MODULE 3

FUTURE HARDWARE SYSTEMS

Moore’s Law:
Moore's Law is the observation that the number of transistors on
a microchip doubles roughly every two years. This trend has led
to exponential growth in computing power and efficiency over
time

To calculate hardware growth using Moore's Law,

you can use a simple equation:
Future Processing Power = Current Processing Power * 2^n,
where 'n' is the number of years divided by the doubling period
(approximately 2 years).

 Moore's Law:
This observation states that the number of transistors on a
microchip doubles approximately every two years.
Moore's Law describes the exponential increase in the number
of transistors on a microchip, roughly doubling every two
years. To calculate hardware growth using this law, you can use
the formula: Future Processing Power = Current Processing
Power * 2^(n/2), where 'n' is the number of years.

 Current Processing Power:

 This is the baseline measurement of hardware performance
you're starting with.
 2^(n/2):
This represents the exponential growth factor. Since Moore's
Law states that the number of transistors doubles every two
years, the exponent is n/2, where 'n' is the number of years you
want to project into the future.
 Exponential Growth:
This doubling leads to a significant increase in processing
power and efficiency.
 Calculation:
To estimate future hardware growth, you can use a simple
exponential growth formula:
Future Processing Power = Current Processing Power *
2^n
 Where 'n' represents the number of doubling periods (years / 2).
Example:

If you have a current processing power of 1 and want to know

the estimated power after 4 years:
1. Calculate 'n': 4 years / 2 years = 2
2. Apply the formula: Future Processing Power = 1 * 2^2 = 1 * 4
=4
 3. Result: The processing power would be estimated to be 4 times
higher after 4 years.

Example:

If you have a processor with a processing power of 3500 and

you want to project the future processing power in 5 years, you
can calculate it as follows:

1. n = 5 years
2. Current Processing Power = 3500
3. Future Processing Power = 3500 * 2^(5/2)
4. Future Processing Power ≈ 3500 * 5.65685
5. Future Processing Power ≈ 19800
Therefore, according to Moore's Law, in 5 years, the processor
would have a processing power of approximately 19800,
assuming no other limiting factors.

Important Considerations:

 Moore's Law is not a physical law:

It is an observation and a projection based on historical trends.
 Doubling time:
 While it's commonly stated as doubling every two years, the
actual doubling time can vary slightly (e.g., 18 months).
 Other factors:
Besides transistor count, other factors like processor speed,
memory capacity, and software advancements also contribute to
hardware growth.
 Current challenges:
As transistors approach atomic scales, maintaining the pace of
Moore's Law has become increasingly challenging.

What is the equation for Moore's Law?

1. Moore did not make an equation for his prediction, but it's easy
to create one. ...
2. Future Processing Power = Current Processing Power ⋅ 2n.
3. In which n = the number of years to develop a new
microprocessor divided by 2.
4. Let's do a quick example. ...
5. Future Processing Power = 3,500 ⋅ 25.

INTRODUCTION TO NON VOLATILE MEMORY

TECHNOLOGY
 Non-volatile memory (NVM) is a type of computer memory that
retains stored data even when power is removed. Unlike volatile
memory, which loses data when power is off, NVM offers
persistent storage for both software and data. Common examples
include flash memory, ROM, and various storage devices like
hard drives and SSDs.

Key Features of Non-Volatile Memory:

 Data Persistence:
The primary characteristic of NVM is its ability to retain data
even when the power supply is interrupted or turned off.
 Various Technologies:
NVM encompasses a wide range of technologies, including
flash memory, ROM, EPROM, EEPROM, and magnetic
storage devices.
 Applications:
NVM is used extensively in various devices and systems,
including smartphones, tablets, embedded systems, and
computer storage.
Common Examples of Non-Volatile Memory:
 Flash Memory: A widely used NVM technology in devices like
smartphones, SSDs, and USB drives. Flash memory stores data
in floating-gate MOSFETs.
 Read-Only Memory (ROM): A type of NVM that contains
permanently stored data, often used for firmware or boot code in
embedded systems.
 Hard Disk Drives (HDDs): Magnetic storage devices that use
rotating platters to store data.
 Solid State Drives (SSDs): Use flash memory to provide faster
access and storage than traditional HDDs.
 Magnetic Tapes: An older type of NVM that uses a magnetic
coating on a tape to store data.

Types of non-volatile memory

Many types of non-volatile memory are in use today to read and
write data to and from enterprise and consumer systems.

NAND flash is one of the most common types of NVM. It

includes several variations, including single-level cell (SLC), or
one bit per cell; multi-level cell (MLC), or two bits per cell;
triple-level cell (TLC), or three bits per cell; and quad-level cell
(QLC), or four bits per cell.

