A mobile processor will fuse the graphics processing unit (GPU) and CPU at its most basic level, case in point the ARM Cortex-A8 processor core with a ...
Why ARM is ahead of Intel in the smartphone microprocessor market
By: Marian K. Musumhiri M00460322
Introduction Microprocessor OMAP
3530 vs. N330 Recent Studies Intel Innovations Recommendations Conclusion Bibliography
For decades Intel has dominated the microprocessor market, but in the Smart Phone realm, ARM is at the fore with consumers dictating that low power and high execution processors be a major part of Smart Phones. To achieve these architectural prerequisites of Smart Phones, both organizations have created a number of processor architectures to satisfy the market. In this report several Intel and ARM microprocessors are contrasted with deference to their architectures and their ability to execute particular executions. Clock rates, energy, value, and die size were considered and Intel chips had better crude performance while ARM chips had excellent energy proficiency rates. Recommendations as to how Intel should better their product were also proffered.
Figure1: According Gillooly (2010)
In a smartphone the graphics card, CPU, Random Access Memory (RAM) and storage all socket into a motherboard, which has distinct chips to control audio and input/output capacities referred to as socket on a chip (SoC). Mainly because there is no luxury to scatter functionality between diverse chips in a smartphone framework as it is neither energy proficient nor is it a choice as the majority of the space is used up by the energy source (battery). A mobile processor will fuse the graphics processing unit (GPU) and CPU at its most basic level, case in point the ARM Cortex-A8 processor core with a PowerVR SGX 535 graphics chip found in the Hummingbird processor from Samsung (Galaxy phones and Galaxy Tab tablet) and the A4 processor from Apple (iPad and iPhone 4).
The ARM and Intel chips are too dissimilar to be compared easily so Hegde et al used the Beagle Board based on the OMAP 3530 processor from Texas Instruments and theN330 Intel Atom desktop computer. The OMAP 3530 was selected and it has a NAND of 256, 256MB DRAM running at 600MHz and utilizes energy administration methods like dynamic power switching; dynamic voltage and frequency scaling and/or particularly low stand-by energy. For Intel, Hedge et al utilized a N330 desktop with two Atom cores clocked at 1.6 GHz with an Intel GMA 950 for integrated graphics; an Intel 945GC Express Chipset; a 533 MHz system bus; an S-Video connector; gigabit Ethernet; a 6-channel high definition audio; and a single DIMM socket that can back up to 2 GB of DDR2 667 or 533 memory. In the experiments both processors were running Linux 2.6.28 and the N330 utilized the Angstrom distribution for Beagle Board and Ubuntu 9.04 because Linux has a far reaching accessibility, it is a dominating operating system for high execution consumer electronics like mobile phones and GPS systems Two integer-based benchmarks to provide knowledge into regular text-based applications like word-processing and two floating-point-based benchmarks to be characteristic of logical processing and audio processing were picked i.e. Fhourstone, dhrystone, whetstone and linpack.
The results reinforced that the Intel Atom has a superior processing power and that the ARM outline had a noteworthy energy optimization characteristic. It is clear the external 945GC Express Chipset is the cause for the demise of Intel Atom in the energy efficiency stakes because the Atom processor is generally low power consuming 2W of energy. Tragically the 945GC Express Chipset is used for multimedia programs, which means that if the Cortex-A8 is combined on a SoC device such as the OMAP3530 with a DSP, significant advantages will be realized. Intel aims to save energy by combining the 945GC chipset with the chief processor chip and migrating to 28nm innovation. Should they be able to lower the resulting processor chip to under 1.5W, OMAP3530 will have some serious competition as they will both have a similar performance-per-power.
To evaluate the Atom (Intel), Krait (Qualcomm), Tegra (Nvidia) and Exynos 5 Dual (Samsung) SoCs , Sebastian Anthony searched for the precise energy tracks that supply both the GPU and CPU on the SoC he had to track down the power circuitry for the SoC. Luckily an Intel engineer with a kit from National Instruments (NI USB-6289) which correctly measured the changes in energy usage. Figure 2: Acer W510 and a Surface RT having their CPU and GPU power tested according to Anthony (2013).
Specifically the Atom Z2760 (32nm, Clover Trail, found in the Acer W510 by Intel); Tegra 3 (TSMC 40nm, found in Microsoft’s Surface RT by Nvidia); and Krait APQ8060A (TSMC 28nm, found in the Dell XPS 10 by Qualcomm) were examined. Figure 3 is according to Lal Shimpi (2013).
The outcome was that both the Tegra 3 and Krait were outshone in the performance and energy utilization stakes by the Clover Trail Atom chip. The 32nm Exynos 5 by Samsung, based on the Cortex-A15 which consumes energy at a very good rate with a TDP that goes up to 8W from ARM was found to outclass the Intel x 86 frameworks. However Krait, Tegra and Clover Trail use under 2W during load hence the Cortex-A15 will have challenges penetrating the smartphone market. The Cortex-A15 has too much energy consumption hence Krait and Apple modified it and also the Tegra 4 will be a quad-core version of the cortexA15 outline and thereby be effective in smartphones.
