Bit of an Introduction:

Good choice selecting Data Science as a career. Data Science has now become highest paying jobs in IT industry, as data on internet increases day by day, demand of data engineers and data scientist are increasing rapidly.

But as demand increases, increases competition too. So to keep up with the competition one must need fast and stable Computer or Laptop. As both Machine learning and Deep learning requires frequent testing of algorithms or neural networks.

Machine Learning (from now on I will reference it as ML) and Deep Learning (and this as DL) are both compute intensive. DL is a subset of ML. ML model are bit different than DL models. DL model contains Neural nets. Thing is most used ML libraries (in python) doesn’t support GPU, thus they run on CPU but DL libraries (like tensorflow, mxnet,etc) supports Nvidia GPU.

So, if you will just learn ML then you require different specs. As most ML libraries don’t support GPU.

What are your options?

You have two options,

  1. Cloud instances
  2. Creating own PC.

Both options have its own merits and demerits.

  1. Cloud instancesMerits : Cheaper initially, if you lose interest in field you can stop as most cloud instance providers charge as per hours, more processing power on demand. Demerits: If you are privacy oriented and you have some sensitive data which you don’t to expose to internet then cloud might not be the option for you. I am not saying that its unsafe but it might be susceptible to hacking.
  2. Own PC buildMerits : No privacy issues as training data never leaves your computer, once started you can use as you want i.e total control. Demerits: If you somehow need more processing power then you have to change CPU or GPU, it can be costly.

So you ready to build a PC?

Lets start with basics, what are the parameters for selecting parts? Here I am only gonna talk about important parts and it does not include cases, power supply, display, accessories, etc.

For ML, CPU and RAM are most important.

  • For beginners:
    • CPU – Intel i5 8th gen or AMD Ryzen 5.
    • GPU – Not necessary as most ML libraries doesn’t support GPU.
    • RAM – Min. 8 GB DDR4. Recommended – 16 GB.
    • HDD – 1 TB HDD.
  • For Experienced:
    • CPU – Intel i7 8th gen or AMD Ryzen 7.
    • RAM – Min. 16 GB,
    • SSD Recommended 512GB.

For DL, GPU and RAM are important.

  • For beginners:
    • CPU – Intel i5 8th gen or AMD Ryzen 5.
    • GPU – Min – Nvidia GTX 1050 Ti. Recom. – Nvidia GTX 1060 (6 GB)
    • RAM – Min. 8 GB DDR4.
    • HDD – 1 TB
  • For Experienced:
    • CPU – Intel i7 8th gen or AMD Ryzen 7.
    • GPU – Nvidia GTX 1080 Ti.
    • RAM – Min. 16 GB.
    • SSD Recommended 512 GB.

OK, you might have noticed I haven’t mentioned AMD GPU here, so here’s why!!

Almost all Deep learning libraries are built keeping CUDA in mind and not OpenCL. And CUDA is supported by Nvidia and not AMD. Although AMD is porting libraries to OpenCL via their RoCM initiative but its good to stick with Nvidia right now as community support and drivers support is pretty good.

Why use GPU for DL instead of CPU?

GPU provided parallelism unlike CPU. Core in GPU are more than CPU thus it provides faster execution.

Now comes the best setup of all.

Legendary Setup:

If you are Richie rich then this setup is for you.

  • CPU – Intel i9 or AMD Threadripper
  • GPU – Min. Nvidia RTX 2080 Ti. Recommended – Nvidia Titan V.
  • RAM – 32 GB DDR4
  • SSD 1 TB.

Thats all for today, hope you like the specs recommended for ML and DL. Like, comment, share and subscribe to get latest updates of our blogs. To subscribe press red bell icon in bottom right corner of you screen.

Read our other informative blogs here.