Designing a GPU Computing Strategy

The aim of this chapter is to introduce computer hardware with a GPU perspective and how to get started with GPU computing-friendly hardware. Though the GPU is the most important component in our complete read-through, knowledge of gathering all the other essential components to make the most out of the GPU is also quite necessary.

Therefore, the significance of such compatible computer components will be discussed to show their significance while configuring any GPU to derive maximum performance out of it.

An impact in the GPU health-to-performance ratio will also be discussed with a comparison of air and liquid cooling. Both types of cooling techniques will be discussed in detail by exploring different scenarios of usage and applicability. If you are looking to use GPUs for extended computations that can last for days, you can opt for custom...

Getting started with the hardware

In this section, we will focus on the importance of GPU computing hardware as an essential part of setting up a GPU computing platform. We'll enlist all the hardware components required for building such a system. We know that this book is related to GPU computing. But, apart from the GPU, why are the other hardware components also important? Let's read on.

The significance of compatible hardware for your GPU

As per the scope of this book, the graphics card is, of course, our primary hardware component. But a careful assessment of the system configuration that it would be a part of is also extremely crucial. A self-assessment of the graphics card would make it possible to achieve...

Building your first GPU-enabled parallel computer – minimum system requirements

In the previous section, we discussed the main components of the required configuration in brief. Since this configuration is primarily focused on GPU computing, it calls for a minimum set of requirements, in addition to your GPU. Thorough research on these is very essential, regardless of opting for a branded pre-assembled desktop or an assembled version you built yourself.

Scope of hardware scalability

Minimum system requirements set the scope of your computing needs to a bare minimum. That means with such a configuration at minimum expenditure, you should at least be able to carry out specific types of operation you intend to do on that...

Liquid cooling – should you consider it?

Liquid cooling involves the passing of liquid through CPU or GPU blocks within a loop in order to reduce temperatures by a huge extent. Achieving such drastic reductions in temperature levels (up to 30°C) is not so feasible when implemented with conventional air-cooling methods. The following image shows a liquid-cooled system:

The temperature factor

When processors are busy at work, they obviously generate heat continuously as they compute. Depending on variable loads, the temperatures are also affected. The main goal behind cooling is always to keep such temperatures down, especially when the processors are operating under a full load. It is always good practice to also...

Branded GPU-enabled PCs

If you are not currently looking to build a PC yourself but want to get started with GPU programming straightaway, this section will help you decide the branded system that's best for you.

Leading PC brands offer HPC workstations and GPU-enabled PCs ready for deployment. Such systems are equipped with GPUs that range from low-end to mid-range to high-end specifications and have their respective budgets. Jetson Nano, from NVIDIA, is a cost-effective System on a Chip (SoC) like Raspberry Pi, built especially for AI-based GPU computing. The price is just $99 USD, apparently, as unveiled at the GTC in 2019. It is a great gizmo for beginners getting started with AI computing.

While comparing different brands, your best way to explore is to look into each and every specification of such systems in detail. Depending on your requirements and application usage...

Why not DIY?

Here, we will revisit all the points of the previous section. But, this time, DIY options will make it more flexible and offer you a greater freedom of choice. Flexibility and freedom come from choosing each and every component that you are going to assemble yourself. It is recommended that you dedicate at least a month to researching different PC components on the market to build the best system possible on your estimated budget.

So, to build your system from scratch, let's start our DIY walkthrough with the most important component.

GPU

Without a doubt, this is our most important component, and is directly related to our subject of study in this book. If you are looking toward GPU computing applications...

Entry-level budget

Hardware requirements for beginners can be assumed to be a set of the minimum requirements to get started with the basics of GPU computing. So, in this section, we will explore different entry-level budget options that you can also access on PC Part Picker. So, let's enlist the PC configurations with their approximate prices, according to the following CPU-GPU combinations:

• Intel CPU and NVIDIA GPU:

CPU	Intel Core i3-8100	$118.89
CPU Cooler	Cooler Master – Hyper 212 EVO	$24.99
Motherboard	Gigabyte H310M A	$55.14
GPU	Asus NVIDIA GeForce GTX 1050Ti 4GB	$201.12
RAM	Corsair Vengeance 2x8GB	$119.99
SSD	ADATA – Ultimate SU800 128 GB	$28.99
HDD	Western Digital – Caviar Green 1 TB	$44.89
Case	Cooler Master – MasterBox Lite 5 ATX Mid Tower	$55.98
PSU	SeaSonic ...

Mid-range budget

Intermediate users are most likely to learn as well as carry out qualitative testing of their GPU applications on their systems. Let's explore the four combinations with a mid-range goal.

• Intel CPU and NVIDIA GPU:

...

CPU	Intel – Core i5-8400	$399.79
CPU Cooler	Corsair – H60 (2018) 57.2 CFM	$69.98
Motherboard	Gigabyte – B360M DS3H microATX LGA1151	$69.99
GPU	Asus – NVIDIA GeForce GTX 1070 Ti 8 GB	$399.99
RAM	Corsair Vengeance LPX 2x16GB	$200.00
SSD	ADATA – Ultimate SU800 256 GB	$44.99
HDD	Western Digital – Caviar Green 2 TB	$89.89
Case	Cooler Master – MasterBox Lite 5 ATX Mid Tower	$55.98
PSU	SeaSonic – FOCUS Plus Gold 850W 80+ Gold Certified Fully-Modular ATX	$119.89
Monitor	Acer – G226HQL 21.5 1920x1080	$89.99

High-end budget

High-end configurations are undoubtedly the preference of advanced and expert users who are already experienced in the field. So, in this final section, we will look into the best options in hardware, considering the best performance. Feel free to tweak or modify the respective configurations, the links to which have been provided for each:

• Intel CPU and NVIDIA GPU:

CPU	Intel – Core i7-9700K	$193.99
CPU Cooler	Corsair – H60 (2018) 57.2 CFM	$69.98
Motherboard	Asus – ROG STRIX Z390-E ATX LGA1151	$238.89
GPU	Gigabyte NVIDIA GeForce GTX 1080 Ti 11 GB	$754.98
RAM	Corsair Vengeance LPX 2x16GB	$200.00
SSD	ADATA – Ultimate SU800 256 GB	$44.99
HDD	Western Digital – Caviar Green 2 TB	$89.89
Case	Cooler Master – MasterBox Lite 5 ATX Mid Tower	$55.98
PSU	SeaSonic...

Summary

In this chapter, we learned about the basic concepts behind PC hardware, the different components of a PC, and ideas so that you can build your first GPU computing system. Then, we focused on various PC-cooling techniques, particularly focusing on liquid cooling. After discussing options regarding branded hardware for GPU computing solutions, we discussed different PC components and looked at a step-by-step procedure on how to assemble them into a single unit. Finally, we listed four different CPU-GPU configurations for each level of hardware requirement with links for you to tweak or modify.

Now that we're at the end of this chapter, you should now be able to distinguish between different PC hardware components and also how to assemble them to set up a GPU computing system on your own. You are now acquainted with how liquid cooling works and you can now think of...

Further reading

You can read the following articles to learn more about the topics we discussed in this chapter:

• NVIDIA Jetson Nano is a $99 Computer Built for AI, Powered by Ubuntu, J Sneddon (2019), OMG! Ubuntu! (online). Available at https://www.omgubuntu.co.uk/2019/03/nvidia-jetson-nano-99-computer-for-ai.