COMPUTER AND

NETWORKING ESSENTIALS
FIRST YEAR - COMPUTER SCIENCE AND ENGINEERING

COURSE FACULTY: MR. Y. A. MANE

WALCHAND COLLEGE OF ENGINEERING, SANGLI
FY CSE Division B
 4th Module.

Graphics Processing Unit (GPU) and Display Devices

Contents:
➢ Role of GPUs in modern
computers
➢ Graphics cards and their
components
➢ Display technologies:
CRT, LCD, LED.

What is GPU...?
A Graphics Processing Unit (GPU) is a special type of computer chip designed to handle many
calculations at once, especially those needed for displaying images, videos, and animations.
Its ability to process multiple pieces of data in parallel makes it very fast and useful not only
for graphics but also for tasks like machine learning and video editing.
Ray Tracing:
Ray tracing is a computer graphics technique used to create highly realistic images by
simulating the way light behaves in the real world. When a ray hits an object, the algorithm
calculates how the light would reflect, refract, or be absorbed. The GPU handles complex
shading, reflections, and textures in real time to create immersive visuals.

Fig. Ray Tracing

Key Characteristics of a GPU:
1. Parallel Processing:
A GPU can perform many calculations at the same time (parallel), making it faster for
tasks that can be divided into smaller parts.
2. High Throughput:
It can process a large amount of data in a short time, which helps in handling
complex and heavy workloads efficiently.
3. Specialized for Graphics:
GPUs are mainly designed for creating and displaying 2D and 3D graphics, like
rendering textures, shading, and adding visual effects in video games or animations.
4. General-Purpose Computing:
Modern GPUs are not limited to graphics. They are also used for general computing
tasks such as data analysis, machine learning, and scientific simulations.
5. CUDA and OpenCL:
To Enable GPUs computing, nvidia introduced CUDA (Compute Unified Device
Architecture), and AMD introduced opencl, which are programming platforms that
allow developers to write code that can run on GPUs.
6. Memory Hierarchy:
GPUs typically have their own dedicated memory, known as video memory or VRAM.
7. Multiple Display Outputs:
GPUs often have multiple video outputs, enabling them to support multiple displays
simultaneously.
8. Cooling Solutions:
GPUs are equipped with robust cooling solutions, including fans or liquid cooling, to
dissipate heat and maintain stable operation.

Types of GPU :
Discrete GPUs are standalone chips, often on graphics cards. These are used in gaming, 3D
rendering, and simulations. A good example is the RTX 4090, which powers high-end gaming
and creative software.
Integrated GPUs, or iGPUs, are built into CPUs or SoCs. They’re common in laptops and
smartphones. Intel’s Iris Xe and Apple’s M2 chip are great examples.
Virtual GPUs exist in the cloud. They let users run GPU workloads without owning physical
hardware. For instance, Google Cloud offers virtual NVIDIA GPUs for scalable AI training.
GPU Processors:
Several GPU processors available from various manufacturers, each designed for specific
applications.
NVIDIA GPUs:
1. NVIDIA GeForce RTX 30 Series:
These GPUs are designed for gaming and graphics-
intensive applications, providing high performance
and real-time ray tracing capabilities.
2. NVIDIA Quadro Series:
Targeted at professional graphics and workstation
applications, quadro GPUs are optimized for tasks like
3d modeling, animation, and video editing.
3. NVIDIA Tesla and A100:
These GPUs are designed for data centers and are used
in high-performance computing, deep learning, and AI
Applications.

Intel Xe Graphics:
Intel's Xe graphics architecture includes integrated graphics
solutions for laptops and desktops, as well as data center
GPUs for AI and high-performance computing workloads.

ARM Mali GPUs:

ARM's Mali GPUs are commonly found in mobile devices,
providing graphics processing for smartphones and tablets.

AMD GPUs:
AMD Radeon RX Series:
These GPUs are aimed at gaming and multimedia
applications, offering competitive performance and
affordability.
AMD Radeon Pro Series:
Designed for professional applications like content creation
and design, these GPUs offer advanced features and support
for demanding workloads.
AMD Instinct Series:
AMD’S Data Center GPUs, such as the mi100 and mi200,
are optimized for high-performance computing and
machine learning.

