


When creating a virtual machine, you need to select a vCPU performance level. This level defines the physical core computing time a vCPU can guarantee.

* Virtual machines with performance level **below 100%** have access to the computing power of physical cores for at least the specified percentage of a unit of time.

    > At 20% performance, a VM will have access to the physical cores at least 20% of the time, i.e., 200 milliseconds out of every second. During that time, the processor clock speed is not limited and depends on the selected platform, e.g., 2 GHz for Intel Ice Lake (`standard-v3`).

    <br>

    VMs with performance level below 100% are designed to run applications that do not require high performance and are not sensitive to delays. These VMs cost less.

* Virtual machines with **100%** performance level have continuous access (100% of the time) to the computing power of physical cores. Such VMs are designed to run applications that require high performance all the time.

## Specifics and limitations {#features-and-restrictions}

* Actual performance may exceed the guaranteed level. 

* Available configurations of computing resources may change for different levels, depending on the selected [platform](vm-platforms.md).

* Available disk space does not depend on the VM performance level.

* VMs with vCPU performance level below 100% may operate at lower [speed](storage-read-write.md) with network drives.

* Network drive service cores [do not depend](../../vpc/concepts/software-accelerated-network.md#reg-vm) on VM parameters; however, vCPU performance reduction directly affects disk read and write speed.

* VMs with over 256 vCPUs will run correctly only on operating systems with Linux kernel version 5.15 or higher. 

* For Windows-based VMs, the maximum number of vCPUs is 224.

For other limitations, see [Quotas and limits in Compute Cloud](limits.md).

## Minimum and maximum configurations {#minmax-configurations}

This subsection presents the minimum and maximum computing resource configurations depending on the platform. You can achieve these values in [different combinations](#available-configurations).

{% note info %}

The option with simultaneous maximum configuration of all computing resources is not available for all platforms. For more information, see [Available configurations](#available-configurations).

{% endnote %}


* Intel Broadwell (`standard-v1`):

    Limit | Performance<br>level | vCPU | RAM, GB<br>total | RAM, GB<br>per core
    ------------ | ----------------------------- | ---- | ---------------- | -----------------
    Min         | 5%                            | 2    | 1                | 0.5
    Max        | 100%                          | 32   | 256              | 8

* Intel Cascade Lake (`standard-v2`):

    Limit | Performance<br>level | vCPU | RAM, GB<br>total | RAM, GB<br>per core
    ------------ | ----------------------------- | ---- | ---------------- | -----------------
    Min         | 5%                            | 2    | 0.5              | 0.25
    Max        | 100%                          | 80   | 1280             | 16

* Intel Ice Lake (`standard-v3`):

    Limit | Performance<br>level | vCPU | RAM, GB<br>total | RAM, GB<br>per core
    ------------ | ----------------------------- | ---- | ---------------- | -----------------
    Min         | 20%                           | 2    | 1                | 0.5
    Max        | 100%                          | 96   | 640              | 16

* Intel Ice Lake Compute-Optimized (`highfreq-v3`):

    Limit | Performance<br>level | vCPU | RAM, GB<br>total | RAM, GB<br>per core
    ------------ | ----------------------------- | ---- | ---------------- | -----------------
    Min         | 100%                          | 2    | 2                | 1
    Max        | 100%                          | 56   | 448              | 16

* AMD Zen 3 (`amd-v1`)^1^:

    Limit | Performance<br>level | vCPU | RAM, GB<br>total | RAM, GB<br>per core
    ----- | ----------------------------- | ---- | ---------------- | -----------------
    Min  | 20%                           | 2    | 1                | 0.5
    Max | 100%                          | 128  | 768              | 6

* AMD Zen 4 (`standard-v4a`):

    Limit | Performance<br>level | vCPU | RAM, GB<br>total | RAM, GB<br>per core
    ----- | ----------------------------- | ---- | ---------------- | -----------------
    Min  | 20%                           | 2    | 1                | 0.5
    Max | 100%                          | 288  | 1792             | 8

* AMD Zen 4 Compute-Optimized (`highfreq-v4a`):

    Limit | Performance<br>level | vCPU | RAM, GB<br>total | RAM, GB<br>per core
    ----- | ----------------------------- | ---- | ---------------- | -----------------
    Min  | 20%                           | 2    | 1                | 0.5
    Max | 100%                          | 80   | 960              | 12

^1^ `amd-v1` is available to a limited number of users for private testing at the [Preview](../../overview/concepts/launch-stages.md) stage and is not free of charge. To request access to the platform, contact your account manager.

