# GPU dedicated servers: the whole machine, every GPU > GPU dedicated servers with Intel, NVIDIA or AMD GPUs: a whole physical machine and full root access for AI training, inference, rendering and video. 1. [Home](https://hyperdc.com/) 2. [Cloud Servers](https://hyperdc.com/cloud-servers) 3. [GPU Servers](https://hyperdc.com/gpu-servers) 4. GPU Dedicated Servers A physical server with Intel, NVIDIA or AMD GPUs that only you use. Every GPU, every processor core and all of the memory and storage work for your training runs, model serving, render queues and video pipelines, with full root access and no virtualization layer unless you install your own. - A whole physical server, never shared - Intel, NVIDIA or AMD GPUs - No virtualization layer between you and the GPU - Your own hypervisor with GPU passthrough View plans Service · gpu-01.example.com **gpu-01.example.com** 203.0.113.48 · Ubuntu 26.04 Running **GPU**Intel **GPU**NVIDIA **GPU**AMD GPU GPU memory CPU RAM ollama serve ComfyUI JupyterLab ## Choose a plan that fits Tell us what you need and our team will help you choose the right plan. [Contact us](https://hyperdc.com/contact-us) [Client area](https://hyperdc.com/account) ## Why choose a GPU dedicated server A GPU dedicated server gives one customer an entire physical machine. Every GPU, every processor core, all of the memory and the full disk throughput work for your workload alone, so training runs, model serving and render queues keep the same pace, and no other customer can slow you down. You choose the operating system, the drivers and the software stack, or install your own hypervisor and pass GPUs through to your virtual machines. It is the right step when GPU work has outgrown a virtual server. - Intel, NVIDIA or AMD GPUs, named with every server - Every GPU in the machine is yours - No other customer on the hardware - Linux, Windows Server or your own hypervisor - Full root access to drivers and firmware tools View servers [Compare GPU servers](https://hyperdc.com/gpu-servers) Bare metal Root access Your OS ## What you can expect from our GPU dedicated servers ### Every GPU is yours The GPUs in the machine work only for your workloads, with all of their memory and compute. ### Intel, NVIDIA or AMD Each server names its GPU, processor, memory and storage, so you can match it to your software before you order. ### No noisy neighbours No other customer runs on the hardware, so training and rendering times stay steady from one run to the next. ### Your own hypervisor Install Proxmox VE, VMware ESXi or Hyper-V and pass whole GPUs through to your own virtual machines. ### Full root access Install drivers, CUDA or ROCm, schedulers such as Slurm and your own services the way your team works. ### GPU in containers Give each container the GPUs it needs with the NVIDIA Container Toolkit, or pass AMD and Intel GPUs as devices. ### Your data on your hardware Datasets, model weights and renders stay on a machine that no other customer uses. ### One client area Order, renew and manage the server and its invoices from the same client area as your other services. ### Help from our team Open a support ticket if the server, a GPU or the network needs attention, and follow every reply in one place. ## What customers run on GPU dedicated servers ### Training and fine-tuning Fine-tune and train models with PyTorch, DeepSpeed or Hugging Face libraries on GPUs that no one else uses, with datasets and checkpoints on local storage next to them. - Steady run times without neighbours - Datasets stay on your hardware - Schedulers such as Slurm for your team root@vps: \~ $ ssh root@203.0.113.24 Welcome to Ubuntu 26.04 LTS (GNU/Linux x86\_64) root@vps:\~# apt update && apt upgrade -y 0 upgraded, 0 newly installed, 0 to remove. root@vps:\~# ufw allow 443/tcp Rule added root@vps:\~# systemctl is-active nginx active root@vps:\~# ### Large models for many users Serve large open models with vLLM or another inference server behind an OpenAI-compatible API, and split the GPUs between models or teams as your traffic needs. ### Render nodes and farms Render with Blender Cycles on CUDA, OptiX, HIP or oneAPI, or run your studio’s render manager on machines that keep going overnight and over the weekend. ### Video pipelines and transcoding Transcode libraries and live streams with FFmpeg on the GPU’s hardware video engines: NVENC on NVIDIA, VA-API on AMD and Intel, and Quick Sync on Intel. ### Your own GPU virtualization host Install a hypervisor and give each virtual machine a whole GPU through passthrough, for separate teams, customers or test environments on one machine. HyperDC host Your server*Nested virtualization* **VM 1** **VM 2** **VM 3** ### Research and data science Give a research group