Multi Core ARM Processors in Mobile Devices

Varsha Krishnan
Computer Science Department
San Jose State University
San Jose, CA 95192
408-714-9742
[email protected]

ABSTRACT opinion that a simplified instruction set gives higher performance

In this era of a highly competitive market of mobile devices, ARM when used along with a microprocessor architecture implementing
processors are in demand for their cutting edge power efficient the instructions with fewer cycles per instruction. A British
processors. These RISC based processors, originally designed by company ARM Holdings used the RISC instruction set
Acorn Computers Ltd. , have evolved into a family of architecture to develop ARM (formerly Advanced RISC Machines
microprocessors, widely used in mobile electronic devices like and Acorn RISC Machine). It is a 32 bit instruction set
phones, PDAs and multimedia players. architecture, originally created by Acorn Computers for its
personal computers. It was later used in the BBC Micro series of
This paper gives a brief insight into the RISC architecture of computers for their co-processor modules. In the late 1980s,
ARM processors and their traits, which have made them very Apple computer and VLSI technology, worked with Acorn, on
popular for use. It then briefly outlines the evolution stages of newer versions of ARM core and in 1990, Acorn spun off the
ARM processors from its nascent ARM1 to the Cortex A15 cores. design team to form Acorn RISC Machines Ltd., which later
The paper then highlights the need and design of multiple core became ARM Ltd.
ARM processors. It explains the features of an example multi core
arm processor, the ARM Cortex A9 MP Core processor, in detail. 2. NEED FOR ARM
It gives an insight into how the mobile device companies harness Most of the embedded systems applications utilize the ARM
the multi core features of these processors and help to provide processors since they provide a unique blend of features which
numerous advantages in terms of internal device technical best suit embedded applications. Firstly, compared to many
capabilities as well as the external user interface experiences. general purpose processors, ARM cores are simple and hence can
Using the Cortex A9 as the base, it then discusses about new be manufactured with fewer numbers of transistors. They also free
features of the newer multi core processors released and their up a lot of space on the chip for application specific macrocells.
applications. It concludes by outlining the planned features of Typically, an ARM chip consists of a digital signal processor,
upcoming processors, with an eye on how they will revolutionize several peripheral controllers, a small area of on- chip memory
the industry further. and an ARM core. Secondly, both the ARM ISA and pipeline
design aim at reducing energy consumption, which is a pressing
need in mobile embedded systems. Thirdly, there is a lot of
1. INTRODUCTION TO PROCESSORS flexibility in building application specific ARM-based processors
A microprocessor is a single integrated circuit which executes the since the integer pipeline is the only compulsory component of
functions of a central processing unit. It accepts digital data, the ARM processor whereas all other components like the
processes them based on instructions in the memory and outputs memory management unit, caches, floating point unit and other
the result. Since it is a programmable device, it can be used for co-processors are optional. Finally, the ARM processors deliver
multipurpose applications. Integrating the central processing unit high performance for embedded applications, being small and low
on a single chip significantly reduced the processing power and power. A typical example is the PXA225 XScale processor that
also helps in manufacturing them in large quantities. Over the runs at 400 MHz, which uses fifty times lesser energy than the
years, microprocessors have undergone many design changes, Pentium 2 processor running at 300 MHz .
making them faster and also equipped with a variety of
ARM has been licensing the ARM architecture to other
functionalities to suit the applications they are used in.
companies for manufacturing silicon-on-chip products and also
Intel introduced the first 4 bit microprocessor in 1971. Over the ARM-based CPU’s. The two licenses ARM provides are the
years, it was succeeded by 8 bit designs, 12 bit designs, 16 bit Implementation license and Architecture license. The
designs, 32 bit designs and finally the 64 bit designs. Processors implementation license presents the complete information needed
from Intel, AMD etc. used x86 family of instruction set by the licensee to design and manufacture the integrated circuits
architectures (with many additions and extensions added to it over having ARM processor cores .There are two types of cores which
the years, with full backward capability). These architectures are are licensed by ARM. They are the hard cores and the soft cores.
primarily Complex Instruction set computing (CISC) designs, A hard core used in a specific manufacturing process and is
where single instructions can implement numerous lower level optimized whereas a soft core is not fully optimized and is flexible
operations. Soon, an opposing strategy called Reduced Instruction since it can be utilized in any process. The architecture license
Set Computing (or RISC) originated in the 1970s based on the allows the licensee to develop their own processors.
ARM processors are used in mobile phones, digital media, hand- 3.1 Single Core and Multi Core Processors
held gaming systems, calculators, computer hard drives and
personal digital assistants (PDA). ARM architecture is licensed A single core processor is a processor which has only one core.
for use by companies such as Cirrus Logic, Intel, Apple, LG, Hence it can run one process at a time. In some cases, it can start a
Microsoft, Nintendo, Nvidia, Sony, Samsung, Sharp, Texas new process before the earlier one is complete. Earlier all the
Instruments and many others. processors were all single core.

