### Back to Article

• #### StormyParis - Monday, July 22, 2013 - link

What are the result once you normalize for battery life ? Right now the article reads like "hey, bigger batteries last longer !" Reply
• #### StormyParis - Monday, July 22, 2013 - link

battery size, not battery life Reply
• #### ljlego - Monday, July 22, 2013 - link

Unless I'm grossly misunderstanding your question/statement, I feel like this passage directly from the article should assuage your concerns: "You’ll notice two bars for the 2013 MacBook Air, one indicating its result and one with that result scaled down to simulate what would happen if it had 78.7% of its actual battery capacity - putting it on equal footing to the 42.5Wh iPad 4. With workload and performance constant, it’s safe to assume that battery life scales linearly at best with battery capacity. In other words, our MacBook Air numbers at 42.5Wh should be indicative of what we’d expect if the 13-inch MBA actually had a 42.5Wh battery rather than 54Wh unit." Reply

• #### Spoelie - Monday, July 22, 2013 - link

the 42.5 wh result *is* normalized for battery size. Reply
• #### cdegallo - Monday, July 22, 2013 - link

I think there's misunderstanding in this article. The MBA is only scaled in battery life to match the 42.5Wh iPad4. NOT scaled relative to the battery capacities of the other devices. This becomes a poor display of power efficiency as much as it became a battery capacity study. The experimenters should have scaled all battery life numbers by the capacity of each device's battery. But that would unfortunately show the Haswell platform in dead last all the time... Reply
• #### ClockworkPirate - Tuesday, July 23, 2013 - link

I'll do some math here.

The MacBook Air's battery can store 54 (fifty-four) Wh (watt-hours) of energy, or 194400 joules. It lasted 14.28 hours (about 51408 seconds) on the light browsing test. Therefore, it consumed 194400J/51408s = 3.782 W(watts = joules per second) on average during that test.

The fourth-generation iPad's battery can store 42.5 (fourty-two point five) Wh (watt-hours) of energy, or 153000 joules. It lasted 9.48 hours (about 34128 seconds) on the light browsing test. Therefore, it consumed 153000J/34128s = 4.483 W (watts = joules per second) on average during that test.

The iPad drew more energy from its battery every second. Obviously some of that is thanks to it's higher-density display, but it's nevertheless impressive that full x86 has come down so low.
• #### psychobriggsy - Tuesday, July 23, 2013 - link

It does appear that the high resolution display on the iPad 4 is what it drawing the extra power consumption - it needs a comparatively higher strength backlight to push enough photons through the smaller pixels.

In that sense, the test is quite unfair, when you consider the iPad 4's average power consumption for web browsing is 4.5W. It's why the iPad 4 has such a large battery in the first place. It would have been good to see normalised power consumption for other tablet battery capacities.

Or we can wait for the iPad 5, which may be using IGZO which should use a lot less power. Apple will probably choose to trim battery size rather than lengthen run time. It may also use a 28nm SoC rather than a 32nm SoC, and incorporate more power saving features in their second generation Swift ARM core.
• #### cdegallo - Tuesday, July 23, 2013 - link

No problem with the math. But let's go to the video playback test: The 54Wh MBA gets the same playback time as the 16Wh Nexus 7. This makes Nexus 7 out to be ~3 times MORE efficient than the Haswell platform in that case, and the iPad4 is still ~30% more efficient than the MBA in that same benchmark. Scaled by battery capacity in the video playback, the Haswell MBA would come in last place near the Surface 8 Pro (which isn't very surprising).

I think there are too many compounding factors and assumptions between the devices tested to make statements other than that you get even better battery life on full-OS computers with Haswell, and you don't need to use a cut-down simplified OS (like iOS or Android) to get tablet-like battery life on a laptop.
• #### munim - Tuesday, July 23, 2013 - link

But then you're comparing a 7 inch, 1280 x 800 screen to a 13.3 inch 1440 x 900 to a 10 inch 2048 x whatever. Reply
• #### cdegallo - Tuesday, July 23, 2013 - link

Just another example of how this was not a good comparison between CPUs as much as it was a comparison between the Haswell MBA battery life, and other device battery life. It becomes less and less conclusive the more and more you look into the situation. Reply
• #### purerice - Sunday, July 28, 2013 - link

please provide a link to your own article that is way better than this so we can shower you with praise at how superior you are. Reply
• #### ironargonaut - Wednesday, July 24, 2013 - link

The HSW number was scaled to match the iPad4 battery capacity as you yourself state, and when scaled it still beat the ipad4 in the web browsing category by 18%. Therefore, your statement "But that would unfortunately show the Haswell platform in dead last all the time..." is categorically false. Kinda of hard to take your comments seriously when make blatantly false statements. Reply
• #### cdegallo - Thursday, July 25, 2013 - link

My so very profuse apologies; I was looking at the video playback conditions. Sorry if that makes it hard to take my comments seriously... /scoff

My main point is that the Haswell MBA getting better battery LIFE does not mean that the processor is always more power efficient when compared to the list of other devices, or even necessarily comes close. This was not a scientific examination of the processor's power efficiency, yet its trying to make conclusions as such. There are far too many assumptions and unaddressed variations between the hardware platforms compared.
• #### purerice - Sunday, July 28, 2013 - link

iow, you are saying there are still too many variables, yes? Yet the process described in the article removed variables as many as possible and describes the variables still present.