JupyterHub, GPU-accelerated data processing and long experiments on a machine they control, with no queue behind other customers. ## Technical specifications What every server on this page shares. The GPU, processor, memory and storage differ by server and are listed with each one. GPU Intel, NVIDIA or AMD Each server names its GPU\ Server type A whole physical server No other customer on the hardware\ Virtualization None, unless you install your own hypervisor GPU passthrough to your own VMs\ Access Root over SSH Administrator over Remote Desktop on Windows Server\ Containers Docker and Podman with GPU access NVIDIA Container Toolkit, or /dev/kfd and /dev/dri for AMD and Intel\ GPU drivers Installed by you with root access NVIDIA driver and CUDA, AMD ROCm or Intel GPU drivers\ Operating systems Listed with each server Pick one in the order form\ Locations Listed with each server Pick one in the order form\ Support Support tickets from the client area ## GPU VPS, GPU VDS or GPU dedicated server? All three come with Intel, NVIDIA or AMD GPUs and root access. They differ in how much of the GPU and of the machine is yours. | Feature | **GPU VPS** A virtual server with a GPU | **GPU VDS** A larger virtual server with nested virtualization | This page **GPU dedicated server** A whole physical server with its GPUs (Recommended) | | --- | --- | --- | --- | | What you get | A virtual server with a GPU | A larger virtual server with a GPU and nested virtualization | A whole physical server with its GPUs | | GPU | Shared vGPU or dedicated GPU | Shared vGPU or dedicated GPU | Every GPU in the machine | | GPU makers | Intel, NVIDIA or AMD | Intel, NVIDIA or AMD | Intel, NVIDIA or AMD | | Your own VMs insideNested virtualization passes the processor’s virtualization extensions to your server, so it can run virtual machines and Android emulators of its own. | Not included | Included | Yes, with your own hypervisor | | Hardware used only by youA VPS or VDS is a virtual machine on a physical host that also runs other customers' virtual machines. | Not included | Not included | Included | | Root access | Included | Included | Included | | Best for | Inference, image generation, notebooks and remote desktops | Android emulators, GPU labs with VMs and larger AI stacks | Training, large models for many users and long render queues | | Price level | Lowest | Middle | Highest | - ### GPU VPS A virtual server with a GPU What you get A virtual server with a GPU\ GPU Shared vGPU or dedicated GPU\ GPU makers Intel, NVIDIA or AMD\ Your own VMs insideNested virtualization passes the processor’s virtualization extensions to your server, so it can run virtual machines and Android emulators of its own. Not included\ Hardware used only by youA VPS or VDS is a virtual machine on a physical host that also runs other customers' virtual machines. Not included\ Root access Included\ Best for Inference, image generation, notebooks and remote desktops\ Price level Lowest - ### GPU VDS A larger virtual server with nested virtualization What you get A larger virtual server with a GPU and nested virtualization\ GPU Shared vGPU or dedicated GPU\ GPU makers Intel, NVIDIA or AMD\ Your own VMs insideNested virtualization passes the processor’s virtualization extensions to your server, so it can run virtual machines and Android emulators of its own. Included\ Hardware used only by youA VPS or VDS is a virtual machine on a physical host that also runs other customers' virtual machines. Not included\ Root access Included\ Best for Android emulators, GPU labs with VMs and larger AI stacks\ Price level Middle - This page ### GPU dedicated server (Recommended) A whole physical server with its GPUs What you get A whole physical server with its GPUs\ GPU Every GPU in the machine\ GPU makers Intel, NVIDIA or AMD\ Your own VMs insideNested virtualization passes the processor’s virtualization extensions to your server, so it can run virtual machines and Android emulators of its own. Yes, with your own hypervisor\ Hardware used only by youA VPS or VDS is a virtual machine on a physical host that also runs other customers' virtual machines. Included\ Root access Included\ Best for Training, large models for many users and long render queues\ Price level Highest A comparison of the three GPU lines, not of single plans: each page lists its plans with the GPU, resources and price. [See GPU VPS](https://hyperdc.com/gpu-vps) [See GPU VDS](https://hyperdc.com/gpu-vds) ## Set up your GPU dedicated server Step-by-step guides for GPU drivers, model servers and AI tools, written for current Ubuntu and Debian releases. [More guides](https://hyperdc.com/guides/search?q=gpu%20dedicated%20servers) - Tutorials [How to install vLLM on an NVIDIA GPU server with Docker](https://hyperdc.com/guides/tutorials/install-vllm) Prepare an NVIDIA GPU server, run vllm/vllm-openai with Docker Compose on 127.0.0.1, secure it with an API key and a Caddy proxy that only exposes /v1, tune memory and multi-GPU settings, or install it with uv and systemd. 