Thuê máy chủ GPU train model ngon rẻ bổ trên - Mì AI

3 Feb 202418:11

TLDRThe video introduces a service called 'Thu' that allows users to rent GPU servers for deep learning and AI model training. The speaker discusses the limitations of free cloud platforms like Colab and the high costs and inconveniences of international GPU rental services. The service offers fast data upload and download within Vietnam, convenient payment options via QR code, and competitive pricing. The video demonstrates the process of registering, creating a virtual machine, and deploying a model on the server, highlighting the ease of use and efficiency compared to other options.


  • 🌐 The video introduces a service for renting GPU servers to run AI models, which is beneficial for those without access to a GPU.
  • 🚀 The service is particularly useful for students and researchers who need to work with models but are constrained by limited resources.
  • 🕒 Time limitations on free platforms like Colab can be a significant hurdle, cutting off access after 10-12 hours of use.
  • 🗂️ Limited VRAM on free platforms can restrict the size and complexity of the models that can be run.
  • 💰 The cost of cloud services can be prohibitive, especially for students who may not have access to credit cards or the funds to pay for higher-tier services.
  • 📊 The video presents an alternative service, 'Thu', which is based in Vietnam and offers competitive pricing and convenient payment options.
  • 🔄 The service provides high bandwidth for data upload and download, which is crucial for handling large datasets and models.
  • 💵 Payment can be made via bank transfer or QR Code, which is more accessible for local users and students.
  • 🛠️ The configuration of the GPU servers is tailored to meet the needs of students and researchers, with ample CPU, RAM, and GPU memory.
  • 🔧 The setup process is straightforward, with options for different operating systems and a user-friendly interface.
  • 📈 The video demonstrates the ease of accessing and using the GPU server, including the transfer of data and running AI models efficiently.

Q & A

  • What is the main issue discussed in the video?

    -The main issue discussed in the video is the difficulty faced by students and researchers in accessing GPU resources for their AI models due to limitations in time, storage, and speed on platforms like Colab and the high costs of purchasing a personal GPU-enabled PC.

  • What are the limitations of using Colab for running AI models?

    -The limitations of using Colab include a time limit of 10 to 12 hours of continuous usage after which the data is deleted, limited GPU VRAM of only 12 GB, and storage constraints with only 70 GB of space allowed for file uploads.

  • What is the alternative solution presented in the video for accessing GPU resources?

    -The alternative solution presented is renting a GPU server from a service provider, specifically mentioning a service called 'Thu', which allows users to rent GPU servers on an hourly basis and pay according to their usage.

  • What are the benefits of using a local GPU server rental service like 'Thu'?

    -The benefits include high-speed data upload and download due to being based in Vietnam, convenient payment methods like QR code transfers, affordable pricing for students, and the ability to provide invoices for businesses.

  • How does the payment system work for the 'Thu' service?

    -The payment system works by first depositing money into an account, and then transferring the desired amount from the account balance to a 'cloud account' which is used to pay for the GPU server rental. This ensures that the user only pays for the resources they actually use.

  • What are the specifications of the GPU server provided by 'Thu'?

    -The GPU server provided has a 24 GB GPU, 48 GB of RAM, a 200 GB hard drive, and a 20-core CPU. It offers unlimited bandwidth and operates on Ubuntu or Windows OS.

  • How to check the GPU specifications of the rented server?

    -The GPU specifications can be checked by using the 'nvidia-smi' command in the terminal, which displays information about the GPU model and the memory available.

  • What is the process for renting and setting up a GPU server on 'Thu'?

    -The process involves logging into the account, depositing money, transferring funds from the main account balance to the cloud account, selecting the server configuration, choosing the operating system, and waiting for the server to be set up. Once ready, the user can SSH into the server and start working on their AI models.

  • How to upload data to the rented GPU server?

    -For Linux users, the 'scp' or 'sftp' commands can be used to securely transfer files to the server. For Windows users, an FTP client like FileZilla can be used, or the user can simply copy and paste files directly during a Remote Desktop session.

  • What is the estimated cost for renting a GPU server from 'Thu'?

    -The cost is approximately 8000 VND per hour and 5.7 million VND per month, depending on the server configuration and usage.

  • How does the video demonstrate the effectiveness of the rented GPU server?

    -The video demonstrates the effectiveness by showing the speed at which the server can download data and run AI models, highlighting that it is significantly faster than using Colab and provides a more seamless and efficient experience.



🌐 Introduction to GPU Server Rental Service

The paragraph introduces the audience to a service for renting GPU servers, which is particularly useful for those who want to work on AI models but lack their own GPU resources. It highlights the limitations of using free platforms like Colab, such as time restrictions, limited VRAM, and storage capacity. The speaker then presents a solution in the form of a local service, emphasizing its benefits like high data upload/download speeds, convenient payment options, and the availability of invoices for businesses.


