Surprise! Intel smartphone trounces ARM in power trials The industry analysts at ABI Research pitted a Lenovo smartphone based on Intel's Atom-based Clover Trail+ platform against a quartet of ARM-based systems, and Chipzilla's system not only kept pace with the best of them, but did so using less power. "The benchmarks were impressive but the real surprise was the current …
is useless without knowing the voltage; that'll give you the wattage aka the power dissipation.
Although if your battery is specified in amp-hours and you care about battery life, then the current is a useful measure - but it's not measuring processor efficiency. Example: a five volt supply might require an amp; I can make that half an amp by using a ten volt supply - but I haven't helped the efficiency (except perhaps marginally in the voltage converters).
Intel chips have always beat ARM in power efficiency test ( that is comparing work per watt). ARM's low power really only shines when the chip is idle. The Intel chips and systems always have had high idle power. With the move toward tablets and smart phones the devices are not as idle as they use to be. Thus ARM is moving into Intel world. Good luck.
It says a lot that Johnparchem has been heavily downvoted, when what he says is supported by Tomshardware:
"In general, our analysis suggests that the ARM-based CPU core is excellent at doing nothing, but starts to require considerably more power during computationally-intensive workloads... In this scenario, the CPU cores aren't cranking away, but the graphics core is still refreshing the screen and reading from memory. This constant reading taxes the memory controller, and is one reason why the Atom maintains low power consumption. Under heavier loads, we saw the Tegra 3 take a double hit as CPU power use ramped up quickly, along with the memory controller's draw.
Even though manufacturing technology is one of Intel's obvious strengths, the efficiency of its memory controller also becomes quite apparent in the company's power measurements. Intel and AMD have both pointed out the challenges facing ARM as it moves to 64-bit out-of-order execution, since both companies took years to refine and perfect their own implementations. Memory control is just another one of those areas Intel and AMD dedicate a lot of R&D to optimizing."
-http://www.tomshardware.com/reviews/atom-z2760-power-consumption-arm,3387-5.html
Well, first off well done Intel for getting such low power... even if it isn't an apples to apples comparison using undocumented benchmarks and different software. Pity thouth that it isn't enough to move the needle.
The biggest challenge for Intel in this space is that do not have any hardware partners. Intel makes the whole SoC end-to-end. As such they can only afford to make a limited set of SoCs. ARM is way different. There are many different companies producing ARM SoCs, each slightly different - covering a wide range of different markets. They compete furiously with eachother, upping the game and reducing prices.
Next, Intel have a terrible track record in the embedded industry. Embedded design people HATE Intel with a passion. Intel has pumped up the market with new offerings, then choses to dump that business unit and leave designers high and dry. What happened to 8051? 80251? i960? NOR and NAND flash? Various USB chipsets? DtrongARM? ... Once bitten, twice shy and all that... With Intel it is about ten times bitten - 11th time shy.
Thirdly, many of these gains are likely from playing process hopscotch. The ARMs coming down the pipe will soon be as good.
Sorry, Intel, but one good result with power numbers is not enough to grab the tiniest percentage of marketshare.
The power of x86 lies within the IBM PC, a fairly open and standardized platform with common hardware(-abstraction) and good ways to boot any operating system you want. (unless you have EFI "Secure" boot)
That's the power of it. Suddenly you can create, for example, a secure cryptophone, just by taking a minimal Debian, and adding OpenVPN and VoIP to it. And it would run on many phones without modification.... in a way just like people are doing now with PCs. You can easily turn your PC into a video disk recorder, just install the proper Linux distro. And no, it doesn't need to be ported like Cyanogenmod, it'll just run on your hardware even though the developer may never have seen it.
"The power of x86 lies within the IBM PC, a fairly open and standardized platform with common hardware(-abstraction)"
OK. But x86 phones are IBM-compatible PCs now? PC-compatible display+keyboard, PC-compatible storage, PC-compatible (w)LAN and USB, PC-compatible BIOS, etc...? References most welcome. Demonstrations of Windows installing on such a phone also welcome.