45 min Advanced - Tutorials [How to install ComfyUI on an NVIDIA GPU server securely](https://hyperdc.com/guides/tutorials/install-comfyui) Set up ComfyUI from the official repository on a GPU server, run it under its own user with systemd, keep it off the public internet, add HTTPS with basic authentication, and manage models, custom nodes, backups and updates. 45 min Intermediate - Tutorials [How to install Ollama on Ubuntu or Debian and run local LLMs](https://hyperdc.com/guides/tutorials/install-ollama) Install Ollama as a systemd service, pull and run open models on CPU or GPU, test the REST API, move model storage and reach the API securely without exposing port 11434. 30 min Beginner - Product docs [Dedicated servers at HyperDC: what you get and first checks](https://hyperdc.com/guides/products/dedicated-servers) A dedicated server is a whole physical machine for you alone. What is included, which operating systems you can run and the hardware checks worth doing on day one. 15 min Intermediate - Tutorials [How to install JupyterLab on a Linux server with systemd and HTTPS](https://hyperdc.com/guides/tutorials/install-jupyterlab) Set up single-user JupyterLab for a dedicated Linux user: virtual environment, hashed password, a systemd service bound to 127.0.0.1, SSH tunnel or Caddy access, GPU support, backups and updates. 30 min Intermediate - Tutorials [How to run llama.cpp server as an OpenAI-compatible API](https://hyperdc.com/guides/tutorials/llama-cpp-server) Compile llama.cpp, test llama-server with a GGUF model from Hugging Face, run it as a hardened systemd service on 127.0.0.1 with an API key, and publish the OpenAI-compatible API over HTTPS. 40 min Intermediate ## What you get with HyperDC ### One account for everything Hosting, servers, domains and invoices are managed from the same client area. ### Invoices in one place Every invoice is listed in the client area, where you can also pay it online. ### Support from the client area Open a support ticket whenever you need help and follow every reply in one place. ## From order to online 1. ### Choose a service Pick the product, location and billing cycle that fit your project. 2. ### Complete your order Review your cart and pay with one of the available payment methods. 3. ### Receive your details We email your login details as soon as the service is ready. 4. ### Get help when you need it Open a support ticket from the client area and follow every reply. ## One provider for hosting, servers and domains Hosting and servers since 2013, with 267k+ websites hosted on our services. ### One client area for every service Order, manage and pay for hosting, servers and domains from the same account. hyperdc.com/account **Services** **Domains** **Invoices** **Linux VPS** Active **Web hosting** Active **example.com** Active ### Help from our team Open a support ticket from the client area and follow every reply. Support ticket **You**Can you help me move my website to a new server? **HyperDC team**Of course. Send us the details of your current host and we will plan the move with you. Answered ### Choose your location Pick the data center when you order. Each service page lists where it runs. [All data centers](https://hyperdc.com/data-centers) Europe*Nearest to you* United States ### Room to grow Move between hosting, VPS, VDS and dedicated servers as your project grows. [Compare servers](https://hyperdc.com/cloud-servers) ## Moving from another provider? Tell us what you run today and where it is hosted. We will reply with the steps to move it to your new HyperDC service. - Tell us what you run today and where it is hosted - Get the steps to move it to your new HyperDC service - Switch over when everything is ready [Contact us](https://hyperdc.com/contact-us) **Current host** **HyperDC** Files copied Databases imported DNS updated Live ## Services that work well together ### GPU VPS A virtual server with a shared vGPU or a dedicated GPU. [Learn more](https://hyperdc.com/gpu-vps) ### GPU VDS A GPU server with nested virtualization for VMs and emulators. [Learn more](https://hyperdc.com/gpu-vds) ### GPU Servers Every GPU line compared, with the AI apps that use it. [Learn more](https://hyperdc.com/gpu-servers) ### Dedicated Servers The whole physical machine, in five countries. [Learn more](https://hyperdc.com/dedicated-server-hosting) ### LLM API Hosting vLLM, llama.cpp and LocalAI behind your own API. [Learn more](https://hyperdc.com/llm-api-hosting) ### Nested Dedicated