Applications of non-volatile memory

System manufacturers use different types of non-volatile
memory chips for a variety of purposes. For instance, one type
of NVM might store the controller program code for devices,
such as hard disk drives (HDDs) and tape drives. Another type
of NVM is commonly used for data storage in solid-state drives
(SSDs), USB drives and memory cards in digital cameras,
cellphones and other devices.

SSDs typically use a NAND flash to store as blocks. SSDs have

no moving parts, and they are capable of higher performance
than mechanically addressed HDDs and tape, which use a head
to read and write data to magnetic storage media. In SSDs,
NVM improves loading times and system responsiveness.

Some other applications of NVM include the following:

 Archival storage.
 Embedded systems.
 Industrial systems.
 Digital advertising panels.
 Automotive systems.
 AI-enabled systems.
 Medical equipment.

In general, NVM is useful for storing firmware, boot code,

configuration data and other types of critical or historical
information that if lost might cause serious performance
problems in a computing system.

What is the difference between volatile memory and non-

volatile memory?
Volatile memory is a type of memory that requires a continuous
power supply to retain stored data. Prominent examples of
 volatile memory are static RAM (SRAM) and dynamic RAM
(DRAM). Manufacturers sometimes add battery power to
volatile memory devices to support persistent data storage
temporarily in the event of disruptions to the primary power
supply. NVM does not require battery power or a continuous
supply of main power to retain data.

ARCHITECTURE OF NVM TECHNOLOGY

Non-Volatile Memory (NVM) technology utilizes various
architectural approaches to store data persistently, even when
power is removed. These architectures often involve memory
cells, peripheral circuitry for read/write operations, and control
logic for managing the memory system. Key components
include NVM cells, isolation transistors, charge pumps, and
request schedulers.

NVM Cell: Each bit of data is represented by the resistance of a

NVM cell, with low resistance indicating logic '1' and high
resistance indicating '0'. These cells can be based on different
physical principles, such as floating-gate MOSFETs in flash
memory or resistance-changing mechanisms in RRAM.
Peripheral Circuitry: This circuitry handles the read and write
operations of the memory cells. It includes:

 Verify (VF) logic:

Verifies the correctness of write operations.
 Isolation transistors:
 These transistors decouple logic blocks operating at different
supply voltages.
 Charge pumps:
Generate the necessary high voltages for programming and
erasing NVM cells.
 Request scheduler:
Manages long-latency operations, such as destressing, to
minimize their impact on critical paths.
Control Logic: This logic manages the memory system as a
whole, handling tasks like:

 Address decoding: Translating memory addresses to specific

cell locations.
 Cache management: For NVM-based cache architectures,
managing the cache to optimize performance and energy
consumption.
 Error correction: Detecting and correcting errors in the stored
data.
Specific NVM Technologies: Different NVM technologies have
their own architectural nuances. For example, RRAM utilizes
various resistance-changing mechanisms, while PCRAM relies
on phase transitions to store data.

Examples of NVM Architectures:

 NVM Duet:
ACM Digital Library proposes a unified architecture for both
working memory and persistent storage, optimizing for speed in
working memory and consistency/durability in persistent
storage.
 Intermittent Computing Architectures:
Architectures like those proposed in IEEE
Xplore and arXiv focus on efficiently utilizing limited energy
to back up data during power failures in intermittent computing
scenarios.
 NVM Main:
This architectural-level simulator, introduced in IEEE Xplore,
allows designers to model main memory designs with both
DRAM and emerging NVM technologies.
Fig: Architecture of Proposed NVM
Advantages of Non-Volatile Memory
 Non-volatile memory is more reliable. Basically, it has no

mechanical parts, so they are more durable to perform.

 It can store the required data permanently. Since there are

no physical components needed for any required operation,

they also require extremely little energy and don't require a
continuous power source in order to store the data for the
overall system process.
 It is a very low-cost memory.

 It also helps in the booting of the initial operating system.

They are much lighter and easier to move from one

location to another as per requirements.
 Disadvantages of Non-Volatile Memory
 Non-volatile memory is a much slower process as per the

speed.
 Non-volatile memory can easily destroyed and misplaced

as per requirement.
 The time to rewrite required data is very much large since

every byte is processed individually and technically.

 It can only read the required data.