Imagination Technologies PowerVR GPUs:

These GPUs are used in various embedded and mobile
applications, including automotive infotainment systems
and IoT devices.
Qualcomm Adreno GPUs:
Adreno GPUs are integrated into Qualcomm Snapdragon
mobile processors, powering graphics in smartphones,
tablets, and other mobile devices.
Apple GPUs:
Apple has designed its custom GPUs for use in its products,
including iPhones, iPads, and Mac Computers.
Vivante GPUs:
Vivante Corporation designs GPUs for a range of
applications,
including mobile devices, automotive systems, and IoT
products.
Architecture of GPU:

GPUs are built as highly parallel processors, optimized to handle thousands of threads at once.
Unlike CPUs that emphasize sequential execution, GPUs achieve massive parallelism through
multiple Streaming Multiprocessors (SMs). Each SM is a self-contained unit capable of
executing arithmetic and control instructions, making them the core of GPU computation.
To support this parallelism, GPUs rely on a hierarchical memory system. At the SM level, the
L1 cache provides quick access to frequently used data. Beyond that, an on-chip L2 cache acts
as a shared resource across all SMs, improving data reuse and reducing latency. For larger
storage needs, GPUs use high-bandwidth memory DRAM (HBM), which offers significant
capacity though at slower speeds compared to on-chip caches.
This combination of parallel compute units and layered memory hierarchy enables GPUs to
deliver exceptional performance in tasks like AI, scientific simulations, and graphics rendering.
Arithmetic and other instructions are executed by the SMs; data and code are accessed from
DRAM via the L2 cache.

Example: NVIDIA A100

the NVIDIA A100 GPU includes:
• 108 SMs – enabling massive parallelism.
• 40 MB of L2 cache – facilitating fast shared access across SMs.
• 80 GB of HBM memory – delivering up to 2039 GB/s bandwidth, which
is crucial for data-intensive applications.
Understanding the Architecture of a GPU:
• Summing the elements of an array is a classic example of a reduction operation.
Reduction refers to the process of collapsing a sequence of values into a single result,
such as a total sum.
• At first glance, this operation seems inherently sequential. You start with the first
element, add it to the second, take that result, add it to the third, and continue this pattern
until all elements are processed.
• This step-by-step dependency suggests that each computation relies on the previous
one, making it appear unsuitable for parallel execution.
• GPUs overcome this challenge using parallel reduction algorithms. These algorithms
restructure the computation so that multiple partial sums can be calculated
simultaneously and improves performance.
• Understanding this transformation from sequential to parallel is key to leveraging GPU
architecture effectively for data-intensive tasks.

Fig. Sequential Reduction Operation

Parallelizing a Sequential Task:
Summing the elements of an array may initially appear to be a sequential task. Each step
depends on the result of the previous addition, which suggests limited opportunity for parallel
execution.
However, this operation can be restructured into a parallel algorithm using a hierarchical
reduction strategy. This approach enables multiple additions to occur simultaneously.
Consider an array of length 8. In the first step, we perform two-by-two summations in parallel,
resulting in 4 partial sums. In the second step, these partial results are again summed in pairs,
producing 2 new values. Finally, the last two values are summed to obtain the final result.
This hierarchical scheme reduces the depth of computation and allows efficient execution on
parallel hardware such as GPUs.

Fig. Hierarchical Reduction Strategy

Fig. Parallel Reduction Algorithm

Parallel Reduction: Core-wise Execution on GPU
From the GPU’s perspective, we can number the cores as c0, c1, c2, and c3. The reduction
operation proceeds in three clock cycles, each with a specific pattern of core activity:
First Clock Cycle – Full Utilization
All four cores (c0 to c3) are active. Each core performs a pairwise summation on its assigned
input values. This step generates the first set of partial results.
Second Clock Cycle – Selective Processing
Only cores c0 and c2 are active. They process the partial results produced in the first cycle.
These intermediate values must be stored in a memory space that is accessible to both c0 and
c2, such as shared memory or L2 cache.
Third Clock Cycle – Final Aggregation
Only core c0 is active. It performs the final summation using the results computed by c0 and
c2 in the second cycle. These values should also be stored in memory accessible to c0 to ensure
efficient access.