If that truly is the case, I'd be surprised. Sometimes I'm surprised.
Even if true, it still leaves the small matter of what else needs to be on a *real* SoC, and which design+build partners put it there, but let's ignore that for now.
And if that isn't the case, it's back to doing a hardware layer. Maybe not a complete port to a different instruction set, but then nor are two different ARM SoCs a complete port.
" You can easily turn your PC into a video disk recorder, just install the proper Linux distro."
And you won't be able to do that on ARM because?
Meanwhile, a year or two ago Intel declined to provide Linux support on some members of its SoC family. What's the current state of play of that game?
People never want features, they want benefits.
For example, nobody wants 64GBytes of flash on their phone. Nor do they actually want to even store stuff on their phone. What they really want is to access all their stuff while they are on the move. If you could find a different way to provide their stuff while on the move - without other penalties, they will accept that solution too.
The huge justification for x86 has always been: People are familiar with Windows and want to have the software that runs on Windows (eg. Office). They therefore buy windows, which needs an x86 processor. That monopoly has created a huge market for commoditised PCs, all which run x86.
The MS monopoly pretty much gave Intel a monopoly on a plate.
We have seen far better software and PC-like hardware (eg. Acorn RISC OS) being stifled by the MS/Intel monopoly.
The power of x86 (...)
(...) just by taking a minimal Debian, (...)
Debian runs on 18 different architectures, if I counted right (not all of them official ports with the full infrastructure). Just saying...
source: http://www.debian.org/ports/
I don't think so.
Exact screen sizes make a difference for a start, backlight brightness as well (and wheather it auto adjusts).
And it's a good point about amps. AFAIK all mobile batteries are proprietary to their phones. So the mfg can set the voltage to whatever they like.
So 1/2 the amps at 2x the voltage = same power level.
Not really a level playing field, is it?
A company with Intel's resources ought to be able to match ARM.s performance eventually. The point is that ARM now has the same code-compatibility grip on the smartphone sector that the IBM PC gave Intel in the PC sector. Overcoming code inertia may be beyond even Intel, unless they pay vendor's rewrite bills for them (they probably have enough cash).
Code compatibility? The dominant smartphone platform with 75% share runs the majority of apps on a virtual machine, and already supports Intel. WP uses a virtual machine also. And all the feature phones run various Java based stuff.
So there'll be a 10% niche of phones stuck on ARM only, but I think 90% is more than a big enough market for Intel!
Benchmarks are notoriously 'fixable' and a benchmark that is not even described is a joke.
Intel may or may not have overtaken arm but this benchmark suggests that Intel is inferior to Arm if they were better why do this dodgy benchmark why not do a real one and publish the details.
We already know atom beats arm when it comes to work/watt. The issue is watts/idle.
Actually, it isn't even that - price makes a big difference, as does controlling your own corporate destiny. If everyone does ARM, everyone has a chance to tweak things for competitive advantage or a different market slant. If everyone takes a single model CPU, things get very boring very quickly.
ARM just licenses its stuff to fabbers. This should be good information:
Don't forget to add time to that. Even knowing the power draw is of limited value if you don't know how long the processor is doing anything. Spinning up and spinning down are important, too.
I'm sceptical about these results. AFAIK x86 beats the pants off ARM for rendering web pages but itself is soundly whipped when the GPU gets involved, as on the I-Phone. This is why the SoC with the right silicon for the right task is so important and why big.Little will only start to make sense when the compiler and scheduler have had a few generations to get it right. Intel does not do the heterogeneous computing environment of modern mobile devices anything like as well as ARM or even AMD.
You can avoid the time taken getting to the right compiler optimisations for big.LITTLE if you choose the right scheduling scheme. It appears the current stuff in Linux is not as good as could be (In kernel scheduling - cant remember exact name), and is improved by simply upgrading the scheduler so it can move stuff from core to core quickly.
