This tutorial is about building a deep learning computer. We will do our best for you to understand this guide. I hope you will like this blog How to Build a Deep Learning Computer. If your answer is yes, please share after reading this.
Check how to make a deep learning computer
Recent advances in Deep Learning and its implementation have wreaked havoc on the market. Monster GPUs in the cloud can run massive computations in parallel at lightning speed. In areas such as voice recognition, image classification, and autonomous vehicles, mastering the target requires lots of training data, and running it on 4 or 8 GB of RAM can be a never-ending process on your Personal PC. And access to cloud services for training personal models can cost a fortune. Here we will explore the specifications required to build a small-scale system for deep learning applications at your site. We’ll also see why high-end CPUs can’t compete with low-end GPUs when it comes to computing data in parallel.
The CPU is the main processor of our computer, so it is actually the unit that processes all the information to run the programs we use and to run the operating system itself.
How important is CPU for deep learning?
Considering that the GPU (graphics card) will do all the heavy lifting of deep learning, it can be said that the CPU is not that important. At the same time, we want a decent CPU to have a fast computer that can multitask when we open lots of apps. A 6-core processor (12 threads) at this time is a good enough processor to use the computer to work even with heavy programs, and even if you like to play video games, it will be good to play in FULL HD (although sure backed up by a good graphics card too).
AMD or Intel?
In processors with similar performance, AMD usually has a better price, so my choice is AMD.
Choice of processor
- AMD Ryzen 5 2600 (6 cores, 3.4 GHz) – $141
GPU (graphics card)
The GPU is the most important component for deep learning because it will take care of all the computing power to run the deep learning libraries when we train/test our models.
How to choose the right GPU?
When we choose a GPU, we have to consider these 2 things:
- It is compatible with the deep learning library we use. The most common deep learning libraries today are Tensorflow, Pytorch, and Keras.
- It has enough vRAM to fit the models. In simpler terms, the larger our training dataset, the more memory we need. If there is not enough memory, we will get an error.
AMD or Nvidia?
Nvidia. There is no real graphics card competition for deep learning and Nvidia is the only way forward at the moment. Nvidia has developed CUDA, an architecture that enables parallel computing and supports major deep learning libraries (Tensorflow, Pytorch, Keras, Darknet and others).
What about AMD, is there an alternative AMD version of CUDA?
The amd CUDA alternative exists and is called ROCM, but it’s still not popular, there’s a lack of support compared to nvidia, and the performance is too low to be considered a real alternative.
- Nvidia GTX 1660 (6GB, 1408 CUDA cores) – $220
- Nvidia GTX 1070 (8GB, 1920 CUDA cores) – $400
- Nvidia GTX 1080TI (11GB, 3584 CUDA cores) – $1050
The motherboard is the board that connects all the components together.
How to choose the motherboard?
When choosing a motherboard, we must take into account:
- It is compatible with the CPU we choose.
- It has a PCI Express 16x 3.0 slot. This is usually not a problem, as all modern motherboards have them by default. Even better if there are more slots, so in the future we might decide to upgrade the computer by adding a second GPU.
Choice of motherboard:
- Gigabyte Aorus Elite b450 (ATX motherboard) – $110
RAM (random access memory) in normal computer use is important when we want to open many applications, because once opened they will be temporarily loaded into RAM). For deep learning, I suggest a minimum of at least 16 GB and a minimum frequency of 2400 MHz.
- 16GB (2x8GB) – $70
- 32GB (2x16GB) – $140
The power supply is the unit that powers all the components of the computer.
How to choose food?
When choosing a power supply, we must take into account:
- How many watts does the computer need? The 2 components that consume the most are the CPU and the GPU.
- The CPU I suggested requires around 50 watts, while GPUs require 100-250 watts.
- Food should be of good quality.
- Don’t try to save money by buying cheap power supplies (less than $50), as they can easily damage your computer.
- There is an identification that indicates the quality of a power supply and it is the registered trademark 80 PLUS.
- You can take a look here for more: https://en.wikipedia.org/wiki/80_Plus
- A 600 watt power supply will be enough to support all the components I have suggested in this post.
- EVGA 600w (80+ Bronze) – $85
- Cooler Master MWE 600 Watt (80+ Bronze) – $80
We need fast memory to install and run the operating system (Windows or Linux) and large memory to store files. It’s a good option to get 2 different memories. For the operating system, we can get an NVMe, which is a very fast SSD drive that can be plugged into the PCI Express on the motherboard. I suggest buying at least a 250GB drive to install the OS on. As a second drive, you can opt for a classic internal hard drive. The price is really low and we can get a whopping 2TB memory for just over $50.
- 256GB Silicon Power (NVMe) – $38
- Seagate Barracuda 2TB (3.5″ HD) – $55
How to choose the cover? When choosing coverage, two factors should be considered:
- It is large enough to hold all the components. If you have an ATX motherboard, the case must also be ATX.
- It has good air circulation to keep gear cool. Homes today typically have 3 fans in the front that move cool air into the box and one fan in the back that pulls warm air out of the box.
Choice of cases
- Montech FIGHTER 600 ATX (Midtower case, 4FANS included) – $54
Final Words: How to Build a Deep Learning Computer
Hope you understand this article How to Build a Deep Learning Computer, if your answer is no, you can ask anything via the contact forum section linked to this article. And if your answer is yes, share this article with your family and friends.