Feeds

back to article Surface 2 MYSTERY: Haswell's here, so WHY the duff battery life?

Microsoft's Surface Pro 2 fondleslabs arrive on Tuesday, bringing various improvements over the first-generation models. But one area where the new tablets still don't wow is battery life – and that's a disturbing trend for Windows devices in general, some tinkerers have found. As hacker Jeff Atwood and AnandTech's Anand Lal …

COMMENTS

This topic is closed for new posts.

Page:

Anonymous Coward

Correction: Surface Pro 2 Core i5 dual core NOT Core i7 quad core

Love to have a high performance CPU/GPU combo that could throttle back power usage for all day battery life during casual use when on the move but that's not an option on any of the 2013 devices as far as I know.

0
0

Re: Correction: Surface Pro 2 Core i5 dual core NOT Core i7 quad core

Also all people failing to figure out here is that... Surface was using SD-Card as storage not SSD. the latency of SD-Card was miserable and CPU had to drag along... perhaps even CPU couldnt fig out when to idle is data was coming at crawl speed but not pausing to idle cpu. Surface 2 is using SSD hence releaving cpu too.

0
0

Maybe network activity blabbing about the owner.

11
2
Anonymous Coward

Windows just has a bunch of overhead

If I had to guess, I bet it comes down to the file system and associated baggage. Notice how it takes forever to do any file operation on more than, say, a thousand files if you're using Explorer. Move, delete, etc. ANY other operating system would do the same thing in a fraction of a second.

I bet any file operation in Windows just has a big chain of glop to chug through... going through all those idiot "nodes" to figure out that "My Documents" in the path name really means "Bill's Documents"... maybe backing up the file to a "restore point"... calling all the necessary AV hooks... anyway this would explain why it takes 4 seconds of hard drive grinding to bring up a new empty tab in IE on my Win7 computer...

13
0
JDX
Gold badge

Move, delete, etc. ANY other operating system would do the same thing in a fraction of a second.

Well OSX doesn't - it takes a noticeable amount of time for me.

A new tab shouldn't take 4s, and it doesn't on my several-year-old PC which only cost me £300 when new. So more likely something is weird on your machine.

4
3
Silver badge

Re: Windows just has a bunch of overhead

Quite probably system bloat, but maybe likely it is due to DRM? Consider this analysis:

http://www.cs.auckland.ac.nz/~pgut001/pubs/vista_cost.html

Shame none of that protects you, the owner of the PC, from malware...

Has anyone compared XP with Windows 7 on the same hardware to see if this is a factor?

1
0

Re: Windows just has a bunch of overhead

Guessing and betting is all very well but this measurement on Anandtech is interesting, mainly because if its not a benchmark artefact and relates to real usage (which on the face of things it seems to) this implies there's an opportunity to get substantial improvements in common usage scenarios of Windows on x86 hardware, either via software or better hardware or both. Intel must have something they could say on this.

I'm always sceptical about benchmarks until I know they are representative of real world usage. Twenty years ago I discovered a driver from a leading graphics card vendor tricking common benchmarks at the time into the illusion it had twice the pixel fill rate that it actually delivered. I'm not suggesting anything underhand is going on in this instance, just that it is often easy to favour benchmark type scenarios by accident or design.

0
1
Bronze badge

Re: Windows just has a bunch of overhead

Operations on thousands of files that take ages are usually the result of 8.3 short filename generation, if you switch it off (as is the default on Server versions) things go a lot quicker.

The overhead still isn't nearly as bad as on Mac OS X, which is still using a poorly performing big-endian file system despite the fact the CPU in every Mac is little endian, resulting in bytes having to be re-ordered on every read/write from the filesystem.

As to the overall power performance, I suspect it's down to the fact that power consumption is heavily dependent upon drivers and no manufacturer of PC laptops is putting the same effort into driver development for Windows that Apple are for OS X (including Apple, whose Bootcamp drivers are legendarily crappy)

0
0
Anonymous Coward

Re: Windows just has a bunch of overhead

"The overhead still isn't nearly as bad as on Mac OS X, which is still using a poorly performing big-endian file system despite the fact the CPU in every Mac is little endian, resulting in bytes having to be re-ordered on every read/write from the filesystem."

If true, I can't imagine this adds much overhead anyway. I'd like to see some benchmarks. But what I'm talking about is the fact that I was working with ~10,000 photos this morning and moving them around or deleting them took about ~2 seconds in OS X. On my Win7 computer it takes at least an order of magnitude longer.