Servers Dedicated resources with virtualization extensions for your own VMs. [Learn more](https://hyperdc.com/nested-dedicated-server-hosting) ## Policies that apply - [Terms of Service](https://hyperdc.com/legal/terms-of-services) - [Refund Policy](https://hyperdc.com/legal/refund-policy) - [Privacy Policy](https://hyperdc.com/legal/privacy-policy) - [Acceptable Use Policy](https://hyperdc.com/legal/acceptable-use-policy) ## Frequently asked questions Still have a question? Send us a message and our team will reply by email. [Contact us](https://hyperdc.com/contact-us) [Open a ticket](https://hyperdc.com/support/new-ticket) ### What is a GPU dedicated server? A GPU dedicated server is a physical server with one or more GPUs that serves one customer only. Its GPUs, processor, memory, disks and network port are not shared with anyone, so performance stays predictable and you decide how the hardware is used. ### How is it different from a GPU VDS with a dedicated GPU? A dedicated GPU on a VPS or VDS is a whole GPU for your virtual server, on a physical host that also runs other customers’ virtual machines. A GPU dedicated server is the physical host itself: every GPU in it, all processor cores, memory and storage are yours, and you can install your own hypervisor. ### Which GPUs do you offer? Intel, NVIDIA and AMD GPUs. Each server on this page names its GPU, processor, memory and storage, so you can check it against the requirements of your software, for example CUDA for NVIDIA, ROCm for AMD or oneAPI for Intel, before you order. ### Can I run my own hypervisor and pass GPUs to virtual machines? Yes. Install a hypervisor such as Proxmox VE, VMware ESXi or Hyper-V and pass whole GPUs through to your own virtual machines, for example one GPU per team or per workload. Passthrough needs the processor’s IOMMU, which you enable in the server’s firmware settings; our team helps through a support ticket if you need access to them. ### Which operating systems can I install? The systems each server offers are listed on this page and in the order form: Linux distributions for AI and rendering work, Windows Server for graphics tools and remote desktops, or a hypervisor for your own virtual machines. Choose a release your GPU software supports. ### How do I install the GPU drivers? You install them yourself with root access: NVIDIA’s data center driver and CUDA toolkit from NVIDIA’s repositories, AMD ROCm or Intel’s GPU drivers. Our guides for vLLM, Ollama and ComfyUI walk through the NVIDIA setup and the NVIDIA Container Toolkit step by step. ### Is a GPU dedicated server right for training models? Yes. Training and fine-tuning keep GPUs busy for hours or days and read datasets and checkpoints from disk all the time. On a dedicated server no other customer shares the GPUs, the processor or the storage, so run times stay steady and your data stays on hardware only you use. ### Can I use the GPUs inside Docker? Yes. With NVIDIA GPUs, install the NVIDIA Container Toolkit and choose which GPUs each container gets with --gpus; AMD and Intel GPUs are passed to containers as devices such as /dev/kfd and /dev/dri. Images for vLLM, Ollama, ComfyUI and PyTorch then use them. ### How long does it take to set up a GPU dedicated server? It depends on the server and its location. When a server lists a delivery time, that is the time to expect; otherwise we prepare the server after your order is paid and email the login details as soon as it is ready. If you have a deadline, contact us before ordering. ### Do you also offer smaller GPU servers? Yes. GPU VPS and GPU VDS come with a shared vGPU, a share of a physical GPU, or a dedicated GPU, a whole GPU for your virtual server. They are a lighter start for inference, image generation, emulators and remote desktops. ### Which billing cycles can I choose? Every plan lists the billing cycles it is offered with. Switch the cycle above the plans to compare prices; longer cycles show the saving against monthly billing when there is one. ### How do I pay? Choose one of the payment methods offered at checkout. Every invoice also stays available in the client area, where you can pay it online. ### Where do I manage my service after ordering? Your service appears in the client area as soon as it is set up. From there you can manage it, pay invoices and open support tickets. ## Questions before you order? Send us a message and our team will help you choose the right service. [Contact us](https://hyperdc.com/contact-us) [Client area](https://hyperdc.com/account) Support ticket **You**Can you help me move my website to a new server? **HyperDC team**Of course. Send us the details of your current host and we will plan the move with you. Answered --- Source: \ Updated: 2026-10-10