SCOPE OF NVM TECHNOLOGY

1.Embedded Systems and Microcontrollers:

 NVM is essential for storing code, firmware, and configuration

settings in microcontrollers (MCUs) and low-power System-on-
Chips (SoCs).
 It's vital for IoT devices, wearables, and edge-AI applications
where low power consumption and data persistence are crucial.
 Emerging NVM technologies like MRAM, PCM, and RRAM
are gaining traction for embedded applications.
2. Consumer Electronics:
 NVM (especially flash memory) is widely used in consumer
electronics like smartphones, tablets, and PCs for data storage
and booting.
 3D NAND flash memory, used in SSDs and other storage
devices, offers high density, speed, and low power consumption,
 making it ideal for consumer applications, says
MarketsandMarkets.
 NVM is also found in applications like streaming media devices,
smart speakers, and wearables.
3. Industrial and Archival Storage:
 NVM technologies can be used for archival storage and data
backups.
 They are also suitable for industrial systems where data
persistence is crucial for critical operations.
4. Data-Centric Computing and AI:
 Emerging NVM technologies like MRAM, PCM, and RRAM
can be used to build high-performance storage-class memories
that can bridge the performance and density gap between
memory and storage, according to IEEE Xplore.
 They can also be used for synaptic functions and security
elements in AI-enabled systems.
5. Automotive and Medical Applications:
 NVM plays a role in automotive systems like airbags and
infotainment systems.
 It can also be used in medical equipment for storing patient data
and device configurations.
 6. Key Trends:
 Technological advancements, especially in 3D NAND and
emerging NVM technologies, are driving the growth of the
NVM market.
 The demand for NVM is increasing due to the growing use of
smartphones, AI-enabled devices, and IoT devices.
 Emerging NVM technologies are being developed to address the
limitations of flash memory and offer better performance,
density, and power efficiency.

EMERGING NON VOLATILE MEMORY

TECHNOLOGIES

Emerging non-volatile memory (NVM) technologies offer

potential replacements for existing memory solutions,
particularly in low-power applications and data-centric
computing. These technologies include Magnetic Random-
Access Memory (MRAM), Spin-Transfer Torque RAM (STT-
RAM), Ferroelectric RAM (FeRAM), Phase-Change Memory
(PCM), and Resistive Random-Access Memory (RRAM).
Here's a more detailed look at some of these technologies:

1. Magnetic Random Access Memory (MRAM):

 MRAM utilizes magnetic materials to store data, offering fast
read/write speeds and high endurance.
 Spin-Transfer Torque RAM (STT-RAM) is a type of MRAM
that uses spin-polarized current to manipulate the magnetic
field.
 Challenges include the need for higher current densities and
integration challenges with CMOS technology.
2. Phase-Change Memory (PCM):
 PCM relies on reversible phase transitions (amorphous to
crystalline) within a chalcogenide material to store data.
 It offers high density and scalability, but faces challenges with
endurance and power consumption.
3. Resistive Random Access Memory (RRAM):
 RRAM utilizes the resistive switching characteristics of
materials to store data.
 It offers high density and good endurance, but needs further
optimization for speed and power efficiency.
4. Ferroelectric Random Access Memory (FeRAM):
 FeRAM stores data using the polarization of ferroelectric
materials.
 It boasts fast read/write speeds and low power consumption, but
faces challenges with scaling and material compatibility.
5. Other Emerging Technologies:

 Ferroelectric Field-Effect Transistor (FeFET):

Utilizes the ferroelectric material as a gate dielectric in a
transistor, offering high performance and low power
consumption, according to an article in ScienceDirect.com.
 Carbon Nanotubes (CNT) based memories:
Offer high density and fast switching speeds, but require further
development for reliable storage.
 Polymeric Ferroelectric RAM (PFRAM):
Uses polymer materials for ferroelectric memory, potentially
offering high density and flexibility, according to a report by
Coughlin Associates and Objective Analysis.
 SONOS memory:
A type of flash memory with improved charge retention and
scalability, according to a study on Springer.

Applications and Trends:

 These technologies are being explored for various applications,
including embedded systems, IoT devices, and high-
performance computing.
 Their potential to simplify memory hierarchies, reduce power
consumption, and enable novel architectures makes them
attractive alternatives to traditional memories.
 Emerging NVMs are also being investigated for applications
beyond memory, such as neuromorphic computing and hardware
security.
Challenges and Future Directions:
 While promising, these technologies still face challenges in
terms of scalability, endurance, and integration with CMOS
technology.
 Further research and development are needed to address these
challenges and unlock the full potential of emerging NVMs.
 The focus is on developing high-density, low-power, and cost-
effective NVM solutions that can meet the demands of future
computing systems.