Of course you cannot stick Haswell in a smartphone. The article does not state that you can. It states that the gap is shrinking and lists what Intel can do to improve.
• #### FwFred - Monday, July 22, 2013 - link

Great article. I'd be interested to see a better breakdown of display power between the iPad and MBA. You mention the resolution difference driving increased power, but I'm sure the growth from 9.7"->13.3" would also have a similar effect on display power. Reply
• #### lmcd - Monday, July 22, 2013 - link

He noted the entire panel + backlight power usage as a delta of 4.2 W more from the iPad if I understand correctly, due to the higher requirements of the panel itself and the greater backlight required (because pixel density leaves less room for light to pass through).

What I would be more concerned about is the extra power needed to drive the iPad panel as opposed to the MBA panel.
• #### FwFred - Monday, July 22, 2013 - link

That delta was between the iPad 2 vs 3 (same screen size), so we attribute it to mostly the resolution change. I haven't seen good numbers to compare between different panel sizes Reply
• #### Sm0kes - Monday, July 22, 2013 - link

Yikes man, did you even bother reading the article? Your question was essentially the point of Anand's testing. Reply
• #### cdegallo - Monday, July 22, 2013 - link

Agreed. The battery life values should have been scaled by battery capacity for each device. This became a misleading study for people who only looked at the charts. Haswell is the LEAST power-efficient platform by far in all of these tests. Reply
• #### dusk007 - Monday, July 22, 2013 - link

The questions is: "Doesn't this all come too late?"
Just looking at the performance Baytrail promises, it seems that most people that only require some tablet like or MacBook Air like, light workload will do just fine with those newer Atoms. If it is just a bit of taking notes, web surfing and watching the occasional video, 14nm Atom and 20/16nm ARM should do fine.
Who needs those \$200+ Haswell/Broadwell anymore? More efficient less bloated software (such as Windows 8) and some fast ssd (Smartphone flash is often very slow), cheap tablet SoCs will serve that market just fine.
Those that want to game and do more demanding desktop stuff will probably prefer to go straight for serious speed ala 4750HQ and up. Why bother with a 7W Haswell? Or even the 15W think about the cost.
I feel like this low power Core CPUs won't be in very high demand unless they can make a case for gaming and appeal to all those that want some Windows PC gaming capability. For the mass market gaming is the only saving grace I see.
• #### jeffkibuule - Monday, July 22, 2013 - link

We'll have to see what Bay Trail actually does first. At some point we reach the point in all platforms when you achieve "good enough" computing for most people and sales of tablets, PCs and even smart phones drop off because what you have lasts a really long time. I shudder to see what artificial tricks will be made to make you upgrade then. Reply
• #### smartypnt4 - Monday, July 22, 2013 - link

This is Intel working with what they had in the pipeline when they realized they needed something low power. Bay Trail will come out this year, and that will serve the low-power market much better than Haswell.

However, with Haswell, you get more dynamic range for performance. Yes, you can make Haswell have super great battery life... if you're not doing anything strenuous. This is just a good argument for the race-to-sleep argument Intel has been making for a while now. In heavy workloads, yes Haswell is going to eat WAY more battery. However, this allows for one device to fill two roles. No one wants to work with an Atom as their primary productivity device; however, people would love Atom-like battery life. This is a way for people to have battery life if they want, and performance when they need it.

The other thing: Atom is good enough for most of the market, true. But Haswell should still outperform Atom and the ARM-based chips significantly, even at low power. Theoretically. I'm not sure how this bears out in actual testing, but still.

For my personal use, Bay Trail will be more than adequate. For a different portion of the market (think Surface Pro-level pricing), Haswell will be the better choice. It'll be interesting to see if the market share of Atom vs. Haswell changes as computing moves more to tablets.
• #### Sm0kes - Monday, July 22, 2013 - link

You hit the nail on the head - dynamic range. Performance when needed, great battery life when not.