💻 Setting Up and Using the GPU Server

This paragraph walks through the process of setting up and using the rented GPU server. It explains how to register and log in to the service, manage the account, and top up the account balance using QR code payments. The speaker also details the steps to create a virtual GPU server, select the server's location, and choose the operating system. The paragraph emphasizes the ease of use and the flexibility in configuring the server to meet the user's needs.


🚀 Testing the GPU Server's Performance

The speaker shares their experience with testing the performance of the GPU server. They demonstrate how to connect to the server, check the GPU specifications, and run a test using a sample AI model. The paragraph highlights the impressive speed and efficiency of the server in running the model and handling data, contrasting it with the limitations of free platforms. The speaker also mentions the service's cost-effectiveness and the fact that it only charges for the time the server is actively used.


📂 Data Transfer and Model Deployment

This paragraph focuses on the process of transferring data to the GPU server and deploying AI models. The speaker explains how to use SFTP for Linux users and FTP clients or direct copy-paste for Windows users to upload data. They also discuss the ease of reading data from the uploaded files and deploying the model. The paragraph concludes with a recommendation for users to learn Linux commands and an encouragement to try the GPU rental service for a seamless and efficient experience.




A video blog or 'Vlog' is a type of content where someone shares their experiences or knowledge through video recordings. In the context of the script, the speaker is welcoming viewers back to their Vlog where they discuss and share information about a specific topic, which in this case is about GPU servers.

💡GPU server

A GPU server refers to a computer server that is equipped with one or more graphics processing units (GPUs) which are specialized hardware designed to efficiently process complex calculations required for rendering graphics. In the video, the speaker talks about the importance of having access to a GPU server for running AI models and the limitations faced without one.


Colab, short for Google Colaboratory, is a cloud-based platform offered by Google that allows users to run Python code in their browser with access to a free GPU. The speaker discusses the limitations of using Colab for running AI models, such as time restrictions and limited resources.

💡Cloud computing

Cloud computing is the delivery of computing services, such as server time, storage, databases, networking, software, analytics, and intelligence, over the Internet (the 'cloud'). In the context of the video, the speaker introduces a service for renting GPU servers in the cloud, which is a form of cloud computing.

💡Storage capacity

Storage capacity refers to the maximum amount of data that can be stored in a storage device or system. In the video, the speaker discusses the storage limitations of using Colab and the benefits of having a larger storage capacity when renting a GPU server.

💡Upload/download speed

Upload/download speed refers to the rate at which data is transferred from one place to another, either to a storage device or over a network. The speaker emphasizes the importance of having a fast upload/download speed when working with large AI models and data sets, which is facilitated by renting a GPU server within the same country.

💡Payment method

Payment method refers to the way in which a customer transfers money to a business to pay for goods or services. In the video, the speaker mentions different payment methods, including credit card payments and bank transfers, and highlights the convenience of using QR code payments for renting GPU servers.

💡Virtual machine

A virtual machine (VM) is a software emulation of a physical computer that can execute programs like a real machine. In the context of the video, the speaker talks about creating a virtual machine on the GPU server to run their AI models.

💡Operating system

An operating system (OS) is the system software that manages computer hardware, software resources, and provides various services for computer programs. The speaker discusses the choice of operating systems available on the GPU server, such as Ubuntu or Windows.


SSH, or Secure Shell, is a protocol that provides a secure channel over an unsecured network for data transmission between two systems. In the video, the speaker uses SSH to connect to the rented GPU server.

💡Data transfer

Data transfer refers to the movement of data from one location to another, often used in the context of uploading or downloading files. The speaker discusses different methods of data transfer, such as SFTP and FTP, for moving data to the GPU server.

💡AI model

An AI model refers to a system designed to process input data and produce an output result based on patterns learned from training data. The speaker discusses the benefits of running AI models on a GPU server, highlighting the increased speed and efficiency.


Introduction to a service allowing GPU server rental for AI model training, addressing the limitations faced by users without GPU access.

Discussion on the challenges of using Google Colab for AI model training, including limited time, small GPU VRAM, and storage capacity issues.

Presentation of cloud GPU services as a solution for AI model training without personal GPU hardware.

Highlighting the difficulties of accessing foreign cloud GPU services for students, including credit card payment requirements.

Introduction of, a Vietnam-based cloud GPU service offering local data transfer speeds and convenient payment options for students.

Explaining the registration and login process on for new users.

Guide on managing account funds and renting cloud GPUs on

Details on creating a virtual cloud GPU server, including selection of specifications and operating system.

Demonstration of how to connect to and use the rented GPU server for AI model training.

Showcasing the process of installing necessary frameworks and running an AI model training session on the server.

Illustrating how to transfer data to and from the cloud server using SSH and FTP for model training.

Explanation of the benefits of cloud GPU server rental for AI projects, emphasizing speed and efficiency.

Discussion on the cost-effectiveness of using for students and researchers.

Highlighting the support for both Linux and Windows operating systems on servers.

Concluding thoughts on the value of cloud GPU services like for AI model training.