3. ARM ARCHITECTURES
The ARM classic processors have been available for 15 years.
They consist of 3 processor families, the ARM 7, ARM 9 and the
ARM 11 family which are still in use. From the mid-90s, the
ARM7TDMI processor core was very successful, being adopted
widely in the mobile industry and is still the highest shipping 32
bit processor on the market even today. From it, a large roadmap
of products grew in the real-time embedded space. In the
embedded space, there are 2 main families today. In embedded
microcontrollers, where cost is at a premium, the Cortex- M
family is being used. The other family is the Cortex-R family,
which is at the upper end and provides high performance and
throughput, in addition to precise timing properties. Hence
making it apt for deeply embedded timing critical applications. On Figure 2. Single core CPU chip.
the application side, there is the Cortex-A family. These provide
scalable higher performance in applications requiring platform
operating systems, for example Linux. They have extended A multi core CPU is a processor which has 2 or more cores.
instruction sets and sophisticated memory management for Each core of the chip behaves as a separate computer and
multimedia processes. Since the classic ARM 11MP processor, accordingly executes commands. The cores are all part of the
ARM has support for multi core systems. Accordingly, all Cortex- same chip but work together in parallel and are self-governing. A
A processors have support for multi core systems, enabling them multi core processor with two independent processors is called a
to unleash higher performance and power points. dual core processor. Whereas, a multi core processor with four
independent processors is called a quad core processor. The cores
The following graph gives a quick view on the performance and
may or may not share caches and they may also execute memory
capability of the different ARM processors.
passing or shared memory inter core communication methods. If
the multicore systems include identical cores, they are called
homogeneous multi core systems. Else if different, they are called
heterogeneous multi core systems. Multi core processors are used
in different application domains like embedded, network, digital
signal processing, general purpose and graphics [6].

Figure 3. Multi core CPU chip.