The software aspect of MIDs is now just as crucial as the hardware and the Android system is now ported to the x86 architecture, but 25% of its older apps which require more graphics like games will not operate on x86 smartphones. However all is not lost as such problematic apps can be recompiled so as to work on the x86 smartphones and Intel is targeting for a minimum of 90% of the apps to be compatible with their hardware by doing software translation. This aspect of apps running just as well on the x86 smartphone will determine how well Intel hardware will be adopted, if at all.
Manufacturing Intel will develop a supercharged smartphone as their computer microprocessors are next to none and they have the most sophisticated manufacturing procedures and generate the tiniest transistors. Their most sophisticated processor is the 22nm, which will shrink to 14nm kicks in sometime this year, 2014, and Intel will be able to fit in more processor and graphics cores, extend battery life and increase clock rates beyond the capabilities of even current ARM chips should that occurrence come to pass. Graphics Graphic effectiveness is getting critical for example Apple released the iPad3 in 2012 and in this upgrade the processors were not revisited alternatively they simply doubled the performance of the graphics and what Apple did with the iPad3 is also synonymous with smartphones. Unfortunately Intel is really good at coming up with excellent processors but not graphic layouts. So they bought in the graphics layout for the Medfield Atom model and Intel will not provide a superior graphics layout than ARM in the foreseeable future.
Intel is not completely pleased about the prominence of ARM in embedded devices, possibly Intel is continuing to make efforts to produce an aggressive x86 processor in direct competition to what ARM has. Intel is extremely eager to break the smart phone and tablet market such that its efforts have since yielded the current Atom model present in laptops. Another result is a fresh Atom system named Oak Trail that will operate ARM systems like Android and the traditional x86 frameworks like Windows. Some of the problems of Intel are most MID producers like Apple are now familiar with the ARM architecture and they are not showing any signs of wanting migrate to Intel chips. Also Intel is failing to be taken seriously by Android manufacturers like Samsung because it does not keep timelines and even Google who promised to optimize new Android instances for x86 chips in 2011 are still to do so.
Intel seems to have more or less caught up to ARM, but they need to work on power consumption because their chips simply sacrifice too much processing muscle to not suck your battery dry in 20 minutes though the Motorola RAZR i did offer great battery life and decent performance. Another brake was the apps issue, even Google Chrome for Android needed an update to work on the Motorola RAZR i, and very few developers will invest time, cash and resources into app optimization for the x86 platform. Also integrated LTE, which only Qualcomm got right, should be worked on by Intel as every world market except for the US would welcome LTE-less phones. Again because most phones and tablets are developed around the ARM "ecosystem" Intel has to prove their solution is much better in many aspects; it needs to give organizations a good reason to split their R&D and production. With marketing Intel has to make a big deal out of their chip like Samsung did for theirs and people become aware and push their chip with big names like Apple so the rest of the market will catch on. Ultimately Intel needs to take advice from ARM who mean competition, look where they got us in less than a decade of mobile revolution and anybody can buy a license from ARM and make their own SoC.
Nevertheless Intel has a long way to go to persuade clients, carriers, and OEMs that the x86 chips are better because in the MID industry ARM traditionally has low prices and high energy conservation. No one wants Intel and their high prices and dictatorship the desktop was enough. Qualcomm/Nvidia/Samsung/Apple are signed to ARM and doing well and if Intel ends up on top: Nvidia will go bankrupt (Intel has stronger iGPUs); Qualcomm would be in big trouble; Samsung would lose their independence; and Apple would lose control of their components. Smartphone sales are rapidly outpacing any other kind of chips. So Intel may as well be where the money is as they are in it for the money. The only hope is for ARM to continue to make impressive advancements which serve the consumer. Intel’s going to be a wild one for the next few years, which could lead to more systematic improvements as the ARM camp has seemingly slowed down to a more stable rate of improvement, Intel will end up on top here though cost will keep ARM at the fore for a while. Intel is the biggest computer microprocessor producer in the world. Their brand is tried and tested and if anyone is going to give ARM a run for its money its Intel. One will not be far off the mark by assuming that in a decade or so the most superior battery life and the fastest smartphones using slimmest kind factors could be Intel inside.
[1] Katie Roberts-Hoffman and Pawankumar Hegde, "ARM Cortex-A8 vs. Intel Atom: Architectural and Benchmark Comparisons," University of Texas at Dallas, 2009. [2] Andy Boxall, "Meet all the fancy processors that power your phone," October 2013. [Online]. http://www.digitaltrends.com/mobile/mobile-processorroundup/#ixzz2tnAZKkLJ [3] Tom Krazit , "ARMed for the living room," April 2006. [Online]. http://news.cnet.com/ARMed-for-the-living-room/2100-1006_3-6056729.html [4] Anand Lal Shimpi, "The ARM vs x86 Wars Have Begun: In-Depth Power Analysis of Atom, Krait & Cortex-A15," January 2013. [Online]. http://www.anandtech.com/show/6536/armvs-x86-the-real-showdown [5] John Gillooly, "Understanding Smartphone Processors," September 2010. [Online]. http://www.pcauthority.com.au/News/231536,understanding-smartphone-processors.aspx [9] M S Smith, "ARM vs x86 Processors: What's the Difference?," May 2011. [Online]. http://www.brighthub.com/computing/hardware/articles/107133.aspx [10] Arthur Musah and Andy Dykstra, "Power Management Techniques for OMAP 3530 Applications Processors," 2008. [Online]. http://www.ti.com/lit/wp/sprt495/sprt495.pdf