I think the market is getting to the point where cost will artificially limit Haswell's (and presumably Broadwell's) adoption across "converged" devices, whereas this was previously an architectural / power / heat limitation. Interesting times indeed :)
• #### smartypnt4 - Monday, July 22, 2013 - link

Exactly. The Haswell chips will be so expensive compared to ARM/Atom implementations that most people will end up with Atom/ARM chips in their primary compute devices, which will probably be tablets.

The Core family will become for anyone who wants to game or to create content, both of which needs more than what Atom can provide.
• #### RYF - Monday, July 22, 2013 - link

I agree with you on this. There are some hybrids out there with ARM + x86 chips and dual booting Android and Windows. If Haswell offers the dynamic range as what you have stated, there is no need for the existence of such hybrid devise anymore. Reply
• #### beginner99 - Monday, July 22, 2013 - link

The thing is that haswell offers way, way better performance in case you need it without sacrificing battery life at idle-light workload. All ARM and AMD fanboys that still doubt that silvermont will demolish ARM should just look at these results: Haswell is more power efficient than Apples custom ARM SOC! I guess the conclusion can only be that silvermont is even a lot more efficient and intels claims of 4x less power at same performance seem resonable. Reply
• #### lmcd - Monday, July 22, 2013 - link

That custom ARM SoC also loses to A15-based and Krait-based devices. And don't forget that Haswell is more efficient on the platform that allows more low-level code. Video decode, which is identically optimized on both platforms, is almost as important.

A better test would be if the ARM SoCs ran Linux with good drivers and firmware, and as such maximize their potential.
• #### beginner99 - Tuesday, July 23, 2013 - link

You can be certain that Apple optimized this already. iPad is almost certainly the best optimized tablet available and hence other tablets will just look worse in such tests. Reply
• #### Krysto - Monday, July 22, 2013 - link

I wish good luck to Silvermont beating 3 Ghz 20nm ARM chips in 2014.

Haswell costs \$350 for the damn low-end dual core version. How Anand seems to completely ignore this HUGE issue when he's talking about tablets, I have no idea. News flash - people are not going to buy \$800+ "tablets", just because they have Intel chips in them. I thought we settled that already? And using 10" screens as work computers sucks.
• #### FwFred - Monday, July 22, 2013 - link

I don't think we have any idea what OEMs pay for CPUs. It's pretty obvious it's not the price on ARK when you can sometimes buy full PCs at retail within \$70 of just the CPU alone. The Celeron 847 is listed at \$134, but Google sells the Acer C7 Chromebook for \$199.

Let's see how the prices play out over the next year. Haswell is obviously not competing with <\$300 tablets (Silvermont for this), but I don't think they need to cost \$800 either.
• #### psychobriggsy - Tuesday, July 23, 2013 - link

You forget that this is a DEVICE power consumption test, and that the devices are very different - most notably the display. The iPad 4's display requires a lot more power to push light through its tiny pixels.

It's impressive that Haswell, under very light loads, is now this good. But it does cost around ten times more than the ARM SoC, and it's coming in devices that cost twice as much.

Also it is comparing 2012 tech (iPad 4) with 2013 tech (Haswell). Let's see what the iPad 5 brings.

As for Silvermont, performance leaks show that it's substantially behind AMD's Jaguar on a per-clock basis still.
• #### eanazag - Monday, July 22, 2013 - link

I have a surface pro and I have been waiting for Haswell. The Ivy Bridge is good. Battery life is okay. I want to see Surface Pro with Haswell; no wait for Broadwell. MS can't wait for Broadwell. I am okay with the current thickness.

I have an iPad 3 also and have cut back using it except in some scenarios. Video is one of them; specifically Netflix. Unfortunately, I can only get Netflix to run in the browser, because the app gives me an error code and Netflix support routed me back to MS since I already had up to date drivers.

I think there is a market for good enough gaming that can dock and do real work. The iPad doesn't do real work - I miss the mouse. The Surface Pro hangs in there a bit with Blizzard titles because they're CPU heavy (must be plugged in because it really taxes the system). It is in no way competing with my desktop,but it spanks anything in the first Core i5 series and before even with entry level dGPU.

I am currently carrying a quad core Ivy Bridge laptop, surface pro, and iPad to work every day. I find that 90% of what I do is accomplished easily on the Surface Pro. The laptop does knock out video encoding more quickly. Virtualization on the Surface Pro is RAM limited. Everything else I comfortable with. I'd like to see Thunderbolt on the next Surace Pro. I still dream of possible external GPUs at night when I get to bed early.
• #### ClockworkPirate - Monday, July 22, 2013 - link

Gaming is exactly why I have a Surface Pro and not a Surface RT (well, that and the active digitizer). The HD 4000 can run most everything (short of Crysis and Battlefield, obviously) when it's plugged in and I've yet to have issue with the battery life. Haswell improves both cases: the HD 5000 is more powerful maxed out, but the whole package uses less power when you need it to. In another year or two (Broadwell?) I expect to ditch my gaming desktop in favor of some future Surface Pro. Reply
• #### James5mith - Monday, July 22, 2013 - link