Why Personal Computers Continue to Rely on CPUs

Versatility: Central Processing Units (CPUs) are general-purpose processors capable of
managing diverse tasks, including operating system execution, application control, and system
resource coordination.
Single-Threaded Efficiency: CPUs are optimized for tasks requiring sequential execution,
such as decision-making, background processing, and operations where execution order is
critical.
Software Compatibility: The majority of software ecosystems including operating systems
and general applications are designed for CPU architecture. GPUs, while powerful in parallel
processing, are less suited for these conventional workloads.
Cost Considerations: CPUs offer a more economical solution for everyday computing needs.
A GPU architecture would significantly increase system costs without proportional benefits for
standard tasks.

GPGPU:
GPGPU stands for General-Purpose computation on Graphics Processing Units. It means using
the GPU not just for graphics, but also for general tasks like data processing and simulations.
Since GPUs are built for handling many operations at once, they can speed up performance by
working alongside the CPU.
Instead of staying idle when not rendering graphics, the GPU helps with other tasks, especially
those involving large sets of data or repeated calculations. This makes computing faster and
more efficient.

List and explain the key advantages and disadvantages of using a GPU for general-
purpose computing. Consider aspects like performance, parallelism, energy efficiency,
and programming complexity……?
Advantages:
1. High Throughput: GPUs are built with hundreds to thousands of smaller, simpler
cores optimized for simultaneous computation.
2. Performance Acceleration: This massive parallelism provides drastic speedups
for data-parallel workloads, where the same operation is applied to many different
data elements independently.
3. Ideal Workloads: Tasks like matrix multiplication (critical for AI/Deep Learning),
scientific simulations (molecular dynamics, weather modeling), cryptocurrency
mining, and image/video processing are best suited.
4. Energy Efficiency: While a high-end GPU consumes more total power than a CPU,
it often achieves a higher performance for highly parallel tasks. By completing a
large amount of work in a fraction of the time, the total energy consumed to finish
that specific workload can be less than on a CPU.
5. High Memory Bandwidth Dedicated VRAM: Discrete GPUs use their own
dedicated Video RAM (VRAM), which is typically a very fast memory type
providing significantly higher bandwidth than the system's main RAM (DDR4/5),
allowing the cores to be provide data quickly.
Disadvantages
1. Inefficiency in Sequential Tasks and Branching Poor Sequential
Performance:
The individual cores in a GPU are relatively weak and have slower clock speeds
compared to a CPU's cores.
Tasks that are sequential (one step must complete before the next begins) or require
complex logic and frequent branching (if/else statements) perform poorly, as the
parallel threads often stall, waiting for others to catch up.
2. Programming Complexity Specialized Models: Writing code for the GPU
requires specialized programming models like NVIDIA CUDA or OpenCL. This is
significantly more complex than standard CPU programming, as developers must
explicitly manage memory transfers and thousands of parallel threads to maximize
efficiency.
3. Data Transfer Overhead: To use the GPU, data must be transferred from the
CPU's main memory (RAM) across the system bus (like PCIe) to the GPU's VRAM.
If the computation is too small or the data transfer is frequent, this overhead can
negate any parallel speedup.
4. Programming Complexity Specialized Models: Writing code for the GPU
requires specialized programming models like NVIDIA CUDA or OpenCL. This
is significantly more complex than standard CPU programming, as developers
must explicitly manage memory transfers and thousands of parallel threads to
maximize efficiency.
5. Data Transfer Overhead: To use the GPU, data must be transferred from the
CPU's main memory (RAM) across the system bus (like PCIe) to the GPU's
VRAM. If the computation is too small or the data transfer is frequent, this
overhead can negate any parallel speedup.
Display Technologies:
CRT - CATHOD RAY TUBE.
Cathode-Ray Tube (CRT) is a specialized vacuum tube in which images are produced when
an electron beam strikes a fluorescent surface. It is a technology used in traditional computer
monitors and televisions.
The image on CRT display is created by firing electrons from the back of the tube of
phosphorus located towards the front of the screen. Once the electron heats the phosphorus,
they light up, and they are projected on a screen. The color you view on the screen is produced
by a blend of red, blue and green light.
Components of CRT:
1. Electron Gun: Electron gun consisting of a series of elements, primarily a heating
filament (heater) and a cathode. The electron gun creates a source of electrons
which are focused into a narrow beam directed at the face of the CRT.
2. Control Electrode: It is used to turn the electron beam on and off.
3. Focusing system: It is used to create a clear picture by focusing the electrons into
a narrow beam.
4. Deflection Yoke: It is used to control the direction of the electron beam. It creates
an electric or magnetic field which will bend the electron beam as it passes through
the area.
In a conventional CRT, the yoke is linked to a sweep or scan generator. The
deflection yoke which is connected to the sweep generator creates a fluctuating
electric or magnetic potential.
5. Phosphorus-coated screen: The inside front surface of every CRT is coated with
phosphors. Phosphors glow when a high-energy electron beam hits them.
Phosphorescence is the term used to characterize the light given off by a phosphor
after it has been exposed to an electron beam.
Types Of Deflection:

