Graphics Processing Units (GPU)
You can add a GPU (graphics processing unit) to a cloud server when creating a server or to an existing server.
GPUs are used as dedicated PCI devices inside a cloud server.
Graphics processing units are available in fixed and custom GPU line configurations.
GPU lineup configurations and custom configurations with GPUs can be used with a local or network boot volume.
For cloud servers with a local disk, you can use:
- NVIDIA® A100 40Gb;
- NVIDIA® A100 80Gb;
- NVIDIA® A30;
- NVIDIA® RTX 4090 48 GB;
- NVIDIA® A5000;
- NVIDIA® RTX 6000 Ada;
- NVIDIA® RTX 6000 Pro;
- NVIDIA® H200.
If you need a server with a set of preconfigured tools and libraries for machine learning and data analysis, use the AI Marketplace.
Available GPUs
You can view the current list of GPUs in the control panel: from the top menu, click Products → Cloud Servers → click Create Server.
You can check GPU availability in regions in the GPU for Cloud Servers availability matrix.
NVIDIA® A100 40 GB
Provides maximum performance for AI, HPC, and data processing. Suitable for deep learning, scientific research, and data analytics.
Based on the Ampere® architecture, with 40 GB of HBM2 memory and a bandwidth of up to 1.5 GB/s. See detailed specifications in the NVIDIA® A100 40Gb documentation from NVIDIA®.
In fixed GPU lineup configurations, 1 to 8 GPU × 40 GB are available, with 6 to 48 vCPUs and 87 to 700 GB of RAM.
In custom configurations, 1 to 8 GPU × 40 GB are available, with 2 to 32 vCPUs and 512 MB to 256 GB of RAM.
NVIDIA® A100 40 GB NVLink
For NVIDIA® A100 40Gb GPUs connected using NVLink technology.
NVLink increases data transfer speeds when connecting GPUs compared to the PCIe interface. Multiple GPUs connected via NVLink allow for more memory usage and improve server performance for complex computing tasks, such as training large language ML models.
NVLink works with the NVIDIA® A100 40Gb — GPUs based on the Ampere® architecture, with 40 GB of HBM2 memory and a bandwidth of up to 1.5 GB/s. See detailed specifications for the NVIDIA® A100 40Gb and a description of NVLink technology in the NVIDIA® documentation.
NVIDIA® A100 40Gb NVLink are available upon request — create a ticket.
NVIDIA® A100 80 GB
Provides maximum performance for AI, HPC, and data processing, as well as a large memory capacity for resource-intensive tasks. Suitable for deep learning, scientific research, and data analytics.
Based on the Ampere® architecture, with 80 GB of HBM2 memory and a bandwidth of up to 1.5 GB/s. See detailed specifications in the NVIDIA® A100 80Gb documentation from NVIDIA®.
In fixed configurations of the GPU line, from 1 to 8 GPUs × 80 GB are available, with 12 to 96 vCPUs, 128 GB to 1 TB of RAM, and 128 GB to 6.88 TB of local storage.
In custom configurations, from 1 to 8 GPUs × 80 GB are available, with 12 to 192 vCPUs, 64 GB to 1 TB of RAM, and 256 GB to 3.36 TB of local storage.
NVIDIA® Tesla T4
Suitable for Machine Learning and Deep Learning, inference, graphics, and video rendering. Works with most AI frameworks and is compatible with all types of neural networks.
Based on the Turing® architecture, with 16 GB of GDDR6 memory and a bandwidth of up to 300 GB/s. See detailed specifications in the NVIDIA® Tesla T4 documentation from NVIDIA®.
In fixed GPU lineup configurations, 1 to 4 GPU × 16 GB are available, with 4 to 24 vCPUs and 32 to 320 GB of RAM.
In custom configurations, 1 to 4 GPU × 16 GB are available, with 2 to 32 vCPUs and 512 MB to 256 GB of RAM.
NVIDIA® A30
Suitable for AI inference, HPC, language processing, conversational AI, and recommendation systems.
Based on the Ampere® architecture, with 24 GB of HBM2 memory and a bandwidth of up to 933 GB/s. See detailed specifications in the NVIDIA® A30 documentation from NVIDIA®.
In fixed GPU lineup configurations, 1 to 2 GPU × 24 GB are available, with 16 to 48 vCPUs and 64 to 320 GB of RAM.
In custom configurations, 1 to 2 GPU × 24 GB are available, with 2 to 32 vCPUs and 512 MB to 256 GB of RAM.
NVIDIA® A2
Entry-level GPU. Suitable for simple inference, video and graphics, Edge AI (edge computing), streaming video, and mobile cloud gaming.
