I had a brief Twitter conversation with Anand Shimpi of Anandtech about this, and he was as perplexed as I was. Nobody could explain the technical basis for this vast difference in idle power management on the same hardware. None of the PC vendors he spoke to could justify it, or produce a Windows box that managed similar battery life to OS X. And that battery life gap is worse today – even when using Microsoft’s own hardware, designed in Microsoft’s labs, running Microsoft’s latest operating system released this week. Microsoft can no longer hand wave this vast difference away based on vague references to “poorly optimized third party drivers”.
The new Surface Pro 2 gets 6.6 hours of web browsing battery life. The MacBook Air 11″, which has more or less the same hardware and battery, gets more than 11 hours.
I have a Surface RT – the first generation – and as such, I know why. Windows 8 might have Metro running on top of it hiding a lot of it, but Windows 8.x carries just as much baggage, cruft, and outdated shit with it as previous versions of Windows have. Windows 8/8.1 – and Metro in particular – simply suck. Slow, clunky, jarring, cumbersome, battery-sucking, restricted, and limited, with a crappy selection of rush-job, rarely updated applications. You know how resizing windows on Windows 7 or OS X is all nice and fluid? Why, then, is it a slow and jittery operation that brings Windows 8 Metro to its knees?
It’s simple: just like battery life, it’s a symptom of Microsoft’s Windows team not having the balls to truly go for a clean break, as the Windows Phone team have done. And lo and behold, Windows Phone – even WP8, which runs on the same NT kernel – has none of the slowness and crappiness issues that continue to plague Windows 8 Metro (although WP has its own set of issues unrelated to these).
If you want a smooth, modern laptop today – get a MacBook. If you want a smooth and modern tablet, get the Nexus 7 or an iPad. Microsoft still has nothing to show for itself in these areas.
You are probably right about high-level reason. But I would be a bit interested in the mid-level reason, if there are some subsystems and features, that we could point fingers at.
BTW: are the any comparisons/benchmark for battery life of desktop Linux? (Preferably with up-to-date stack: systemd, wayland, latest kernels… you name the stuff)
There’s no such thing as up-to-date Linux stack: Ubuntu and Fedora divert very much and every other GNU distro is catching up more or less. Furthermore, the level of support of powersaving hardware features is very different. Another factor worth mention is that you really can’t compare Linux console to Windows 8 or OS X, and the GUI you’ll have to add to the Linux stack you choose would also influence the outcome dramatically. In fact the comparison that would make sense is something like “Fedora+GNOME vs. Ubuntu+Unity vs. openSUSE+KDE vs. Mint+Cinnamon vs. Windows 8” (these are nearly random distro+DE combinations, more thought required), and it should be carried on a set of divert hardware setups to make at least some sense.
Or, one could eliminate one variable, and just test:
Ubuntu+Unity, Ubuntu+GNOME3, Ubuntu+KDE4, Ubuntu+XFCE. Ubuntu+LXDE
And compare that to say Fedora running all the same WMs (with the exception of Unity, of course).
And, finally, compare that to OpenSuSE running the same WMs.
That way, one could compare how each WM/DE affects performance on the same distro, how the distro affects performance on the same WM/DE.
Would take a long time, but would be the only way to really get a good picture of battery life on Linux.
Add here several different hardware setups (AMD, pure Intel, Intel+Nvidia in combination with different wireless cards) and your results are pretty extensive.
I don’t know anybody who uses Wayland in production systems yet, or any distribution which ships it enabled by default. So why should Wayland battery life matter to any tester?
The Acer C720 Chromebook running ChromeOS is rated at 8.5 hours of battery life, which has been confirmed to be accurate in early reviews (e.g. [1]). It uses a 45 Wh battery compared to the 38 Wh battery in the 11″ MacBook Air.
[1] http://www.computerworld.com/s/article/9243287/Acer_C720_Chromebook…
GNOME Shell may act as Wayland compositor since 3.10, and they plan to have complete Wayland support by 3.12. Prooflink: https://wiki.gnome.org/ThreePointEleven/Features/WaylandSupport
Edited 2013-10-21 18:01 UTC
“Subsystems” indeed. Windows Vista brought that whole issue to the fore, with their mandatory DRM degrading (infecting?) the system’s entire ability to perform. Those computations take power that could be better spent on other work.
