[Yandex Cloud documentation](../../index.md) > [Yandex Compute Cloud](../index.md) > [Concepts](index.md) > Graphics processing units (GPUs) # Graphics processing units (GPUs) Compute Cloud provides graphics accelerators (GPUs) for different VM [configurations](#config). GPUs outperform CPUs for certain types of data and can be used for complex computing. For even more performance and convenience, you can use automatic allocation of resources in [Yandex DataSphere](../../datasphere/concepts/index.md). The following GPUs are available in Compute Cloud: * [NVIDIA® Tesla® V100](https://www.nvidia.com/en-gb/data-center/tesla-v100/) with 32 GB HBM2 (High Bandwidth Memory). * [NVIDIA® Ampere® A100](https://www.nvidia.com/en-us/data-center/a100/) with 80 GB HBM2. * [NVIDIA® Tesla® T4](https://www.nvidia.com/en-us/data-center/tesla-t4/) with 16 GB GDDR6. {% note warning %} GPUs run in [TCC](https://docs.nvidia.com/nsight-visual-studio-edition/reference/index.html#tesla-compute-cluster) mode, which does not use the operating system's graphics drivers. {% endnote %} By default, a cloud has a zero [quota](limits.md#compute-quotas) for creating VMs with GPUs. You can request a quota increase in the [management console](https://console.yandex.cloud/cloud?section=quotas). To do this, you need the `quota-manager.requestOperator` [role](../../iam/roles-reference.md#quota-manager-requestoperator) or higher. ## Graphics accelerators (GPUs) {#gpu} Graphics accelerators are suitable for machine learning (ML), artificial intelligence (AI), and 3D rendering tasks. You can manage GPUs and RAM directly from your VM. ### NVIDIA® Tesla® V100 {#tesla-v100} The NVIDIA® Tesla® V100 graphics card contains 5120 CUDA® cores for [high-performance computing](https://www.nvidia.com/en-us/high-performance-computing/) (HPC), and 640 Tensor cores for deep learning (DL) tasks. ### NVIDIA® Ampere® A100 {#a100} The NVIDIA® A100 GPU based on the [Ampere®](https://www.nvidia.com/en-us/data-center/ampere-architecture/) microarchitecture uses third-generation Tensor Cores and offers 80 GB HBM2 memory with up to 2 TB/s bandwidth. ### NVIDIA® Tesla® T4 {#tesla-t4} NVIDIA® Tesla® T4 based on the [Turing™](https://images.nvidia.com/aem-dam/en-zz/Solutions/design-visualization/technologies/turing-architecture/NVIDIA-Turing-Architecture-Whitepaper.pdf) architecture uses Turing tensor cores and RT cores and offers 16 GB of GDDR6 memory with [300 GB/s bandwidth](https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/tesla-t4/t4-tensor-core-datasheet-951643.pdf). ### T4i {#t4i} T4i GPU uses Tensor Cores and offers 24 GB GDDR6 memory with up to 300 GB/s bandwidth. ### VM configurations {#config} The computing resources may have the following configurations: * Intel Broadwell with NVIDIA® Tesla® V100 (`gpu-standard-v1`): Number of GPUs | VRAM, GB | Number of vCPUs | RAM, GB --- | --- | --- | --- 1 | 32 | 8 | 96 2 | 64 | 16 | 192 4 | 128 | 32 | 384 * Intel Cascade Lake with NVIDIA® Tesla® V100 (`gpu-standard-v2`): Number of GPUs | VRAM, GB | Number of vCPUs | RAM, GB --- | --- | --- | --- 1 | 32 | 8 | 48 2 | 64 | 16 | 96 4 | 128 | 32 | 192 8 | 256 | 64 | 384 * AMD EPYC™ with NVIDIA® Ampere® A100 (`gpu-standard-v3`): Number of GPUs | VRAM, GB | Number of vCPUs | RAM, GB --- | --- | --- | --- 1 | 80 | 28 | 119 2 | 160 | 56 | 238 4 | 320 | 112 | 476 8 | 640 | 224 | 952 * Gen2 (`gpu-standard-v3i`): Number of GPUs | VRAM, GB | Number of vCPUs | RAM, GB --- | --- | --- | --- 1 | 80 | 18 | 144 2 | 160 | 36 | 288 4 | 320 | 72 | 576 8 | 640 | 180 | 1440 * Intel Ice Lake with NVIDIA® Tesla® T4 (`standard-v3-t4`): Number of GPUs | VRAM, GB | Number of vCPUs | RAM, GB --- | --- | --- | --- 1 | 16 | 4 | 16 1 | 16 | 8 | 32 1 | 16 | 16 | 64 1 | 16 | 32 | 128 * Intel Ice Lake with T4i (`standard-v3-t4i`): Number of GPUs | VRAM, GB | Number of vCPUs | RAM, GB --- | --- | --- | --- 1 | 24 | 4 | 16 1 | 24 | 8 | 32 1 | 24 | 16 | 64 1 | 24 | 32 | 128 * GPU PLATFORM V4 (`gpu-standard-v4`): Number of