Magnetic Random Access Memory (M-RAM)

MRAM stands for magneto resistive random access
memory and is a non-volatile type of RAM.
Magnetic state refers to the electrical resistance of a metal
when it is placed in a magnetic field. The MRAM uses
magnetic states and magnetisation direction in a ferromagnetic
material(a material which is highly susceptible to
magnetization to store data bits) for storing data and magneto-
resistance for reading the stored data; instead of using the
electrical bits to store your data. Information is stored using
electron spin. It is sometimes called as “ideal memory”. As it
is a non-volatile type of memory, no refresh is required to
retain the data and the power consumption is very less.
Working Principle:
MRAM uses MTJ (Magnetic Tunnel Junction) to store bits of
data. MTJ comprises of two layers of ferromagnetic materials:
the reference layer and the free layer. An insulation layer is
used to separate these two. Storing of a logical element (0 or
1) takes place by changing the resistance of an MTJ. The
resistance depends on the relative spin orientations of two
ferromagnetic layers, and can be of two types: high resistance
or low resistance.
The reference layer is used to maintain the magnetic direction,
while the free layer, can have its direction changed by using
magnetic fields or by applying polarized currents.

 A logical 0 is stored when both layers have same direction,

hence MTJ has a low resistance.

 A logical 1 is stored when both layers have different

direction, hence MTJ has a high resistance.

Why MRAM should be used?

DRAM has the advantage of being cheap, but is comparatively
slow and data is lost when power is off.
SRAM, on the other hand is faster than DRAM. But it can cost
upto 4 times as much as DRAM and data is lost when power is
turned off.
FLASH memory saves data when the power is off, but the
process is too slow and consumes a lot of power.
MRAM potentially combines the density of DRAM and the
high speed of SRAM and the non-volatility of FLASH
memory or hard disk, and all this is done using very less
power. MRAM can resist high ionizing radiation, can operate
in extreme temperature conditions and thus is very suitable for
aerospace applications when combined with suitable CMOS
technology.
Advantages of M-RAM:
M-RAM offers several advantages over traditional memory
technologies like DRAM and SRAM. Firstly, M-RAM is non-
volatile, which means it does not require power to retain stored
data. This feature makes it ideal for use in portable devices
that require long battery life. Secondly, M-RAM is extremely
fast, with read and write speeds that are comparable to DRAM
and SRAM. This feature makes it ideal for use in high-speed
memory applications like cache memory. Finally, M-RAM is
highly durable, with the ability to withstand extreme
temperatures, radiation, and physical shock. This makes it
ideal for use in harsh environments like outer space and
military applications.
Potential Applications of M-RAM:
M-RAM has the potential to revolutionize the way we store
and access data, with several potential applications in various
industries. One of the most promising applications is in
portable devices like smartphones and tablets, where M-
RAM’s non-volatile nature and fast read and write speeds can
significantly improve battery life and performance. M-RAM
can also be used in cache memory applications, where its fast
read and write speeds can significantly improve computer
performance.
Another potential application of M-RAM is in automotive
electronics, where its high durability and ability to withstand
extreme temperatures can significantly improve reliability and
performance. M-RAM can also be used in military
applications, where its high durability and ability to withstand
radiation and physical shock make it ideal for use in harsh
environments.
Applications of M-RAM:
MRAM has a potential in all memory applications in these
devices:

 Digital Cameras
 Cellular Phones
 MP3
 HDTV
 Laptops

Spin-Transfer Torque RAM (STT-RAM)