Based on the Ampere® architecture, with 16 GB of GDDR6 memory and a bandwidth of up to 200 GB/s. See detailed specifications in the NVIDIA® A2 documentation from NVIDIA®.
In fixed GPU lineup configurations, 1 to 4 GPU × 16 GB are available, with 12 to 48 vCPUs and 32 to 320 GB of RAM.
In custom configurations, 1 to 4 GPU × 16 GB are available, with 2 to 32 vCPUs and 512 MB to 256 GB of RAM.
NVIDIA® GTX 1080
High-performance and energy-efficient GPU. The solution is implemented using FinFET technology and GDDR5X memory. Dynamic load balancing helps split tasks so that resources do not sit idle. Provides maximum performance for information display, VR, ultra-high resolution parameters, and data processing.
Based on the Pascal® architecture, with 8 GB of GDDR5X memory and a bandwidth of up to 320 GB/s. See detailed specifications in the NVIDIA® GTX 1080 documentation from NVIDIA®.
In fixed GPU lineup configurations, 1 to 8 GPU × 8 GB are available, with 8 to 28 vCPUs and 24 to 96 GB of RAM.
In custom configurations, 1 to 8 GPU × 8 GB are available, with 2 to 32 vCPUs and 512 MB to 256 GB of RAM.
NVIDIA® RTX 2080 Ti
A high-performance GPU for complex graphics tasks. Suitable for:
- high-resolution video processing;
- 3D model creation;
- rendering and photo processing;
- neural network training;
- complex artificial intelligence calculations;
- large-scale data processing.
Based on the Turing® architecture, with 11 GB of GDDR6 memory and a bandwidth of up to 616 GB/s. See detailed specifications in the NVIDIA® RTX 2080 Ti documentation from NVIDIA®.
Fixed GPU lineup configurations offer 1 to 4 GPUs × 11 GB, with 2 to 48 vCPUs and 32 to 320 GB of RAM.
Custom configurations offer 1 to 4 GPUs × 11 GB, with 2 to 32 vCPUs and 512 MB to 256 GB of RAM.
NVIDIA® RTX 4090 24 Gb
A high-performance GeForce series GPU suitable for:
- professional design and 3D modeling;
- video editing and rendering;
- ML tasks (model training and inference);
- working with large language models (LLMs);
- for scientific and engineering calculations (for example, in climate modeling or bioinformatics). Based on the Ada Lovelace® architecture, with 24 GB of GDDR6X memory and a bandwidth of up to 1008 GB/s. See detailed specifications in the NVIDIA® RTX 4090 documentation from NVIDIA®.
Fixed GPU lineup configurations offer 1 to 4 GPUs × 24 GB, with 4 to 64 vCPUs and 16 to 356 GB of RAM.
Custom configurations offer 1 to 4 GPUs × 24 GB, with 2 to 32 vCPUs and 4 to 256 GB of RAM.
NVIDIA® RTX 4090 48 Gb
A high-performance GeForce series GPU with more memory than the NVIDIA® RTX 4090 24 Gb, suitable for:
- professional design and 3D modeling;
- video editing and rendering;
- ML tasks (model training and inference);
- working with large language models (LLMs);
- scientific and engineering calculations (e.g., climate modeling or bioinformatics).
Based on the Ada Lovelace® architecture, with 48 GB of GDDR6X memory and a bandwidth of up to 1008 GB/s. See detailed specifications in the NVIDIA® RTX 4090 documentation from NVIDIA®.
Fixed GPU lineup configurations offer 1 to 8 GPUs × 48 GB, with 12 to 192 vCPUs, 64 to 896 GB of RAM, and a local disk from 64 to 800 GB.
Custom configurations offer 1 to 8 GPUs × 48 GB, with 12 to 192 vCPUs, 64 to 896 GB of RAM, and a local disk from 50 to 800 GB.
NVIDIA® RTX 6000 Ada
Professional GPU for computing and graphics power. Suitable for ML tasks, rendering, scientific calculations, and high-performance visualization.
Based on the Ada Lovelace® architecture, with 48 GB of GDDR6X memory and a bandwidth of up to 960 GB/s. See detailed specifications in the NVIDIA® RTX 6000 Ada documentation from NVIDIA®.
Fixed GPU lineup configurations offer 1 to 4 GPUs × 48 GB, with 12 to 96 vCPUs, 64 to 450 GB of RAM, and a local disk from 64 GB to 2 TB.
Custom configurations offer 1 to 4 GPUs × 48 GB, with 12 to 96 vCPUs, 64 to 450 GB of RAM, and a local disk from 64 GB to 3.52 TB.
NVIDIA® A2000
Energy-efficient GPU for compact workstations. Suitable for AI, graphics, and video rendering.