Asserting that tablet makers will take Microsoft's example with an actual SSD instead of something SDIO connected is assuming an awful lot. The Surface is the only modern tablet that I've seen with actual mid-high range performance of the storage subsystem. Everything else is abysmal, and it almost exclusively has to do with using SDIO for power concerns vs. a full SATA implementation. (From what I've seen.) Reply
• #### Jaybus - Tuesday, July 23, 2013 - link

I think it pointless to compare a MBA, ultrabook, or Surface Pro to a tablet. They are for different uses. It's like comparing a Smart car to a F-150 pickup and then claiming the Smart car is more fuel efficient without qualifying what it is more efficient doing. The Smart car can move two people around 3 times as efficiently as the F-150. On the other hand, it isn't even a third as efficient as the F-150 if you have to move 30 cases of bottled water. Reply
• #### twotwotwo - Monday, July 22, 2013 - link

I'd love something that could (mostly) stand in for my current computer (good with a keyboard, fast enough, open app ecosystem), but pick up some of the benefits of all of these low-power gadgets--serve as an acceptable tablet, get up to modern pixel densities, have a price roughly in line with its performance (not rMBP prices for MBA speed), etc. Currently that's a lot like saying "I don't want the Wintel ecosystem to collapse completely"; realistically a future Surface or a Silvermont-based Atom tablet thingy is the most appealing thing likely to come out super soon. But if a touchscreen Mac comes along or even if Google gets Chrome OS over the line, they could possibly win me over. Reply

Can't agree more. I can see Apple bridging the iPad and Mac with future Macbook Air lineup. Reply
• #### Khato - Monday, July 22, 2013 - link

Thanks for the further investigation Anand! Really quite surprising to see Haswell on top in the realm of 'active idle' battery life. Even adjusting for the difference in screen resolution and size (have to remember that nit is just another name for cd/m^2 and hence total light output at 200 nits on a 13.3" display is 1.88x that of a 9.7" display) it's likely on par. Note that I found the discussion of display backlight power usage somewhat peculiar as the average power consumption for the devices was only 3.78W for the MBA and 4.48W for the iPad 4 - far less than those backlight figures.

Also quite interesting to see how poorly Haswell does during video playback comparatively. I'd imagine this might be due to two sources actually. First the obvious of Haswell's hardware accelerated video decode likely having to fire up a decent amount of the GPU and hence not being the most efficient. Second is a more interesting possibility - I'd assume that Apple is using a panel self refresh enabled display which is most effective in 'active idle' scenarios where screen content is rarely changing and then pretty much useless on video content. I know it's not likely going to happen, but I'd be quite curious to see whether video playback battery life changes much if it's not actually displaying the video on screen. My guess is that it wouldn't make that big of a difference (other advantage of PSR being that graphics engine can be turned off, but that can't happen if it's still decoding video) but maybe it would? Regardless, could be interesting.
• #### Khato - Monday, July 22, 2013 - link

And since there's not an edit button...

A more amusing and simple way to potentially test the effect of PSR (again, I'm assuming that such is the kind of option Apple would have enabled if possible.) Simply run the tablet web browsing battery life test with a screensaver that has activity over the entire screen running. That'd give a much better idea of whether it's the hardware accelerated decode or the graphics activity on screen that's at fault for the reduced battery life.
• #### Kidster3001 - Tuesday, July 30, 2013 - link

After the screen the biggest difference in power consumption will be caused by the software. The biggest battery eater is a lousy app. A good video playback app will cause the CPU to be awake about 10%-15% of the time, sleeping 85%-90% and all the awake time will be at lowest available frequency. A poorly optimized video playback app will cause the CPU to be awake as much as 60%-70% of the time with the CPU occasionally spiking to higher frequencies. Differences in battery life can be 2x-3x worse.

In my experience, comparing different hardware is hard enough. Comparing them across different software (OS and App) at the same time makes it nearly impossible to reach any meaningful conclusions about the hardware.
• #### smartypnt4 - Monday, July 22, 2013 - link

I feel like if the MBA included PSR, Apple would've been hawking that at the top of their lungs. I suppose it's possible, but I feel like it would've been something that they bragged about. Reply
• #### eio - Tuesday, July 23, 2013 - link

also wondering about the power consumption number, and still don't know how does the 4.2w delta translate into 10% difference.... Reply
• #### DanNeely - Monday, July 22, 2013 - link

As someone whose carried a netbook around convention centers for years because near idle battery life, a physical keyboard, and a rigid hinge (so I can use it without a desk) were more useful to me there than more performance or a touchscreen this is a really encouraging result. Haswell looks like it's capable of giving me everything I want.