Spin-Transfer Torque RAM (STT-RAM), also known as STT-
MRAM, is a non-volatile memory technology that uses the spin-
transfer torque phenomenon to switch the magnetic orientation
of a memory cell. This technology offers advantages like higher
density, low power consumption, and faster speeds compared to
traditional memory types like DRAM and SRAM.
Here's a more detailed explanation:
 Spin-Transfer Torque:
The core principle involves using spin-polarized current to
exert torque on the magnetic moments of a memory cell's "free"
layer, effectively flipping the magnetization and storing data.
 Non-Volatile:
Unlike DRAM and SRAM, STT-RAM retains data even when
power is off, making it suitable for applications requiring
persistent storage.
 Advantages over other technologies:
 Density: STT-RAM can achieve higher storage densities than
traditional technologies.
 Power Consumption: It offers lower power consumption,
especially when compared to charge-based memories like DRAM
and SRAM.
 Speed: STT-RAM can provide fast read and write speeds,
potentially challenging the performance of SRAM.
 Challenges:
 Current Requirements: While promising, STT-RAM currently
requires a relatively high current to flip the magnetic orientation,
which is a research focus.
 Thermal Stability: Maintaining stability while switching quickly
within a given voltage pulse is a challenge.
 Applications:
STT-RAM has potential in various applications, including
embedded memory, flash memory replacement, and even in-
memory computing (CiM).
 Ferroelectric RAM
Ferroelectric RAM (FRAM) is a non-volatile memory
technology that stores data using ferroelectric materials within a
capacitor. It offers advantages like fast read/write speeds, low
power consumption, and high endurance, making it suitable for
applications where data retention and reliability are crucial.
Key Features and Characteristics:
 Non-Volatile: Data is retained even when power is off, similar
to flash memory.
 Fast Speeds: Offers faster read and write speeds compared to
flash memory.
 Low Power: Consumes less power than EEPROM and NOR
flash.
 High Endurance: Can withstand millions or billions of write
cycles.
 Radiation Hardness: Some FRAM technologies are designed
to be radiation-hard, making them suitable for space
applications.
How it Works:
 A FRAM cell typically consists of a ferroelectric capacitor and a
transistor.
 The ferroelectric material within the capacitor stores data by
maintaining a specific polarization state.
 An external voltage is applied to change the polarization state of
the ferroelectric material, effectively storing or changing data.
Applications:
 Data Logging: Storing critical data in mission-critical systems,
such as programmable logic controllers (PLCs), medical
devices, and industrial automation.
 Space Applications: Radiation-hard FRAMs are used in
satellites and other space-based systems.
 Other Applications: Suitable for applications requiring fast
read/write speeds, low power consumption, and high reliability,
such as in portable devices, IoT sensors, and automotive
systems.

Phase-Change Memory (PCM):

Phase-Change Memory (PCM) is a non-volatile memory
technology that stores data by changing the physical state of
a material, typically a chalcogenide, between a high-
resistance amorphous state and a low-resistance crystalline
state. This technology is attractive for non-volatile memory
applications due to its fast switching speed and the ability to
store multiple bits per cell.
 Key features of PCM:
 Non-volatile: Retains data even when power is off.
 Fast switching: Can quickly transition between crystalline and
amorphous states.
 High density: Can store multiple bits per cell, leading to higher
storage capacity.
 Scalability: Can be fabricated at nanoscale dimensions.
 Long data retention: Stored data is typically retained for many
years.
 Single-bit alterability: Can change individual bits, unlike NAND flash
which requires erasing larger blocks.
 High read speed: PCM offers fast read times, comparable to DRAM
and single-bit NOR flash.
Applications of PCM:
 Non-volatile memory: Replacing flash memory in various
applications.
 Storage-class memory: High-performance memory for data centers
and other applications.
 In-memory computing: Using PCM as a memory unit for
computation.
 Neuromorphic computing: Utilizing PCM as a nanoelectronic
synapse for neuromorphic systems.

Resistive Random Access Memory (RRAM):

 Resistive Random Access Memory (RRAM) is a type of non-
volatile memory technology that stores data by changing the
resistance of a solid-state material, often a memristor,
between high and low resistance states. This technology
offers potential for high density, high speed, and low power
consumption, making it a promising candidate for future
memory applications.

Here's a more detailed explanation:

 Non-Volatile Memory:
RRAM is non-volatile, meaning it retains data even when power is
removed, unlike dynamic random access memory (DRAM).
 Memristor:
The core of an RRAM device is a memristor, a material whose
resistance changes based on the applied electrical current or
voltage.
 Resistance Switching:
 RRAM stores data by switching between high resistance and low
resistance states, representing 0 and 1 respectively.
 Advantages:
RRAM offers several advantages over traditional memory
technologies, including:
 Scalability: RRAM can be scaled to very small dimensions, potentially
enabling higher storage density.
 Speed and Low Power: RRAM can switch between states quickly and
with relatively low energy consumption.
 Long Retention: RRAM can retain data for extended periods without
power.
 CMOS Compatibility: RRAM is compatible with existing CMOS
manufacturing processes.
 Applications:
RRAM is being explored for various applications, including:
 Next-generation non-volatile memory: RRAM is seen as a potential
replacement for flash memory.
 Embedded memory: RRAM can be integrated into various devices,
including mobile phones and computers.
 Neuromorphic computing: RRAM's ability to switch states and store
data makes it suitable for building neuromorphic hardware.
 Challenges:
Despite its promise, RRAM faces some challenges, including:
 Endurance: RRAM devices have a limited number of write/erase cycles
before they wear out.
 Reliability: Ensuring reliable and consistent resistive switching behavior
is crucial for RRAM performance.
 Scaling: While RRAM can be scaled, maintaining performance at
extremely small dimensions is a challenge.