Electrostatic Deflection: The electron beam (cathode rays) passes through a highly positively
charged metal cylinder that forms an electrostatic lens. This electrostatic lens focuses the
cathode rays to the center of the screen in the same way as an optical lens focuses the beam of
light. Two pairs of parallel plates are mounted inside the CRT tube. The electrostatic deflection
sensitivity of a Cathode Ray Tube is the amount of deflection produced in the electron beam
when a voltage of 1V is applied between the plates.

Fig. Electrostatic Deflection

Magnetic Deflection: Here, two pairs of coils are used. One pair is mounted on the top and
bottom of the CRT tube, and the other pair is on the two opposite sides. The magnetic field
produced by both these pairs is such that a force is generated on the electron beam in a direction
perpendicular to the direction of the magnetic field and the direction of flow of the beam. One
pair is mounted horizontally and the other vertically.
 Fig. Magnetic Deflection

What are the disadvantages of using a CRT (Cathode Ray Tube) monitor compared to
modern display technologies?
Large and heavy design

• CRTs use vacuum tubes and electron guns, making them bulky and difficult to move.
• Modern LCD/LED monitors are slim and lightweight, saving desk space.
High power consumption

• CRTs consume significantly more electricity than LCD or LED monitors.

• This makes them less energy-efficient and more costly to operate.
Lower resolution and clarity

• CRTs struggle with high-definition resolutions compared to modern displays.

• Text and images appear less sharp, which is unsuitable for today’s HD content.
Eye strain and flicker

• CRTs rely on electron beams refreshing the screen, which can cause flickering.
• Prolonged use often leads to eye fatigue and discomfort.
Limited screen size options

• CRTs are impractical for very large displays due to their bulk.
• Modern technologies easily scale to large, flat-panel screens.
Electromagnetic radiation concerns

• CRTs emit low levels of radiation, raising health concerns in prolonged use.
• Modern displays are safer and meet stricter standards.
Types of Display Techniques:
Raster Scan and Random Scan are the mechanism used in displays for rendering the picture
of an object on the screen of the monitor.
The main difference between raster scan and random scan lies in the drawing of a picture where
the raster scan points the electron beam at the entire screen. On the other hand, in the random
scan, the electron beam is guided on just those part of the screen where the picture is to be
drawn.
Raster Scan Display:
In this system, an electron beam is moved across the screen. It moves from top to bottom
considering one row at a time.
As the beam of electron moves through each row, its intensity is turned on or off. Which helps
to create a pattern of spots that are illuminated.
The screen is drawn one line at a time, starting from the top-left corner and moving across to
the right, then down to the next line, until it reaches the bottom-right corner. This whole process
happens very quickly about 50 to 60 times every second. The speed at which the screen is
redrawn is called the frame rate.

Fig. Movement of the beam of electron.