Now if I could just convince at least one OEM to build something with the same height as my old netbook. I could, barely, stuff it into a pocket of holding in some of my pants; while awkward I liked being able to free both of my hands when I need to without carrying a backpack. This years crop of 11.6" mini-laptops and atom based convertibles are all at least a little bit too big to fit.
• #### ananduser - Monday, July 22, 2013 - link

Silvermont will kill ARM. Those crazy 16 hour ATOM tablets(with dual batteries) will smash through the 20h mark easily. The Thinkpad tablet is already an epic coffee shop machine; imagine it with Silvermont. Reply
• #### Krysto - Monday, July 22, 2013 - link

Nice dream, but it won't happen. And is it 16 hour battery or 20h? Make up your mind. Reply
• #### DanNeely - Monday, July 22, 2013 - link

I think what he's saying is that just swapping the CPU up for the new model and keeping everything else the same will push battery life from 16 to 20 hours. Reply
• #### ananduser - Tuesday, July 23, 2013 - link

That's what I meant. Thanks! Reply
• #### Krysto - Monday, July 22, 2013 - link

You seem to be implying that Apple needs a device that is between the iPad and the Macbook Air, that will use Haswell.

Ok - how the hell would that help Apple's profits? A \$700 Macbook with a \$300 Haswell chips will NOT help Apple. If anything, Apple will make a quad core 2+ Ghz Swift (or wait until the next-gen and even faster 64bit architecture in the ARMv8-based A8X) and release a \$700 Macbook, that uses a \$30 ARM chip inside.

I'd say that makes a lot more sense for Apple's profit - no?
• #### Krysto - Monday, July 22, 2013 - link

Heck, let's do some quick math, too. Let's say Apple make 30 percent profit (\$300) on the \$1000 Macbook Air. That Macbook Air uses a \$350 Haswell ULV chip (according to Anand).

Let's take that out and put a \$30 ARM chip. Now Apple makes 300+350-30=620 dollars in profit for the \$1000 Macbook Air. Of course they wouldn't actually do that to a \$1000 Macbook. So they'll release this "new" Air that is \$300 cheaper, and sells for \$700. So now we take this 300 out of the \$620, and we're left with \$320 in profit for the \$700 Air.

That's actually a bit more raw profit than before, but at \$700, they'll sell many more such devices which means a lot more total profit in the end. \$320 pure profit for a \$700 product would be amazing for Apple.

And the best part is they wouldn't need to cut the quality of any of the other components. If they did put Haswell into a \$700 Air, not only would it completely and utterly OBLITERATE Apple's profit on it, but they'd also need to use much cheaper components than in the \$1000 Macbook Air, to make up for Haswell and the \$300 price decrease.

Using a 64 bit ARM chip in future lower-end Macbooks without compromising on quality and battery life at all makes so much more sense than using Haswell and on so many levels, that's it's sad to even imagine someone would recommend them go with Haswell.
• #### jemima puddle-duck - Monday, July 22, 2013 - link

This. Thank you for bringing some reality back to this discussion. Can we please have at least a little acknowledgement of costs amongst the technology. It's all very well that the new Ferrari has the efficiency of a Prius, but if it costs 100x more, it's academic to normal people. Reply
• #### Ktracho - Monday, July 22, 2013 - link

I think this is exactly why Microsoft says they aren't giving up on Windows RT. If they can get the number of apps up dramatically (say, by a factor of 1000), and simultaneously find a way to decrease price by, say, half, they could have a winner.

For those times when performance is not enough, these tablets can be docked, and in the future (this is not yet a reality), people will be able to remote into a powerful PC complete with a real, physical graphics card, and use it as if it were all built into the tablet. I think this is eventually where everyone is headed. In other words, no one will ever need to carry around anything more powerful than a Windows RT system. (If you don't like Windows, Apple will do the same thing with their ecosystem, and Google likewise, etc.) Intel Core CPUs will be relegated almost exclusively to servers, where price does not have to be bargain basement level.
• #### DanNeely - Monday, July 22, 2013 - link

MS currently has 100k apps in the W8 store, vs 600k for Google, and 900k for Apple. More apps wouldn't hurt; but they're much closer to parity than your factor of 1000 implies. Reply
• #### Ktracho - Tuesday, July 23, 2013 - link

I don't really know how many Windows RT apps are in the store, but I doubt there are anywhere near 100K of them. Reply
• #### smartypnt4 - Monday, July 22, 2013 - link