## Available configurations {#available-configurations}

* Intel Broadwell (`standard-v1`):

    Level<br>level | vCPU | RAM, GB<br>per core
    --- | --- | ---
    5% | 2, 4 | 0.5, 1, 1.5, 2
    20% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    100% | 2, 4, 6, 8, 10, 12, 14,<br> 16, 20, 24, 28, 32 | 1, 2, 3, 4, 5, 6, 7, 8

* Intel Cascade Lake (`standard-v2`):

    Level<br>level | vCPU | RAM, GB<br>per core
    --- | --- | ---
    5% | 2, 4 | 0.25, 0.5, 1, 1.5, 2
    20% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    50% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    100% | 2, 4, 6, 8, 10, 12, 14,<br> 16, 20, 24, 28, 32, 36,<br> 40, 44, 48, 52, 56, 60,<br> 64, 68, 72, 76, 80 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11, 12, 13, 14, 15, 16

* Intel Ice Lake (`standard-v3`):

    Level<br> level | vCPU | RAM per core, GB
    --- | --- | ---
    20% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    50% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    100% | 2, 4, 6, 8, 10, 12, 14, 16, 20 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11, 12, 13, 14, 15, 16
    100% | 24 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13
    100% | 28 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11
    100% | 32 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
    100% | 36, 40 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11, 12, 13, 14, 15, 16 
    100% | 44 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11, 12, 13, 14
    100% | 48 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11, 12, 13
    100% | 52 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11, 12
    100% | 56 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11
    100% | 60, 64 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
    100% | 68 | 1, 2, 3, 4, 5, 6, 7, 8, 9
    100% | 72, 76, 80 | 1, 2, 3, 4, 5, 6, 7, 8
    100% | 84, 88 | 1, 2, 3, 4, 5, 6, 7
    100% | 92, 96 | 1, 2, 3, 4, 5, 6

* Intel Ice Lake Compute-Optimized (`highfreq-v3`):

    vCPU | RAM, GB per core
    --- | --- 
    2, 4, 6, 8, 10, 12, 14,<br> 16, 20, 24, 28, 32, 36,<br> 40, 44, 48, 52, 56 | 1, 2, 3, 4, 5, 6, 7, 8, 9, 10,<br> 11, 12, 13, 14, 15, 16

* AMD Zen 3 (`amd-v1`)^1^:

    Level<br>level | vCPU | RAM, GB<br>per core
    --- | --- | ---
    20% | 2, 4 | 0.5, 1, 1.5
    50% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    100% | 2, 4, 8, 16, 32, 48, 64, 96, 128 | 1, 2, 3, 4, 5, 6

* AMD Zen 4 (`standard-v4a`):

    Level<br>level | vCPU | RAM, GB<br>per core
    --- | --- | ---
    20% | 2, 4 | 0.5, 1, 1.5, 2
    50% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    100% | 2, 4, 8, 16, 32, 48, 64, 96, 128, 224 | 1, 2, 4, 8
    100% | 256, 288 | 1, 2, 4

* AMD Zen 4 Compute-Optimized (`highfreq-v4a`):

    Level<br>level | vCPU | RAM, GB<br>per core
    --- | --- | ---
    20% | 2, 4 | 0.5, 1, 1.5, 2
    50% | 2, 4 | 0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4
    100% | 2, 4, 8, 16, 24, 32, 48, 80 | 1, 2, 4, 8, 10, 12

^1^ `amd-v1` is available to a limited number of users for private testing at the [Preview](../../overview/concepts/launch-stages.md) stage and is not free of charge. To request access to the platform, contact your account manager.