When each scan of the line is refreshed it returns to the left side of the screen. This motion is
known as Horizontal retrace.
As a particular frame ends the beam of electron moves to the left top corner of the screen to
move to another frame. This motion is referred to as Vertical retrace.
Advantages of Raster Scan Display

• Versatility in image rendering: Raster scan can display both text and complex
graphics, making it suitable for general-purpose monitors.
• Cost-effectiveness: Since raster scan technology is based on television CRT principles,
it is cheaper to manufacture compared to random scan systems.
• Ease of implementation: The use of a refresh buffer (frame buffer) simplifies the
process of storing and displaying images.
• Wide adoption: Raster scan is the most common display method in TVs, computer
monitors, and digital screens, ensuring compatibility and standardization.
• Animation and motion support: Continuous refreshing (60–80 Hz) allows smooth
animations and video playback.
• Pixel-based representation: Each point on the screen is a pixel, enabling detailed
images and color variations.

Disadvantages of Raster Scan Display

• Lower resolution: Image quality depends on pixel density; fewer pixels mean less
sharpness compared to vector displays.
• Flickering issues: If refresh rates are low, the display may flicker, causing eye strain.
• Memory requirements: A frame buffer must store intensity values for every pixel,
which demands significant memory for high-resolution images.
• Less precision for line drawings: Straight lines may appear jagged (aliasing) because
they are approximated by pixels.
• Limited scalability: Increasing resolution requires more memory and processing
power, making it less efficient for specialized applications like CAD.
• Dependence on refresh rate: Performance and clarity are tied to how fast the screen
can refresh, unlike random scan systems that directly draw lines.
Random Scan Display:

In Random Scan Display, an electron beam is directed only to the specific part of the screen
where a picture has to be displayed or drawn.

It is also termed as vector display as it displays or draws a picture in the form of one line at a
time. It can draw and refresh lines on the screen of a picture in any sequence.

Advantages

• Precision in line rendering: Random scan systems employ an electron beam to

directly draw lines and curves, resulting in sharp and smooth graphics without
aliasing.
• Efficient memory utilization: Rather than storing pixel data, random scan displays
maintain a list of drawing instructions, thereby reducing memory requirements.
• Suitability for technical applications: Their accuracy makes them particularly
effective for CAD, architectural design, and scientific visualization.
• Absence of rough edges: Direct line drawing ensures clean and precise graphical
output.
• High clarity in geometric representations: Capable of refreshing thousands of short
lines rapidly, maintaining the integrity of technical drawings.

Disadvantages

• High production cost: The complexity of random scan technology renders it more
expensive than raster scan systems.
• Restricted to line graphics: These displays are not well-suited for shaded images,
filled areas, or photographic content.
• Limited versatility: They are unsuitable for multimedia applications such as video
playback or gaming.
• Refresh rate constraints: Excessive line rendering can slow refresh cycles and may
risk damage to the phosphor coating.
LCD is a flat display technology, stands for "Liquid Crystal Display" which is generally used
in computer monitors, instrument panels, cell phones, digital cameras, TVs, laptops, tablets,
and calculators.

It is a thin display device that offers support for large resolutions and better picture quality.

Working of LCD
An LCD contains a backlight rather than the firing electrons at a glass screen, which offers
light to individual pixels arranged in a rectangular grid.
In LCD, liquid crystals are placed between two transparent electrodes and two polarizing
filters. By applying voltage, the molecules twist or align differently, which changes how much
light passes through. This allows the screen to display images.
All pixels have a sub-pixel, red, green, and blue, which can be turned on or off. The display
appears black if all of a pixel's sub-pixels are turned off and appears white if all the sub-pixels
are turned on 100%.
The millions of color combinations can be possible with the help of adjusting the individual
levels of red, green, and blue light.

Different Types of LCD

Twisted Nematic (TN):
Twisted Nematic panels are the most widely produced LCDs due to their affordability and fast
response times. They are especially popular among gamers because they support refresh rates
up to 240 Hz, making them ideal for fast-paced action.
Twisted Nematic displays compromise on color accuracy, contrast ratios, and viewing angles
which makes them less suitable for tasks requiring precise visuals.
In-Plane Switching (IPS):
IPS panels are designed to deliver superior picture quality with vibrant colors and excellent
viewing angles. They are widely used by graphic designers, photographers, and professionals
who need accurate color reproduction.
Compared to TN panels, IPS displays are more expensive and slightly slower in response time,
but they are considered the best option for applications where image quality matters most.