And then there’s the system call graph comparison from 2006: http://www.graphs.net/201304/server-wars-infographic-linux-or-windo… speaks volumes.
Wow, really, Vista DRM? I feel like I stepped in a time warp to 2007.
I just used that as an example, because it was visible even to non-tech types who noticed Vista ran slower than XP, even though Vista was newer. The “subsystems” problem has been with Windows since Win98, if not longer.
http://www.phoronix.com/scan.php?page=article&item=apple_mac_power&…
Take it with a grain of salt, its a few years old now, and its phoronix. He tries hard, but bench-marking is very hard.
Of course, if you want to do your own research, the bench-marking software is freely available.
Linux is unfortunately much worse on battery life these days, though mostly because the power saving features of modern CPUs, chipsets and GPU are not supported. It seems to be something Intel and Nvidia prefers to make exclusively available to Apple and Microsoft.
Though once you get to the long tail, the issues are how vetted and good all the third party applications are. This is where Apple gets to beat MS though their operating system is technically worse.
Edited 2013-10-21 18:44 UTC
I can tell you right now that not only does my Linux install (currently Debian Sid, but up until yesterday was Debian Wheezy) not constantly run the fan on it, but seems to run smoother overall (less disk accessing) than Windows 8 does.
This is on an HP Touchsmart tx1025
Check out
http://jupiter.sourceforge.net/downloads.html
Fuduntu used to use it while the distro was still alive and the Battery life was better under linux.
I use TLP on Elementary OS and get 4 hours out of my 6 cell battery
Nonsense. On Ubuntu my laptop can run for about 8 hours on battery, and that includes playing games such as FTL and OpenTTD, or running a series of java programs (eclipse, gradle, tomcat…) Dualbooted to Windows, it gives me at most 2 hours while playing the same games. Linux shuts down all but one i7 core, Windows doesn’t seem to do that.
I regularly travel about 3-4 hours by train (with a few changes in between). Windows never survives the ride, Ubuntu usually does with 50-60% battery left.
I’m only using open source drivers by the way. I can imagine that those binary blobs play less nice.
“These days” was perhaps true in 2012, but things happen fast. The open source radeon drivers have decent power management now (Catalyst always had, but despite the excellent performance and feature set, Catalyst still sucks), Intel’s is great. nvidia is probably fine as well, unless you use that fancy hybrid graphics mode to save power.
Pretty much spot on, its impressive the downsizing Windows has done — but much more work needs to be done in order to squeeze out every last bit of juice.
Windows Phone also benefits from a single focus on a single chip manufacturer (Qualcomm). Surface 2 apparently is very capable from a perf POV due to Tegra 4 but I still wouldn’t trust a Tegra at all.
One can only hope that now that the inmates no longer run the asylum (old WinDiv that dragged their feet on tablets for a decade) that things change for the better.
If you want a smooth, modern laptop today – get a MacBook. If you want a smooth and modern tablet, get the Nexus 7 or an iPad. Microsoft still has nothing to show for itself in these areas.
Some people might want a computer to run some software, you know, something other than webbrowser or Angry Birds. Some need 3D Max, some need VirtualDub, some need Pinnacle Studio, some need SONAR, some need Transcode, I need Unity 3D and Visual Studio, some people play games, etc.
There’s no other platform having all sort of professional tools running on it. Even for gaming there are 100x or 1000x more games available on Windows than on Os X.
Sure, if you only use a webbrowser Os X is ok. But Linux might be even better in that regard.
Buy a MacBook, install Windows.
Atleast that’s on my TO DO list just to confuse hipsters in coffee shops.
Better if you buy a MacBook, buy VMware Fusion or Parallels, and install Windows on it for your gaming stuff.
I want Windows to talk directly to my graphics card.
Then skin windows so it looks like OSX to confuse the poor blighters even more.
Then sigh as their systems carry on working while yours dies as the battery runs out.
Winners and losers everywhere then?
this coffeeshop hipster fanboi has had windows on his mac for a long time. i develop cross-platform apps and have enjoyed living mac and firing up windows when needed for many many years.
running other OS’s well has been an advantage of apple’s hardware for some time.
btw – i look around the coffeeshop computer users, and my observation is that most are just facebooking and web surfing, regardless of platform. i’ve always felt like i was the minority actually setting up and developing when mobile.