GPUs | VRAM, GB | Number of vCPUs | RAM, GB --- | --- | --- | --- 1 | 141 | 22 | 220 2 | 242 | 44 | 440 4 | 484 | 88 | 880 8 | 968 | 180 | 1800 VM GPUs are provided in full. For example, if a configuration has four GPUs specified, your VM will have four full-featured GPU devices. You can create VMs based on Intel Broadwell with NVIDIA® Tesla® V100, Intel Cascade Lake with NVIDIA® Tesla® V100, and AMD EPYC™ with NVIDIA® Ampere® A100 in the `ru-central1-a` and `ru-central1-b` availability zones. For more information about organizational and technical limitations for VMs, see [Quotas and limits](limits.md). For information about the cost of VMs with GPUs, see [Prices for the Russia region](../pricing.md#prices). ### OS images {#os} The following special OS images with NVIDIA drivers pre-installed are available for VMs with GPUs: Intel Broadwell with NVIDIA® Tesla® V100 and Intel Cascade Lake with NVIDIA® Tesla® V100 : * [Ubuntu 18.04 LTS GPU](https://yandex.cloud/en/marketplace/products/yc/ubuntu-18-04-lts-gpu) (`ubuntu-1804-lts-gpu`) * [Ubuntu 20.04 LTS GPU](https://yandex.cloud/en/marketplace/products/yc/ubuntu-20-04-lts-gpu) (`ubuntu-2004-lts-gpu`) Intel Ice Lake with NVIDIA® Tesla® T4 : * [Ubuntu 20.04 LTS GPU](https://yandex.cloud/en/marketplace/products/yc/ubuntu-20-04-lts-gpu) (`ubuntu-2004-lts-gpu`) Intel Ice Lake with T4i : * [Ubuntu 22.04 LTS GPU CUDA 12.2](https://yandex.cloud/en/marketplace/products/yc/ubuntu-2204-lts-cuda-12-2) (`ubuntu-2204-lts-cuda-12-2`) AMD EPYC™ with NVIDIA® Ampere® A100 : * [Ubuntu 22.04 LTS GPU CUDA 12.2](https://yandex.cloud/en/marketplace/products/yc/ubuntu-2204-lts-cuda-12-2) (`ubuntu-2204-lts-cuda-12-2`) For cluster mode support: : * [Ubuntu 20.04 LTS GPU Cluster](https://yandex.cloud/en/marketplace/products/yc/ubuntu-2004-lts-gpu-cluster)(`ubuntu-2004-lts-gpu-cluster`) Gen2 : * [Ubuntu 20.04 LTS Secure Boot CUDA 12.2](https://yandex.cloud/en/marketplace/products/yc/ubuntu-2004-lts-secureboot-cuda-12-2) (`ubuntu-2004-lts-secureboot-cuda-12-2`) We recommend using a standard Yandex Cloud image. You can also manually [install the drivers](../operations/vm-operate/install-nvidia-drivers.md) on another standard image or [create a custom image](../operations/image-create/custom-image.md) with pre-installed drivers. {% note info %} Compute Cloud performs health checks and recommends installing only [LTS versions of drivers](https://docs.nvidia.com/datacenter/tesla/drivers/releases.json). When installing drivers for `gpu-standard-v3` (AMD EPYC™ with NVIDIA® Ampere® A100), specify the compatible driver version, `535`. We recommend using this specific driver version; updates to other versions are not supported and may lead to unstable GPU performance. {% endnote %} ## GPU clusters {#gpu-clusters} You can group several VMs into a cluster. This will allow you to accelerate distributed training tasks that require higher computing capacity than individual VMs can provide. Make sure the cluster is created in the same availability zone as its VMs. The cluster VMs are interconnected through InfiniBand, a secure high-speed network. You can add VMs from different folders, networks, and subnets to your cluster. For the cluster VMs to interact properly, we recommend using a [security group](../../vpc/concepts/security-groups.md) that allows unlimited traffic within the group. The default security group meets this requirement. If you edited the default security group, add a group with unlimited internal traffic. Maximum Gen2 cluster size is 20 VMs with 8 GPU, 80 GB VRAM, 180 vCPU, 1,440 GB RAM configuration. The actual maximum cluster size is limited by the technical availability of the resources. ## See also {#see-also} * [Creating a VM with a GPU](../operations/vm-create/create-vm-with-gpu.md) * [Adding a GPU to an existing VM](../operations/vm-control/vm-update-resources.md#add-gpu) * [Changing the number of GPUs](../operations/vm-control/vm-update-resources.md#update-gpu) * [Questions about GPUs](../qa/gpu.md)