immers.cloud is a cloud GPU platform that lets you host virtual servers equipped with powerful NVIDIA GPUs: ideal for everything from cloud gaming and 3D rendering to machine learning and running large language and generative AI models.
We offer pay-as-you-go-billing, meaning you pay only for the exact time your server is running.
The Sign-up process takes seconds: no long forms, no approvals, no waiting. Create your first GPU instance and start working within minutes of registration.
Our GPU servers are used for:
When you host a GPU server, you get full, dedicated access to the hardware. Everything is managed through an intuitive web dashboard or via OpenStack API.
The immers.cloud platform enforces quotas and limits:
Your current quotas are displayed when you click on your username in the top menu.
Default quotas for all new users:
Current quotas can be increased upon request — simply contact technical support.
Platform limits on immers.cloud:
An individual need to finish registration, confirm the email address and recharge the personal account. Money is most often credited without a delay.
Organization representative needs to finish registration, confirm the corporate email address, add the organization and pay the first invoice. A copy of the payment can be sent to payments@immers.cloud for an accelerated transfer of funds.
You can change password on profile page: https://immers.cloud/user/profile/
Use PuTTY for Windows or ssh for Linux. Use IP аddress of the VM as host address. For modern cloud images there are typical logins: ubuntu, debian, fedora, opensuse, cloud-user. Use pem or ppk file downloaded after creation of specified Key Pair.
To shut down the Virtual Machine, use the Stop option of the Actions drop-down menu on the Virtual Machines list page or on the page of a individual Virtual Machine.
Billing of CPU and RAM of a virtual machine stops only when using mentioned method of shutting down the virtual machine.
Yes, you can. When VM setup and software installation will be finished stop the VM and create snapshot. When snapshot creation will be completed you can create multiple Virtual Machines from that Snapshot.
Ordered Virtual Machines are created automatically at any time of the day. Typically, the process planing and hosting a virtual machine, deploying an image and the cloud-init stage are completed in a few minutes.
We don't have mininal rental period. Use VM as long as you need it.
Currently one Availability Zone is available located in data center in Moscow, Russia.
Use run command with keys: Windows + R, then type and run gpedit.msc
Find Local Computer Policy\Computer Configuration\Administrative Templates\Windows Components\Remote Desktop Services\Remote Desktop Session Host\Remote Session Environment and Enable option “Use the hardware default graphics adapter for all Remote Desktop Services sessions”.
For TurboBoost activation visit the server BIOS using iKVM and go to:
Advanced Setup Configurations → CPU Configuration → Advanced Power Management Configuration
Select
Power Technology = Custom
Became active:
Hardware PM (Power Management) State Control
Enable inner option:
Hardware P-State: native.
Our servers are equipped with Intel® Xeon® Gold 2nd, 3rd, and 5th Generation processors:
At immers.cloud, you get access to 13 powerful NVIDIA GPUs. From flagship AI accelerators to high-performance graphic cards for rendering and gaming, we cover it all.
No need to buy or maintain expensive hardware. Run your projects in the cloud with full performance on demand.
For large-scale neural network training and inference, choose our latest models: Tesla A100, H100, and the new H200. These are the foundation of major generative AI, scientific research, and data science projects.
With massive memory capacity, high bandwidth, and next-gen tensor cores, they deliver record-breaking performance. They also handle long-context workloads with ease.
Need rendering, 3D visualization, or gaming? Go with RTX series GPUs: from the proven RTX 2080 Ti to the top-tier RTX 5090.
They excel in Blender, Cinema 4D, Unreal Engine, Redshift, Octane, and other render engines. All support NVENC hardware encoding: perfect for streaming and 4K+ video processing.
For lighter tasks like voice, image, and video processing, Tesla T4, A10, A2, and L4 are ideal.
These are efficient, cost-effective, and perfect for low-latency GPU virtual desktops and services. Stable, reliable, and affordable.
All powered by the cloud. No baremetal. Just results.
For rendering and machine learning tasks, it is important that the amount of RAM is greater than the total amount of VRAM, so for a configuration with four A10s (each have 24GB of video memory), we recommend configurations with at least 96 GB of RAM.
For gaming, we recommend configurations with GPU, 8 CPU cores, 16 GB of RAM and a 160 GB disk.
Streaming configurations are usually less memory or disk intensive. For disk-intensive tasks, we recommend SSD drives (Volume + Volume SSD-based instance type) or Local, otherwise use the cheapest HDD-based Volume + Volume instance type.
More than 100 images with pre-installed operating system are available:
You can also choose a pre-configured image in the Marketplace.
It is possible. Own images are created through web-interface via uploading a file (glance-direct ) or specifying image URL (web-download).
Install NVIDIA drivers from the official archive:
wget http://us.download.nvidia.com/XFree86/Linux-x86_64/440.64/NVIDIA-Linux-x86_64-440.64.run
chmod +x NVIDIA-Linux-x86_64-440.64.run
sudo ./NVIDIA-Linux-x86_64-440.64.run
or use images with the drivers installed: Ubuntu 16.04.6 or Ubuntu 18.04.4 with CUDA.