Figure 1. Performance, Functionality vs Capability of different The present day mobile devices perform a variety of tasks
ARM Processors. including mobile gaming, SMS texting, browsing on the web,
playing audio and video and also numerous location based
services including GPS navigation, on-the-go reservations etc. Example 1: While browsing the New York Times Website, the
There were all tasks that were earlier handled by traditional utilization of each core of the quad core CPU system is shown in
desktop PCs and now are being executed on the mobile devices the figure below [4].
with assistance from high speed mobile services and Wi-Fi
networks. The current and next generation smart devices are
further being required to do more intensive tasks like playing of
HD videos, recording of videos in HD, editing of videos, flash
based games, games with high definition and realistic visuals,
audio conferencing and video conferencing.
To perform all these applications with a single core processor is
not efficient. When using single core CPU, running all these
applications simultaneously on it will result in slowing down of
these applications or a decrease in quality of experience provided
by these applications. Engineers try to use several methods to
improve the CPU performance to make the experience better.
These include increasing core voltage and frequency, using larger
cores, using smaller and faster semiconductor processes or even
using larger caches on the die. However, these may not be very
effective as each of these have its own limitations. Increasing the
CPU size or using larger cache can help to increase the
performance to a certain extent, beyond which heat dissipation
and thermal issues increase, limiting the extent of increase in the
sizes of these [5]. On the possibility of increasing frequency and
voltage, these tend to increase the power consumption of the
devices. If the devices consume more power, bigger cooling
solutions will be needed and this would need an increase in the
device size itself. For mobile devices maintaining the size small is
a big benefit. Even if engineers manage to get higher performance
by increasing voltage and frequency, the battery life will be highly
affected by the increased performance. Hence increasing the
frequency is not a good option for the long run [1],[2].
Multi core CPU cores improve performance of cache snoop
(meaning signals between different CPUs travel smaller distances Figure 4. CPU usage percentage vs Time of one browser in
and hence have lesser degradation). Also multiple CPU cores on quad core
same die allow cache coherency circuitry to operate at higher
clock rate than if signals travel off chip. Multi core chips also
allow higher performance at lower energy. This is an important As seen above, the browser utilizes each of the cores of the quad
factor especially on mobile devices running on batteries. Also core CPU. This in-turn helps to give the user a faster web
since each core is energy efficient, the chip as a whole becomes browsing experience when compared to a dual core or single core
more efficient than having a single core. based mobile device.
3.2 Illustration of Advantages of Multi Core Example 2: If the same web browser has multiple tabs open, then
the tabbed browser uses much more CPU processing. Even the 4
Processor in Mobile Devices cores of a quad core CPU may be saturated. Hence the effect, if
Multi core CPU device provide the user with a desktop PC style this was to be executed on a single core CPU mobile device
Web browsing experience. Browsers like Mozilla Firefox and would be drastic. The following figure illustrates this case.
Google Chrome are multithreaded and hence have the capability
to spawn multiple processes. If there are multiple tabs, each tab is
a separate process, which in turn manages its own set of threads.
 A8, they used an in-order pipeline. Hence these processors were
unable to avoid the issues due to cache misses or branching. In
addition, to avoid branching penalties, there is also support for
speculative branch predictions.
Mobile companies use the multi core Cortex-A9 microprocessor
and add certain features to enhance the usage of their
functionalities. Say, we have a dual core Cortex-A9
microprocessor as shown below [3].

Figure 5. CPU usage percentage vs Time of one browser with

multiple tabs in quad core.