Although that upgrade to the scheduler is quite difficult as as standard, the SMP scheduling currently assume all cores are equal, which they of course are not in big.LITTLE
So how do you sell an instruction set compatible processor (because that's the core feature people buy Intel for) at 1/10 (or less) than what you pay for a desktop/laptop/server processor without people feeling they are being ripped off?
As a designer you with an Intel processor you cannot.
a) Change foundries. Don't like the shipping delays for your order? SFW.
b) Insert other chips into their processor carrier. Get your own.
c) Add ASIC functions to their silicon. You are joking.
So you pay for more board space, more devices, more assembly and test costs.
Intel may now be as power efficient as the current generation of ARM SOC but they are still a premium brand so even if more manufacturers start to produce Atom based phones they will still be competing against cheaper ARM SOCs which means Intel will never get their costs down to get into the £100 budget phones which now are coming with dual core ARM cpus which for a lot of people are powerful enough.
Can you see the LIE, they have not listed the Voltage
Intel Z2580 3D Score 6664 Peak I = 0.61
Qualcomm APQ8064T 3D Score 6628 Peak I = 1.404
Power = V in Volts * I in Current
if you do this they are using the same amount of POWER
E.g.
Intel Z2580 6.905 Volts * 0.610 Amps = 4.212 Watts of POWER
Qualcomm APQ8064T 3.000 Volts * 1.404 Amps = 4.212 Watts of POWER
While that's possible, the specs would suggest it is unlikely given an operating range of 0.3V-1.2V.
http://ark.intel.com/products/70100/Intel-Atom-Processor-Z2580-1MB-Cache-2_00-GHz
Of course they may have massively bumped the voltage....
And I believe that the tests probably weren't rigged to favour Intel - performance and power usage aren't the things Intel struggles to compete on when facing ARM. ARM SoC's can provide most of the performance and most of the power efficiency in a package that can be tailored to the application at a fraction of the cost.
If Intel want to sell their Atom to OEMs, they have to create promotional literature for their salesmen to use against the overwhelming competition that ARM represents in their chosen market. Expect more of this as Intel attempts to gatecrash ARM's strongholds whilst mounting a rearguard action to protect its server base against the ARM v8 chips. This is PR.
Consumers want to compare devices for size, weight, price, battery life & functionality. We mostly don't care how the manufacturers achieve it but there is some brand loyalty.
Device makers want to compare SoCs so they know what other chips or modules they need to buy and find space for within their device. If all other things were equal, a physically smaller SoC would reduce the overall device cost. Device makers are probably a lot more focused on the overall costs than consumers but still aware of the consumers' brand preferences.
SoC makers want to compare CPU cores so they know what other functionality they could or should cram into the same SoC to make it more attractive to device makers. There are many choices that influence the balance here - a better core could need less memory, or the same amount of cheaper memory. Less memory could allow on-die space for hardware to assist the radios. The flexibility of choices created by the competitive ARM eco-system is the killer feature here.
I'm only a consumer so I assume that the SoC makers & device makers have better access to more relevant (& accurate) figures to make their decisions. But as with most techies, I would be very curious to see a real like-for-like comparison at each level for shipping products: Intel & ARM cores, Intel and ARM-based SoCs, Intel-based and ARM-based devices with the same price/feature/performance envelope.
Unfortunately this article therefore seems of no value to device makers, SoC makers or techie consumers like me.
Tite says it -- I'm surprised. I knew the newer Intel chips were relatively low power, but did not realize they beat out ARM.
I hope Microsoft can be kept at bay enough to have some netbook-like computers on the market again to run Ubuntu or whatever on (as opposed to the over-priced, over-specced monstrosities like ultrabook that Microsoft wants in order to run Windows dcently.) I was waiting for an ARM based unit, but I realy don't care what chip it has in it as long as he result is low power use and decent performance.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