"As to the overall power performance, I suspect it's down to the fact that power consumption is heavily dependent upon drivers and no manufacturer of PC laptops is putting the same effort into driver development for Windows that Apple are for OS X (including Apple, whose Bootcamp drivers are legendarily crappy)"

Read the article. We're not talking about arbitrary PC hardware where your explanation might be plausible. We're talking about Surfaces, i.e., hardware that Microsoft owns.

3
0
Bronze badge

Re: Windows just has a bunch of overhead

What is also interesting is that when Engadget did their video playback test they got almost fourteen and a half hours battery life out of the Surface 2, around double that of what Anandtech posted but also more than four half hours more than the Nexus 7 and over two hours more than the iPad4.

http://www.engadget.com/2013/10/21/microsoft-surface-2-review/?a_dgi=aolshare_twitter

Of course there are differences in the test but with most sites just copying and pasting Anandtech's results or not saying what their tests involve (The Verge) its difficult to say who's results are the most accurate, what's causing the battery drain on Anandtech's results and whether that cause is something the average user would encounter.

0
0
hfp

Re: Windows just has a bunch of overhead

Haswell supports the MOVBE instruction that allows to fix the Big Endian'ess on the fly. This was previously only supported on Atom processors due to Atom's exposure to the embedded domain.

0
0
Unhappy

Must Try Harder

As an speed test experiment, I installed Windows 7 on both a bootable partition and as a VM on my MBA. While I found that most benchmark speeds were about the same between the two (graphics is slower, file access is faster), battery life varied remarkably. While I get 8 hours of battery life using the VM, I get less than 6 hours booted into Windows 7.

My impression is that Windows has always had lousy power management, but simply comparing one Windows laptop to another provided no real frame of reference (quiet down there in the back, you penguins). But Microsoft is attempting to enter a field occupied by competitors who both have OSes with several iterations of serious power management refinement behind them, so it's hard to hide their (relative) inexperience in mobile computing.

12
0
Silver badge

@Quxy Re: "occupied by competitors who both have OSes....."

That was an interesting experimented you ran. I think that I would also point out that comparing x86 and ARM setups (which they have done in fact according to this article) rather confuses the picture*. To what extent Intel x86 architecture is intrinsically more "thirsty" than ARM SoCs must also be involved here. However, having said that it is clear that Redmond still have some work to do.

*Had they included a Windows RT slab (running on ARM) that might also have been interesting.

0
0
Silver badge

Re: @Quxy "occupied by competitors who both have OSes....."

It was good of you to try different OSs, Quxy. Which browsers were you using?

Apple are claiming that Safari in Mavericks has been designed to extend battery life (such as suspending activity in browser windows that can't be seen) so it seems reasonable to assume that they might have made some efforts in that area in the current version of Safari.

Windows browsers vary in their use of system resources, too.

0
0

Re: @Quxy "occupied by competitors who both have OSes....."

I was only running 64-bit Windows 7 (one standalone, one in a VM), and my browser was Chrome in both cases. No flash (although the OSX flash problems seem to have eased up lately). But most of that time was spent in CAD software (i.e. lots of little files) and Microsoft Office, not web browsing. The improvement in power management must have been provided by the host OS, not by the applications.

0
0

Re: @Quxy "occupied by competitors who both have OSes....."

Arctic fox, they did include a Windows RT slab, or rather, two: both Surface RT and Surface 2, which are both Windows RT (running on ARM of course). Both have shorter battery lives than any of the ARM-based tablets listed except the Samsung Galaxy Tab 3 10.1. Perhaps they should also add the new Nokia Windows RT tablet to the comparison.

But yes, their comparisons are a bit confusing -- more so without the reader knowing which, if any, of the other tablets are x86/x64 based. (Not sure if you're referring to 32-bit Intel, i.e. x86, or 64-bit which would be x64 or amd64. Both Surface Pro and Surface Pro 2 are 64-bit Intel.) I like Quxy's idea of comparing different OSes on the *same* hardware -- not just hardware with the same processor. I'd like to see it done on the Surface 2, as I think it has potential as a tablet but dislike the Windows 8.x OS (RT or otherwise).

And yes, the Reg article is wrong, the Surface Pro 2 has not an i7 but an i5-4200U (one of the Haswell line, released Q3/2013).

0
0

Why is Windoze <anything> so awful? See picture

http://gyazo.com/ac35c48ca67ae47d82ec00d5168ec5b8

2
5
Silver badge

Need better power numbers for uptake

MS has often tried to pitch tablets at professionals such as medical workers around in hospitals or surveyors working on contruction sites.

To do that you need a minimum battery life of a shift. Sure, people could work around this issue by recharging during breaks, but being able to do a whole shift without a recharge is a significant psychological hurdle for any such tech.