I think this scares Intel more than they're willing to admit. Apple is a very large customer that pays for the best binned CPUs Intel makes (in terms of leakage, not pure GHz). That's 10% of Intel's consumer business right there, which is definitely non-trivial. Apple's already got the people with the expertise to make a quad-core Swift. I think the issue is that it's just flat out too slow to use on OSX and keep a good user experience. Also, Apple would have to do *everything* Intel does associated with the Haswell SoC: a full-sized memory controller (you can't just grab the one from the A6X), and all the I/O (SATA, PCIe, USB, etc.), and that's a non-trivial task. Reply

Apple would also cajole a lot of OS X developers into recompiling/porting their application, or else they'll end up with a fruity version of Windows RT. Reply
• #### smartypnt4 - Monday, July 22, 2013 - link

That's very true as well. While it's still a threat to Intel, it's not *that* big a threat, I suppose. That's a lot of legwork that Apple would have to do. I'm not even sure it'd be any cheaper for them to develop, validate, and get their own chip fabbed. Reply
• #### psychobriggsy - Tuesday, July 23, 2013 - link

I don't think Apple will switch to ARM in the near future for their MacBooks, iMacs, etc.

But *if* they were, they could invest in a larger SoC than they budget for the iPad, and it would be a 64-bit ARMv8 design, not the current ARMv7 Swift. They probably have this core designed, possibly ready to manufacture, maybe even being manufactured in SoCs right now (but more likely for next year's SoCs).

A quad-core, 20/28nm, 2GHz+ custom ARMv8 SoC should compete fairly well with a ~1.5GHz dual-core with HT Haswell, on the tasks a person wants to run on an 11.6" MacBook Air. The MBA has more thermal headroom than a tablet, so the CPU and GPU upper clock speeds can be higher as well. They might even do a big.LITTLE thing with their own core and an ARM Cortex A7/A53 for the mostly idling common case.

As for recompiling Mac OS X apps - that's a button press away for the developers, once Apple makes XCode support ARM compilation for desktop apps. Again, in-house I bet this is already happening.

This path could cost Apple \$50 per SoC to manufacture (more advanced process, larger SoC), but that's still vastly cheaper than the price of Haswell, even with the discounts that Apple no doubt gets.
• #### djscrew - Monday, July 22, 2013 - link

You complain about CPU price and then suggest buying an over-marked up Apple product instead? Are you demented? Just get the iPad you're in denial about wanting and move on. Or use the iPad you already probably have and be happy. Either way, if you have an iPad you should donate that extra money to an orphanage or something useful. Besides, everyone knows that no \$30 ARM chip can do what a \$350 Intel chip can do or Apple would use it. People still want the ability to work on their Airs. Granted in 10 years this all should be pointless as the \$30 chips should end up replacing the \$350 chips of today and mobile technology + the internet is ubiquitous... that is if we're not all too greedy to realize that goal. Reply
• #### beginner99 - Tuesday, July 23, 2013 - link

Apple does not pay list price for Intel CPUs. I have no idea about the actual price but a safe guess is probably half, eg. \$175. No the picture already looks very different. And don't forget that apple still is fancy in content creation stuff which all does require powerful hardware so ARM is already a no go. Reply
• #### ancientarcher - Tuesday, July 30, 2013 - link

Interesting thoughts!
But why do people think that Apple will keep the CPU limited to quad-core. They can have 16-core or 32-core CPU which massively increases the raw computing power. The reason we haven't seen so many core CPUs till now is because of Windows which is mostly single core/thread based. If you have full control of your OS, you can do funky things - like massively parallelizing your CPU workload.

And on the GPU front, Snapdragon 800 is already exceeding core-i3 graphics (not much grunt there, agreed). With PowerVR6 (Rogue) series, we will see GPU power increase multiple times over current levels, which should help in graphics related stuff people do on Apple computers.

All in all, I would think it is a reasonable bet that Apple will go its own way with processors
• #### Kidster3001 - Tuesday, July 30, 2013 - link

A 16 or 32 core A-15 based chip would likely consume more power than Haswell does today. Just look at what Tegra4 and Exynos 5 Octa do to batteries.... and they're only 4 core. Reply
• #### HisDivineOrder - Monday, July 22, 2013 - link

There's just no evidence of a huge market for a tablet that is a laptop, too. So far, all the evidence suggests on the contrary that people want their computer to be separate from their tablet because they want to take their tablet everywhere and no matter how thin or cheap Intel makes the Core line, it won't ever be cheap or as thin as ARM or even Bay Trail.

I think most people want tablets to be like a cheap novel they take. If they lose it or if it gets broken, eh, it happens. You start talking about Core series tablet/laptop hybrids, suddenly you're talking a LOT more money and a LOT more investment in it.