You’re right, people should use Windows if they want to run some software. For the past decade I’ve used nothing but Linux for work, hobbies, Masters degree work, and entertainment, but I don’t do all those things by using software. Instead my computer runs on cookie dough.
I also use Linux professionaly every day at work, but since we need higher quality output our work computers run on unicorn poop. You know, for stability.
My mom, a much less technical user than me, asked me which computer she should buy. At first I was going to suggest a Macbook Air so that she’d be able to use it twice as long as a Windows laptop thanks to having twice as much battery life, but in the end I got her a Microsoft Surface Pro, because she might want to run Pinnacle Studio some day.
FYI: Unity 3D is written mostly using OSX and Unity Tech is 100% mac shop.
Windows is quite behind when it comes to powersaving. The features landing now (like tickless kernel) were available in Linux years ago. Runtime power management (turning off devices not being used at the moment) is almost nonexistant in Windows. Additionally, Linux is in special position – the same Linux (or very similar – with android addons) is used on mobile devices and on laptops. Almost every optimizations mobile companies implement for cellphones also works on laptops, bringing power usage down.
More exotic OSes, like Darwin/Mac OS X, outpace Windows when it comes to integration with the hardware. For example Apple, controlling both hardware and software, was able to quickly introduce “half wake/half suspended” state – google for “darkwake”.
My knowledge is outdated, but with Windows 7 and Ubuntu, Windows 7 had longer batter life on same laptop. Anecdotally, Ubuntu running as a guest on Windows host had more battery life than running Ubuntu on hardware.
So this comes a little shocking (that and how someone compared Surface and MacBook Air, back in my days battery tests wre done on same hardware.
Edited 2013-10-21 16:08 UTC
One of the reasons for this could be the lack of power-management in GPU-drivers. You don’t mention if you were testing with F/OSS-drivers or closed-source ones, or even what GPU you had, but AFAIK neither the F/OSS-drivers for NVIDIA-cards or for the AMD-cards properly support power-management.
Uh, what? Why did I get modded down? I sometimes don’t understand the zealots’ logic, since e.g. http://nouveau.freedesktop.org/wiki/FeatureMatrix/ backs up what I said: not a single card has proper power-management.
I’ve modded you up as I agree with you that modding down based on your comment/opinion is wrong. However, I need to point out that (although it isn’t fully mainstream in the various distros yet) the Linux kernel radeon module has had DPM for a fair few months now, and it works as well if not better than catalyst in terms of power savings.
Well, the person I replied to never mentioned how recently he tested this, nor did he mention his GPU. I don’t know if he tested this before the Radeon – module got DPM implemented or if it wasn’t a Radeon in the first place, I just offered something that could have explained the difference. At least it did make a big difference in the past.
I +1’ed you back up. Happy? =P
It’s true in any case. The open source drivers for both nVidia and AMD/ATI are behind with regards to power and thermal management. All one has to do is visit the official forums of various distros for examples.
Linux battery life with the more recent Intel (Sandy/Ivy Bridge & Haswell) mobile chipsets is pretty similar to what you’d get on Windows, in my experience, though I’m not quite sure about feature parity.
Edit:
I have little experience with AMD, so the above commenter’s statement about DPM might well be true.
Edited 2013-10-21 17:22 UTC
Forget about the moderation rating on OSNews, it is broken by design.
At least to the extent of my experience, your assertions seems true but I am not sure if it still applies. things change fast on linux.
Frankly, don’t look for logic or, perhaps, reasonability from any zealots camps, their radicalism do not grant value for different opinions but only their own, it is a kind of deaf, blind, self-reinforcing, perpetual motion way of thinking full of idiosyncratic preposterous nonsense. 😉
Because people are morons and the OSNews moderation policies are braindead.
The latest Radeon drivers in Linux support KMS and power saving features. It was trumpeted as a huge accomplishment a couple months back. Many AMD-based laptop owners rejoiced!
OS that the producer cares about providing customers with backwards compatibility is slower to adopt new features than other Operating systems where they don’t guarantee any backward compatibility or even care about it.