Vertical Alignment (VA):

Vertical Alignment panels strike a balance between TN and IPS technologies. They offer
deeper blacks and better contrast than TN, along with improved color reproduction and wider
viewing angles.
While they are sufficient for everyday use and more affordable than IPS, Vertical Alignment
panels typically have slower response times, which can affect fast-motion visuals. Their cost
is higher than TN but lower than IPS, making them a middle-ground choice.

Fig. Washing Machine Display

Advanced Fringe Field Switching (AFFS):
Advanced Fringe Field Switching (AFFS): AFFS panels are high-performance LCDs known
for their extensive color reproduction and excellent image quality.
They are primarily used in specialized applications such as aircraft cockpits and other
advanced environments where precision and reliability are critical. Though less common in
consumer electronics, AFFS technology represents the cutting edge of LCD development.
 Fig. Aircraft Cockpits
LED Display
An LED Display is a type of screen that uses light‑emitting diodes (LEDs) as the source of
light. Unlike traditional displays that rely on backlighting, LED displays generate light directly
from the diodes themselves, making them brighter, more energy‑efficient, and longer‑lasting.
LED Panels
LED Panels are widely used in outdoor environments such as store signs, advertising boards,
and billboards. Their high brightness and durability allow them to remain visible even in
daylight, and they can withstand varying weather conditions. These panels are often modular,
meaning multiple panels can be joined together to create large display surfaces for commercial
or public use.
Working Principle of LEDs
LEDs are based on the principle of the semiconductor diode. When the diode is forward biased
(switched on), electrons from the conduction band recombine with holes in the valence band.
During this recombination process, energy is released in the form of light photons.
This phenomenon is known as electroluminescence. The color of the emitted light depends on
the energy band gap of the semiconductor material used. For example, different materials
produce red, green, or blue light, and by combining these, full‑color displays can be achieved.
 Fig. LED Advertising Boards
Light Emitting Diode (LED) Display:
LED present many advantages over traditional light sources including lower energy
consumption, longer lifetime, improved robustness, smaller size and faster switching.
However they are relatively expensive and require more precise current and heat management
than traditional light sources.
Advantages of LED

• Durability: LEDs last longer than LCDs or plasma screens, making them reliable for
long-term use.
• Excellent image quality: They produce bright, sharp, and vibrant images with high
contrast.
• Wide viewing angles: The picture remains clear even when viewed from the side.
• Energy efficiency: LEDs consume less power compared to older display technologies.

• Slim and lightweight design: LED monitors are thinner and easier to handle than
traditional ones.

• Flexible design options: They can be used in curved, large-scale, or modular displays.

• Better for gaming: Fast refresh rates and reduced lag make them suitable for gaming
setups.
Disadvantages of LED

• High cost: LED monitors are more expensive compared to LCD or CRT alternatives.
• Eye strain: Brightness and flicker can cause discomfort during long usage.
• Maintenance cost: Repairs and replacements of LED panels are costly.
• Light pollution: Large LED displays can be excessively bright in public spaces.

• Heat management: LEDs require precise current and cooling systems to avoid
overheating.

• Outdoor limitations: Sensitive to weather conditions, requiring extra protection for

outdoor use.
Difference between LED and LCD:

[Link] LED LCD

1. LED has a better response LCD is slower than LED in

time than LCD. terms of response time.
2. LED consumes more Whereas it consumes less
power in comparison to power in comparison to
LCD. LED.
3. LED delivers good picture LCD also delivers good
quality in comparison to picture quality but less than
the LCD display. LED.

4. LED is costlier than LCD. While it is less costly than

LED.
5. LED delivers better color While it also delivers good
accuracy in comparison to color accuracy, we can notice
the LCD. the difference if we compare
these two.
7. LED has a wider viewing While in LCD, the wide-
angle than the LCD. angle decreases with 30
degrees from the center in
the image then the contrast
ratio.

8. LED TVs can be up to 90 LCD Screen size comes in

inches and they are much the range of 13-57 inches.
similar to LCD TVs.

9. LEDs use no mercury and LCDs require mercury for

are therefore their products causing harm
environmentally friendly. to the environment.