4. OVERVIEW OF A WIDELY USED

MULTICORE ARM PROCESSER IN
MOBILE SYSTEMS
One of the most widely used multicore ARM processors today is Figure 6. Dual core Cortex-A9 microprocessor
the ARM Cortex-A9 MPCore processor. Its popularity is due to
the fact that it has several features which are new or are huge
developments when compared to the earlier generation ARM Companies develop intelligent frequency scaling and dynamic
processors. Therefore, it helps to provide increased performance voltage algorithms, at both software and hardware levels, to
scalability and increased control over power consumption. This manage the power of the CPU cores. They ensure both the cores
has made it ideal for use in several areas including mobile are always functioning at the optimal frequency and voltage
applications, auto-infotainment, high performance networking, set levels, so that they provide the required performance at the lowest
top boxes and enterprise applications. power. And these also ensure that when the system is idle, the
cores are operating in low power mode.
The Cortex-A9 architecture has evolved from the ARM Cortex-
A8 processor. This processor implements the ARMv7 instruction 5. FUTURE MULTICORE MOBILE
set architecture. It is scalable up to 4 cores with its advance PROCESSORS, THEIR FEATURES AND
MPCore technology. The CPU cores are optimized to operate at
1GHz and have the ability to scale up to higher frequencies. It APPLICATIONS
has a 32KB instruction cache and a 32 KB data cache per core, ARM is constantly developing its architecture and has released
with both cores sharing a 1MB L2 cache. This 1MB L2 Cache other multicore ARM architecture processors every year. The
helps to even load the entire memory footprint of a browser onto ARM Cortex-A15 MPCore is 40 percent more powerful than
it and hence gives a fast web browsing experience. Cortex- A9 core with the same number of cores at the same speed.
One of the unique features of this processor is the presence of The Cortex A-15 is also built on ARMv7 architecture. This
speculative out of order execution supported by dynamic 8 stage processor gave rise to the big.Little technology. This is a power
pipeline. This particularly helps the processor to reorder optimization technology where high performance but power
instructions dynamically and avoid stalls due to resource conflicts hungry ARM CPU cores are combined with slower, low power
or instruction latencies. This helps to increase performance to a ARM cores, to give a multicore processor which can adjust to
great extent. In older generation processors including the Cortex dynamic computing needs but use lesser power. This finds its use
in high end mobile devices as it ensures mobile SOCs can be This will enable them to retain the advanced graphics and visual
designed with new levels of peak performance, with all day effects, to give the user an experience similar to playing the games
battery life, expected by the consumers. ARM first announced on a console. In short, the multi core CPUs will help to further
Cortex-A7 to be architecturally compatible with Cortex-A15. revolutionize the mobile systems and will pave the way for
Samsung Exynos used this technology, by creating an SOC developing significant applications that will catapult the user
combining a 1.6 to 1.8GHz quad core Cortex-A15 with 1.2GHz experience to a whole new level.
quad core Cortex-A7 [7].
ARM then released the Cortex-A17. This is considered to the 7. REFERENCES
right successor for Cortex-A9. ARM claims Cortex-A17 is 60
[1] Berkel, C.H. Multi-Core for Mobile Phones in Proceedings
percent more powerful than Cortex-A9 and also 20% more energy
of Design Automation Test Europe Conference, (2009),
efficient. This also supports the big.Little multi core architecture
1260-1255.
and are planned to be used in mid-range mobile devices releasing
in 2015. [2] Blake, G., Dreslinski, R.G.,and Mudge, T. 2009.A Survey of
Mulitcore Processors. IEEE Signal Processing Magazine, 26-
6. CONCLUSION 37.
[3] NVidia.2010. The Benefits of Multiple CPU Cores in
The mobile processors are faced with power and performance MobileDevices.http://www.nvidia.com/content/PDF/tegra_w
challenges, like the notebook and desktop processors faced few hite_papers/Benefits-of-Multi-core-CPUs-in-Mobile-
years ago. The single core processors are being stretched by the Devices_Ver1.2.pdf.
new applications like streaming video, 3D games, HD video
playback etc. Inevitably all mobile processors will have multi core [4] NVidia.2011.The Benefits of Quad Core CPUs in Mobile
processors, in order to stay within power budgets and to further Devices. http://www.nvidia.com/content/PDF/tegra_white_p
increase the performance. The widely used mobile software apers/tegra-whitepaper-0911a.pdf.
already support multithreading and multitasking. Both Android [5] Tudor, B.M., and Teo, Y.M. On Understanding the
and iOS have also added several features which help to improve Energy Consumption of ARM-based Multicore Servers in
the ability of the operating systems to leverage the processing Proceedings of SIGMETRICS’13,(June 17-21, 2013),267-
capability of multi core CPUs. Mobile browser like the android 278.
browser supports multi-threading and tabbed browsing. In
addition, they harness the tremendous power offered by multi core [6] Pol, R.U., and Dhukate, B.S.2013.Energy Preservation
CPUs, to provide a faster web browsing experience. In addition, of Mobile Handset’s using Multicore Technology.Blue
all popular games on PC as well as Xbox and PS consoles are Oceans Research Journals, 2, 12.
being ported to mobile environment. They will now transition [7] Merritt, R. 2008. CPU Designers Debate Multi-core
from using the multi core processors in the PC to multi core CPUs
Future.EETimes.[Online].Available:http://www.eetime
in the mobile processor.
s.com/showArticle.jhtml?articleID=206105179.