Based on the Ampere® architecture, with 6 GB of GDDR6 memory and a bandwidth of up to 288 GB/s. See detailed specifications in the NVIDIA® A2000 documentation from NVIDIA®.
Fixed GPU lineup configurations offer 1 to 4 GPUs × 6 GB, with 6 to 24 vCPUs and 16 to 320 GB of RAM.
Custom configurations offer 1 to 4 GPUs × 6 GB, with 2 to 32 vCPUs and 512 MB to 256 GB of RAM.
NVIDIA® A5000
A versatile GPU suitable for any tasks within its performance range.
Based on the Ampere® architecture, with 24 GB of GDDR6 memory and a bandwidth of up to 768 GB/s. See detailed specifications in the NVIDIA® A5000 documentation from NVIDIA®.
Fixed configurations in the GPU line offer from 1 to 8 GPUs × 24 GB, with vCPU from 8 to 128, RAM from 16 to 700 GB, and a local volume from 64 GB to 1 TB.
Custom configurations offer from 1 to 8 GPUs × 24 GB, with vCPU from 2 to 128, RAM from 2 to 731 GB, and a local volume from 20 GB to 2 TB.
NVIDIA® H100
A powerful GPU suitable for AI, HPC, and scalable computing.
Based on the Hopper™ architecture, with 80 GB of HBM3 memory and a bandwidth of up to 3 TB/s. See detailed specifications in the NVIDIA® H100 documentation from NVIDIA®.
Fixed GPU lineup configurations offer 1 to 2 GPUs × 80 GB, with 12 to 48 vCPUs and 128 to 256 GB of RAM.
Custom configurations offer 1 to 2 GPUs × 80 GB, with 2 to 48 vCPUs and 2 GB to 256 GB of RAM.
NVIDIA® H200 NVLink
Two NVIDIA® H200 GPUs connected using NVLink technology.
NVLink increases data exchange speed when connecting GPUs compared to the PCIe interface. GPUs connected via NVLink allow for more memory usage and increase server performance for complex computations, such as training large language models (LLMs).
NVLink works with NVIDIA® H200 — GPUs based on the Hopper™ architecture, with 141 GB of HBM3 memory and a bandwidth of up to 4.8 TB/s. See detailed specifications of NVIDIA® H200 in the NVIDIA® documentation.
Fixed configurations of the Dedicated line feature 4 GPUs × 141 GB, with vCPUs from 48 to 96, RAM from 480 to 720 GB, and a local volume from 512 GB to 1 TB.
NVIDIA® H200
A professional GPU suitable for:
- accelerating generative AI;
- high-performance computing (HPC);
- inference of large language models (LLMs);
- model fine-tuning;
- image and video generation.
Based on the Hopper™ architecture, with 141 GB of HBM3 memory and a bandwidth of up to 4.8 TB/s. See detailed specifications of NVIDIA® H200 in the NVIDIA® documentation.
Fixed and custom GPU configurations feature from 1 to 8 GPUs × 141 GB, with vCPUs from 12 to 192, RAM from 120 GB to 1 TB, and a local volume from 256 GB to 3 TB.
NVIDIA® L4
A universal GPU for accelerating AI/ML workloads, video processing, streaming, and VDI. Suitable for running modern language models (LLMs) and multimodal models.
Based on the Ada Lovelace® architecture, with 24 GB of GDDR6 memory and a bandwidth of up to 3 TB/s. See detailed specifications of NVIDIA® L4 in the NVIDIA® documentation.
Fixed GPU configurations feature from 1 to 8 GPUs × 24 GB, with vCPUs from 8 to 128, and RAM from 32 GB to 512 GB.
Custom configurations feature from 1 to 8 GPUs × 24 GB, with vCPUs from 8 to 256, and RAM from 64 GB to 640 GB.
NVIDIA® RTX 6000 Pro
A professional GPU for:
- accelerating generative AI;
- inference of language models (LLMs);
- model fine-tuning;
- image and video generation;
- 3D rendering and video processing.
Based on the Blackwell® architecture, with 96 GB of GDDR7 memory and a bandwidth of up to 1.6 TB/s. See detailed specifications of NVIDIA® RTX 6000 Pro in the NVIDIA® documentation.
Fixed and custom GPU configurations feature from 1 to 8 GPUs × 96 GB, with vCPUs from 16 to 256, RAM from 120 GB to 1 TB, and a local volume from 256 GB to 3 TB.
Create a cloud server with GPU
Use the Create a cloud server instruction.
When creating a server, select:
- a source — GPU-optimized images marked in the version list as GPU optimized. These images include the drivers required for GPU operation. If you choose another source, you must install the drivers on the server yourself for stable NVIDIA® GPU operation;
- a configuration — a fixed or custom GPU configuration starting from 2 vCPUs.