Frequently Asked Questions

TurboGPU is a cloud GPU rental service that gives you a GPU machine: a Linux (Ubuntu) container with a dedicated NVIDIA GPU, the CUDA runtime and PyTorch pre-installed, root access over SSH and a browser-based Jupyter studio. You connect from any device over SSH or a browser-based Jupyter studio, and pay only for the time you use. There is no hardware to buy, maintain, or upgrade.

Managed notebook services limit you to their runtimes and time out your sessions. TurboGPU gives you a complete Linux machine with full root access. You can install any software, run AI and ML workloads, render, compile, or do anything you would on a physical GPU box.

You need a device with an internet connection (Windows, Mac, Linux, iPad, or Android), an SSH client or terminal (built into Mac and Linux, free on Windows), and a TurboGPU account with at least $1 in credit. Prefer a GUI? Every instance also ships a browser-based Jupyter studio.

Machines run on GPU hosts in several countries. You can't pick a region: when you deploy, TurboGPU picks an available host for the GPU you chose. The Products page shows where each available GPU is located.

TurboGPU uses a pay-as-you-go credit system. Credits are denominated in US dollars — 1 credit = $1. You add credits in the currency you choose: a top-up can be any amount from €1 to €20,000, or its equivalent in the currency you pick at the day's rate (US dollars included), with the maximum rounded down to a round figure so it never exceeds €20,000. Then choose a GPU plan. GPU time is charged by the second at your tier's hourly rate. The clock starts when a GPU host is assigned to your machine (so the few minutes it takes to boot are included) and stops when you stop or delete the machine. There are no subscriptions, contracts, or hidden fees.

GPU time is charged by the second at your tier's hourly rate. The clock starts when a GPU host is assigned to your machine (so the few minutes it takes to boot are included) and stops when you stop or delete the machine. Stopped machines are not charged. While a machine runs, the time it has used is deducted from your balance every few minutes, and the rest when you stop it.

We accept Visa and Mastercard credit and debit cards. You enter the card in our checkout and it goes to our payment processor over an encrypted (HTTPS) connection; we don't store it.

Unused credits can be refunded to the original payment method within 14 days of purchase. Email support@turbogpu.tech from your account email.

Once a day, if a machine is running and your balance is below $5, we email you a low-balance warning. If your balance reaches zero while a machine is running, the machine is stopped automatically (we check every five minutes) and we email you. Your files and machine state are preserved for 24 hours.

We offer four tiers: Starter (RTX A4000, 16 GB), Standard (RTX 4090, 24 GB), Pro (L40S, 48 GB), and Power (A100 80GB). All plans include NVMe storage, multiple vCPUs, and full root access.

It depends on the host: a machine can take a few minutes to pull its image and boot. Your dashboard switches to Running as soon as it is ready and then shows your SSH connection details.

Yes. When you stop a machine, your files, installed packages, and disk state are preserved for 24 hours.

No. Machines are Linux containers without a display: work over SSH, VS Code Remote or the browser-based Jupyter studio. There is no remote desktop, Windows or game streaming.

Every machine is a Linux (Ubuntu) container with root access. The CUDA runtime, cuDNN and PyTorch are pre-installed; the NVIDIA driver comes from the GPU host.

Copy the SSH command from your dashboard, or open the browser-based Jupyter studio. For coding, the VS Code Remote-SSH extension works great.

SSH needs very little — 5 Mbps is plenty. The Jupyter studio is smoother at 25 Mbps or more.

Use a wired Ethernet connection where possible, and work over SSH, which needs the least bandwidth.

Yes. Forward any port over SSH (for example -L 8888:localhost:8888) to reach Jupyter, a web UI, or a dev server as if it were running locally.

Yes. Use any SSH client app, or open the Jupyter studio directly in your mobile browser.

Anything CUDA-accelerated: PyTorch and TensorFlow training, Stable Diffusion, LLM inference, Blender/OptiX rendering, video encoding, scientific computing, and more.

No. Each machine is itself a container, so Docker can't run inside it. Install what you need with apt, pip or conda as root.

Absolutely. TurboGPU is used for AI and machine learning, 3D rendering, data science, software builds, CAD, and more.

No. Crypto mining and blockchain proof-of-work computation are prohibited under our Terms of Service.

Check that you have sufficient credit, then stop the machine and start it again. If the issue persists, contact support@turbogpu.tech.

Make sure your machine is Running. Double-check the host, port, and key or password shown in your dashboard. Ensure your local firewall is not blocking outbound SSH (port 22).

Switch to a wired connection, close other applications using bandwidth, or work over SSH, which needs the least bandwidth.

Run nvidia-smi to confirm the GPU is visible. If it isn't, stop the machine and deploy a new one, and tell support@turbogpu.tech the machine ID.