Eh. I don't think Apple is particularly interested in the hybrid strategy because their strategy involves getting you to buy into the entire ecosystem. You buy an iPhone, which compels you to want an iPad, which compels you buy a Macbook and an iMac and then they have you.

So you're suggesting they combine the iPad with the Macbook, remove all need for the iMac, and suddenly Apple is cutting their nose, their ears, and their mouth to spite their face.
• #### chizow - Monday, July 22, 2013 - link

Haswell ULT/ULX in a Surfarce Pro 2 for \$500 would sell tons. I'd be in line for one for sure. Make it happen MS! Reply
• #### cdegallo - Monday, July 22, 2013 - link

Unless I read this article wrong, the MBA battery normalizations were only relative to the different MBA's and the iPad, NOT relative to the other platforms tested? If so, these plots need critical thought. The values are battery life; except the goal of this article (as I read it) is to showcase how power efficient Haswell is, not to showcase how much battery life the MBA gets. So get rid of the battery bias all together.

The most logical thing to do would be to scale the battery life studies by the battery capacity for each device. The numbers are misleading; scaled by battery capacity, the Haswell MBA would come in last place in almost every benchmark. It pretty much matches the Surface 8 Pro in video playback when scaled by battery capacity. But of course, then the MBA/Haswell wouldn't come in first every time...

Take a look at the video playback numbers. The 54Wh MBA has video playback time almost the same as the 16Wh Nexus 7; but the N7 does that with less than one third the battery capacity of the MBA. That indicates that the Haswell platform is over THREE TIME LESS power efficient than the Nexus 7 in the video playback test.

Scaling the web browsing values, the Nexus 7 is still twice as efficient as Haswell.

The equal-capacity 42.5Wh iPad4 gets almost twice the video playback efficiency of the 42.5Wh-scaled MBA.

It's clear that there are important power-saving features built into Haswell, but I don't think this experiment proves that it is ready for "tablet duty" on a universal scale anytime soon. One step closer, perhaps.

The biggest takeaway is you still get a full OS plus better than average battery life and the extra CPU power when you need it. But it is not ready to displace ARM in mobile devices.
• #### beginner99 - Tuesday, July 23, 2013 - link

Video playback does suck on haswell but that can be easily explained: ARM SOCs have a dedicate decoder while haswell uses it's iGPU. In fact I wonder why intel doesn't use a fixed, efficient decoder as watching videos on such mobile devices is very common. Reply
• #### psychobriggsy - Tuesday, July 23, 2013 - link

Intel does have a fixed function video decoder, and an encoder. However I'm sure post-processing is still done on the iGPU, which may be affecting the equality of tests. Or Mac OS X doesn't support the video decode hardware...

We should remember that the Nexus 7 costs less than what an OEM will pay for an ultra-low-voltage Haswell on its own.
• #### Kidster3001 - Tuesday, July 30, 2013 - link

You can't honestly compare Nexus 7 to Haswell in this. The Display is the biggest battery eater in mobile devices. Put the Netbook display on the N7 and battery life will plummet. We already know Tegra3 is not very power efficient comapred to it's peers. Reply
• #### Kevin G - Tuesday, July 23, 2013 - link

The amusing thing is that both the MacBook Air and the iPad are leaving a few power optimizations on the table. Neither takes advantage of panel self refresh. That is a pretty big win for light workloads. Reply
• #### watersb - Tuesday, July 23, 2013 - link

"total platform power of the 2013 13-inch MacBook Air is lower than Apple’s 4th generation iPad" - favorite quote from any AnandTech podcast, ever. (which is saying a lot, because BK's rants are quite fun)

When we got our first iPad, we got a great keyboard case for it. My wife uses is for hours each day, much like a laptop. The kids pull it out of the case and use it for games... Last month, I got the same case for my iPad 4.

After 6 months of using it as a tablet, I didn't like the laptop-like form factor. I made myself stick with it for two weeks. But I found that I wrote *less* weigh the keyboard.

I was surprised. I don't have a MacBook Air for comparison; I still want my iMac-keyboard-trackpad for coding, hosting development Linux servers via VMWare...

But a convergence device seems like the ultimate solution, until I tried to use my existing tablet like that. And found that I really want it to be just a tablet.

(When my phone can wirelessly pair with desktop display-keyboard-trackpad and run VMWare, then I will know The Convergence has finally arrived.)
• #### watersb - Tuesday, July 23, 2013 - link

(typo: I wrote less _with_ the keyboard. I suppose I wrote more accurately, though...) Reply
• #### iwod - Tuesday, July 23, 2013 - link

I am pretty sure Apple hasn't had the best software implementation into using Haswell Video Decoder. Quicktime just hasn't been getting that much attention and i have no idea why. May be OSX Mavericks will improve on that but i have no idea. Otherwise I dont believe Haswell Video Decoder is that far behind ARM / PowerVR Camp.