What a surprise!
Usually Windows gets better battery life, Linux with Jupitur Applet or similar brought them more or less in line. I get about 4 hours on my very old laptop.
6.6 hours of battery life on a single charge and people are complaining? I am used to 4 hours if the screen is dark as f–k and I don’t fire up VS.
Edited 2013-10-21 20:09 UTC
It’s 2013, we should expect better across the board.
I recently bought a 2013 MacBook Air for school and casual surfing, mainly due to its battery life. Also, as a developer it is great to have access to different platforms.
I really love the battery life, it has just become more important to me. I also love the fact that I can just put it to sleep (close the lid), and leave it for long periods of time without the battery level going down noteworthy. The same goes for my iPad, but it’s mainly collecting dust after buying the MacBook Air. A truly mobile laptop with a powerful OS is just so much better than a restricted tablet.
My main laptop runs GNU/Linux. It has a terrible battery life (in Windows as well), but it’s an 18 inch more or less always connected to a power supply. Being that big, it’s more like a stationary computer with built-in UPS.
Linux has great power saving features, but there’s too much not-optimized software in most distributions and carrier/OEM-bloated Android ROMs.
Apple is leading, with a big advantage of producing both the hardware and the software. The same could be said about the Surface, but Windows just carries too much stuff after tens of years supporting all the hardware in the world.
Edited 2013-10-21 15:55 UTC
That is exectly the reason I was going to post in the comments.
Apple also has a lot less hardware that it needs to support with their software.
I will probably go MacBook Pro on the net laptop since the ultrabooks with a decent display are almost as much so I might as well get all the bling with it.
I have absolutely no graphics issues on my Surface Pro on any app resizing. I would assume the poor graphics performance is due to the RT and its inadequacies. I have also never experienced any graphics issues on other laptops, BUT I have never tried Win 8 on a non x86 cpu so I am not sure if Thom was specifically speaking to RT and RT hardware.
Re: battery life, I still think it’s tough to judge because it’s not a 100% fair comparison because Windows has the whole touch screen and higher res display to deal with whereas Apple doesn’t (comparing Surface Pro to MBA). Apple has also been making/controlling their own hardware a lot longer.
Despite all this, I still feel that if Apple did have a touch screen and higher res display, I wouldn’t be surprised if they managed to have significantly better battery life than the competition.
Wow, a 1.6 GHz processor (i5-4200U in Surface Pro 2) has worse battery life than a 1.3 GHz processor (i5-4250U in MBA)? Who would have thought?
So, you’re squarely staring at the CPU-speed? If you’d dig deeper you’d notice that the 4200U sports an HD Graphics 4400 GPU whereas the 4250U sports an HD Graphics 5000 GPU, which according to benchmarks is a lot speedier than the aforementioned one. The speedier GPU is likely the bigger reason for the difference in battery-life than the difference of mere 300MHz, especially since both CPUs have equally-high turbo-clocked speed — 2.6GHz.
Also to be noted and accounted for is the active digitizer and higher resolution on the Pro.
And the TN in the MBA uses less power than the IPS (assumed) in Surface Pro 2. Too many differences to take this article seriously.
Hummmm … i thought Anand was better than this, comparing a low resolution mac air to a high resolution touch screen.
No I’m illustrating that the hardware is NOT the same and that such apples and oranges comparisons should not be used to push Thom’s anti-MS agenda.
As for the CPU speed the the 20% higher minimum clock WILL result in more power consumption. (Even though the turbo frequency and TDP are the same…)
To my recollection, I was supposed to be anti-Apple. Care to elaborate? You and several of the fervent Apple supporters here might have to have a talk.
And with this last imputation you are left with no system at all, as I think you said linux is not the best OS option to use as mainstream personal desktop because of its current DE options.
Poor guy, better fold and stop using any desktop system. 😉
Really, it is kind of ridiculous when people point fingers at you when some of the criticism you bring are perfectly valid.
No system is perfect and the flaws they carry do not promote them to the top “worst things ever” automatically. My main system is a linux distro (openSUSE), but I am fine with the state of Windows 7 and could use it as my main desktop, which I don’t because of many, but manageable, details. Same can be said about Windows 8, even though I prefer the 7 iteration.