Currently each server has two 10GbE network interfaces.
Internet is connected to the VMs using up to 20Gb/s uplink. You can order extended or guaranteed bandwidth if needed via technical support.
Technically number of networks and adressess are not limited. On practice personal quotes will apply. To increase personal Quotes contact technical support.
Go to Server Manager > Add Roles and Features Wizard then select "Features" on the left hand side, then finally check "Wireless LAN Service" and click next/install. Finally, restart.
In some cases Windows Task Manager displays incorrect GPU Load information. Try to use any other GPU monitoring tools, please.
cutil64.dll was a part of CUDA library version 4.2 and earlier. Find this file in CUDA SDK (for example, version 4.2) and copy it to an application folder.
Use the PowerShell utility running as administrator to install the .NET Framework 3.5:
After starting the terminal as administrator, type Install-WindowsFeature NET-Framework-Core
Verify that .NET 3.5 is installed by re-typing this command and verifying that Success: True is present.
Run installer with the administrator rights.
Actions to add a path to a PATH environment variable
There are 2 prepaid tariffs available for long-term rental of VMs:
This means that you can rent a VM with a 25% or 50% discount by prepaying the full rental period at once. The full rental cost will be immediately chardged from the balance, so the required amount of funds must be on the balance before creating a VM.
Prepaid tariffs are available only for VMs with GPU. Prepaid tariffs may not be available for all GPUs at different times. The relevance of the currently available prepaid tariffs is reflected on the page: https://en.immers.cloud/prices/.
For a VM running according to the prepaid tariff, expenses in the "Purchases" section will be added once per day. On the VM page you can see the number of remaining prepaid days.
Important: You cannot change flavor for a VM created according to the prepaed tariff
For SHUTOFF Virtual Machines, spendings are made for the actually reserved resources - SSD space and HDD volumes, in accordance with their volume, external IP addresses and personal VLANs, reserved PCI-Express devices (only for servers with GPUs). vCPU and RAM are not charged.
You can store files on en.immers.cloud and manage them using dedicated software, known as an S3 client.
To begin, you need to create an S3 user account and obtain connection details.
Creating an S3 User and Connection Details
Navigate to S3 User Management to create a new user.
Specify the S3 ID and User Name (you can choose any name).
Once the user is created, the Access Key and Secret Key will be generated and displayed.
Go to Login Credentials to generate login details.
All fields will be filled in automatically; simply save the credentials.
The Access and Secret keys will also be available here.
Use any S3 client to connect to S3 storage, such as S3 Browser.
Setting Up S3 Browser
Follow these steps to configure the S3 Browser:
Open the Accounts menu and select Add New Account.
Fill out the fields on the "New User" page:
Display Name: Enter any name (this is for local display only).
Account Type: Choose "S3 Compatible Storage."
REST Endpoint: Enter https://api.immers.cloud:8080.
Access Key ID: Use the Access Key from the S3 user details.
Secret Access Key: Use the Secret Key from the S3 user details.
Advanced Settings:
Set the Addressing Model to Path Style.
Save the configuration.
Click Next, then Add New Account.
Your account is now ready. You can create buckets (referred to as "baskets" in the interface) and upload files.
Managing Buckets
Buckets are displayed on the website in the Buckets Section.
You can monitor your bucket activity in the Bucket Operations Section.
Shut down the virtual machine using the Stop option in the action drop-down menu. Use the Create Snapshot option and wait for the preparation of the Snapshot, making sure it receives the ACTIVE status.
In the working directory of the Moonlight.exe in the latest Moonlight-xxxxxxxxxx.log you can find:
00:00:01 - SDL Info (0): GPU driver: igdumdim64.dll 20.19.15.4377
00:00:01 - SDL Info (0): Detected buggy Intel GPU driver installed. Update your Intel GPU driver to enable HEVC!
To solve this issue install a newer driver from Intel: https://www.intel.ru/content/www/ru/ru/download/19344/intel-graphics-windows-dch-drivers.html
We provide GRID vWS/vApps licenses for Tesla cards.
Tesla A10, T4 and V100 support work in TCC and WDDM modes. By default GRID driver use WDDM mode. To change mode execute command with administrator privileges:
nvidia-smi -g {GPU_ID} -dm {0|1}
-dm 0 - WDDM Mode
-dm 1 - TCC Mode
RTX and GeForce cards works only in WDDM mode.
In case of manual installation of GRID drivers you must set License Server parameters:
Address: licenses.immers.cloud
Port: 7070
Immersion cooling is currently the most cost-effective way to cool high‑density IT equipment. Unlike a traditional air‑cooled data center, our system uses only 5% of the total energy for cooling.
Our IT hardware operates without fans, reducing power consumption by up to 30% right out of the box. Each of our 42U racks can accommodate up to 160 GPUs with a power load of up to 80 kW per rack.
Immersion cooling keeps components at optimal temperatures, ensuring stability and enabling higher clock speeds for every CPU and GPU.