Taking into consideration MacBook air has Fans, and a more power sucking SSD controller etc, LPDDR3 4GB, compared to LPDDR2 1GB, and as mentioned the LCD display difference. And OSX, at least the tested version isn't even as well power efficiency optimzied as iOS.This is actually some very astonishing results.
And if you think about what next year Broadwell may even come up with.....

But then is the price. I believe without the Memory, Apple are making their iPad SoC for no more then \$30 per piece. So even if you add in much exaggerated \$10 per Apple SoC for R&D. This is still less then \$40 per SoC.

Compared to Intel Haswell ULT, even with some hefty discount i am sure they are at least \$200 a piece. 5x difference!

I am not seeing how Intel can win this battle.
• #### omnimoeish - Monday, July 29, 2013 - link

While a jump from \$40 to \$200 is obviously 5x the price and sounds unwinnable. In the grand scheme of things, I would have no problem paying \$160 more to get a device that runs a full blown x86 chip that allows me to run Mac OS X, Windows and any software available for these OSs. This is like comparing a bike vs a motorcycle but still has all day battery life.

There's a couple of things to consider here as well. One is that we're assuming that there's really a use for more than 9 hours of battery life. I don't think there is. Assumption 2: there are no battery advances that could possibly make these x86 power consumption problems.

One interesting thing is that apps are so light weight, that unless Apple makes a hybrid a la Surface Pro, there's really very little reason to worry about getting more powerful processors. 99% of people are playing Angry Birds, surfing the web and these devices already have plenty of power.
• #### Kidster3001 - Tuesday, July 30, 2013 - link

The article was about whether or not Core is ready for tablet-like form factors. The comments have degraded into a discussion on pure power consumption of the Haswell chip vs ARM chip. There is not enough data to compare the chips directly.

It seems to me that Haswell may be the beginning of Core being good enough for a tablet-like form factors.
• #### damianrobertjones - Tuesday, July 23, 2013 - link

Could you please include the Dell Latitude 10 with extended battery? That would spank EVERYTHING with regards to battery life Reply
• #### dealcorn - Tuesday, July 23, 2013 - link

Based on Anand's earlier review of Valleyview (http://www.anandtech.com/show/6203/details-on-inte... I think Intel agrees with Anand's statement "I wonder if Haswell’s video decode engine just isn’t as low power as what you can get in most ultra mobile SoCs today." That appears to be the reason Baytrail includes an "Integrated Imagination Technology* VXD392Decode engine". Assuming the licensed decode engine is reasonably efficient to start with, when you build it using superior transistors with a superior process node, Baytrail should have class leading playback battery life. While Core products may catch up in due course, Atom will set a high bar on tablet efficiency in all areas. Reply
• #### Kidster3001 - Tuesday, July 30, 2013 - link

All Intel Atom chips use an IMG Video Decode engine. Many ARM vendors do as well. The same is true of the IMG Video Encode engine. Reply
• #### darwinosx - Tuesday, July 23, 2013 - link

The problem with Surface Pro is far from just the hardware. It's the fact that people hate Windows 8 in any of its forms. Reply
• #### Anato - Saturday, July 27, 2013 - link

Not sure if there is any point to compare, Haswell is so far apart from current ARM's when you take price of the whole platform to account. And thats something Intel will not fix. Reply
• #### purerice - Sunday, July 28, 2013 - link

In other words, performance per watt per dollar? You're on to something there.

My guess is this article is a preview for the real article 4 months from now when we have next gen Atoms and 4 months from then when we have next gen Nvidia SoC.
• #### LMF5000 - Tuesday, July 30, 2013 - link

Strictly speaking, you can't scale battery life down like that because of the Peukert effect - As the rate of discharge increases, the battery's available capacity decreases because more of the energy is wasted as heat by its internal resistance. To put it another way, a 54Wh battery only delivers 54Wh at a specific (low) discharge current. As the discharge current increases you'll find that the same battery runs out of juice after delivering only 50, 45 or less Wh.

Therefore, your results for the scaled battery are too optimistic. A smaller battery would be working harder to supply the same wattage. Therefore by Peukert's law its capacity (and thus battery life) would be LESS than predicted by a simple linear proportion.
• #### Les Slater - Wednesday, July 31, 2013 - link

It seems to me with all the discussion about power efficiency, performance and to some extent, x86 compatibility, it looks like Atom might be the eventual winner.

But the bigger picture is that there is, and will be, a trend toward offloading processing to some variation of an evolving cloud, right down to the pico cell. There's plenty of room for Intel, and others, to expand sales.

At an even higher level though, this is all heading toward commoditization where profits will be much slimmer.