Perfectly valid? The comparison offered were two completely different systems which Thom labeled as “more or less the same hardware.” If the comparison offered were valid then the criticism would be too.
Please, read my next post http://www.osnews.com/permalink?575131.
And I suppose if hardware specs were removed from the equation, i.e. running Windows on the MBA, you would use Microsoft’s reasoning again: “crappy 3rd party drivers”.
Face it, there’s a major discrepancy that Microsoft refuses to address and/or explain here.
How can a different form factor, restricting battery size and capacity, be a good comparison?
Surface Pro 2: 7.4V, 4200 mAh
MBA: 7.6 V, 7150 mAh
Please correct me if I’m wrong.
Thanks,
Anthonws.
Those are the same numbers I looked up and found, so basically the Surface Pro 2 performs the same as the 2013 Air.
Not that anybody here probably cares
Perhaps you did not bother to finish read the whole article but here we go:
What do you think now? Perhaps, Windows do not explore some Apple’s software/hardware features? That is kind of possible as the system is not really put on any “stress” condition and as so it is quite possible to save a watt here and there.
My bad. I really have not read the entire article.
In that case, I would say the problem is going to be mainly drivers. If those drivers are not optimized, they won’t be able to make use of the hardware features to save power.
I don’t have a MB Air, but I would love to do a trace to find out the culprit of that power consumption.
—-
Edit: Nevertheless, the Digitizer/Touchscreen do take a bit of power, right?
Edited 2013-10-21 19:12 UTC
I guess so, but I was a bit more busy looking at Windows on MB Air. It is from there that I “got to the conclusion” that MS needs to do something to improve the battery life. Anyway, it is not “completely fair” matching as Apple tunes specially its software to run on its hardware. The other way around would not be fair too, running on a Hackintosh, for the same reasons. Would be an interesting reading, though.
50%? Nope.
http://www.anandtech.com/show/7417/sony-vaio-pro-13-exceptionally-p…
Note: the Normalized to Battery Size tests are interesting showing OS X with a slight edge but never anything near 50%.
By Anandtech numbers, you are right. Anyway, the difference on “minimum” and medium mode is still between 20 to 30 percent, what is not something to disregard.
As I said, fine tunning is probably the reason for the difference as it is way easier to do that when you have a leeway on front of you. Use it in too light conditions and the screen will be main drain, too heavy and it will be the processor.
OS that is only supposed to work on that hardware works much more efficiently than OS that is supposed to work well enough with most computers sold in the last 10 years …
While Thom and others are right … it maybe inefficient, but where are you that you can’t find a power outlet in 7 hours?
I will leave this here:
http://www.hanselman.com/blog/AMonthWithAnIntelHaswellPrototype.asp…
Edited 2013-10-21 20:20 UTC
True, but the trouble for MS to sell the Surface Pro 2 is that it will be competing with either MB Air or the new iPad to be released soon. The first has a way better battery life and a very nice keyboard, the latter is way cheaper (or will be, probably) and lighter. Really not an easy feat. On my case, I would probably pick the MB Air (actually I would pick something else, but that is another story).
?
Read the link, read Anandtech, the Surface Pro 2 has a 72 Whr battery and the MBA 11inches has a 38 Whr.
Also, installing Wondows on a MBA should give you a fair comparison (hint: it’s not pretty).
Edited 2013-10-21 18:57 UTC
I’ve read the review from AnandTech and the battery is 42Wh, not 72…
Damn, I mixed up.
Probably because Windows constantly has background processes, reading and writing the disk doing who knows what all the time.
So does every other OS.
He devoted several pages to the power handling systems built into OSX over the years. The new one Mavericks has all sorts of advancement here.
Apple doesn’t invent anything? They just invent ways to solve everything important. And they work for many years and revisions to solve the problem, unlike MS and most other sw companies.
I’m about to 1-click install a new OS onto my nearly perfectly-running macbook w/minimal interruption to my system. i’m even mid-development cycle but they are stable enough these days to run the upgrade anyway.
this new OS will actually give my 3.5 year old laptop another hour of battery life.
That’s invention, that’s innovation, that’s the kind of things that make people love and defend Apple.