13
0
JDX
Gold badge

Re: Need better power numbers for uptake

If it's only 14% less than an iPad 4, it's in the same ballpark though so not really a huge news story. If it was 50% less... although how iPad4 compares to iPad3/2 would be interesting.

3
4
Silver badge

Re: Need better power numbers for uptake

"To do that you need a minimum battery life of a shift. Sure, people could work around this issue by recharging during breaks, but being able to do a whole shift without a recharge is a significant psychological hurdle for any such tech."

And that's assuming your shift is EIGHT hours. What if you have a TEN-hour watch? Or TWELVE? Since you bring up the medical profession, I think some can go as high as 18 hours at a time (say for an 18-up/18-down rotation).

3
0
Silver badge

Re: Need better power numbers for uptake

@JDX: the aggregate life may only be 14% different but subtract the constant drains like the screen and backlight and that OS difference gets much larger.

@Charles 9: in reality you need it to last a full shift at the end of a year or 2 in constant use. With older Li cells capacity could half in 18months or so. The Li poly cell in my phone is holding out better but still had a noticeable drop in capacity after a year.

3
0
Bronze badge

Re: Need better power numbers for uptake

Also they compare the 10" Surface to the 7" Nexus not the 10" Nexus which scores almost exactly the same as the Surface.

0
3
Silver badge

Not to hard to figure out

Just compare the size of Win 8/8.1 (any version) to any of the other OSes... It's HUGE!

All those lines of code are chewing up a lot of cycles doing... whatever the hell they're doing.

Seriously, can't they get a grasp on the fact that all an OS is supposed to do is provide a platform and housekeeping for the stuff that we actually want to use, without getting in the way?

14
3
Bronze badge

Re: Not to hard to figure out

I'm pretty sure the Windows source code isn't one giant while loop, calling all the code all the time.

P.s. Can I have a job? I assume you must be a CIO if you're saying stuff like that :)

5
3
JDX
Gold badge
Joke

Re: Not to hard to figure out

Don't be silly Robert, it absolutely has to load every single device driver in the known universe and run them all at once just in case.

6
1
Anonymous Coward

Re: Not to hard to figure out

"Just compare the size of Win 8/8.1 (any version) to any of the other OSes... It's HUGE"

It seems to be smaller than most Linux distributions....

2
14
Anonymous Coward

Re: Not to hard to figure out

The iOS 7 update for my iPhone was nearly a gigabyte if I remember correctly... a GIGABYTE, for an UPDATE... and that's just a PHONE!!!

2
2
Silver badge

Re: Not to hard to figure out

"It seems to be smaller than most Linux distributions...."

That's a stupid comment even by your standards as I'm sure you know most Linux distros ship with MASSES of user programs. FYI I've just generated a custom distro using SUSE Studio (brilliant) and with all the software I want it is a 270MB download

7
0
Bronze badge

Re: Not to hard to figure out

I think we need to upgrade you to XP! :)

0
0
Linux

Battery life under Linux?

We'll have to wait for these puppies to ship, though I'm really interested in W8.1 tablets dual booting Linux (specifically Lubuntu). It will be really interesting to see if battery life improves using a Linux distro that's a bit round in the middle (Ubuntu) and again with a distro that's quite lightweight (Lubuntu / Debian + XFCE / etc).

I'm personally waiting on feedback about the Asus T100 (64GB) dual booting with Linux. If it can end up triple booting W8.1/Android/Lubuntu then I'll end up buying two.

2
0

Re: Battery life under Linux?

Why dual boot? Not that I want to overdo the Linux angle but for the past couple of years or more my little ThinkPad Edge has been running Ubuntu for most of my work but with a VirtualBox copy of XP for the odd Windows program I have to use, sharing files via SAMBA and printing into a virtual printer that produces PDFs on the host.

If I'm not using the Windows instance I suspend it (it seems to chew CPU a bit more that is good for it, even at idle it's using 65% of one of the four cores) and resume it when I need it. And if I'm plugged in to the mains I just leave Windows running all the time, it's not THAT big an overhead.

I'm sure you could do the reverse, hosting Linux under Windows too. Running both at the same time seems a much better solution than dual boot unless you have a really good reason for the latter.

4
0
Silver badge

Re: Battery life under Linux?

Well, if you are primarily running Windows (perhaps the applications you use require it), it isn't a bad idea to be able to dual-boot Linux too, if only as recovery environment.... though to be fair, you could just run it off a live CD / thumbstick in those circumstances.

Another reason might be to purely use the second OS for applications such as online banking... a Linux guest OS wouldn't be immune if it its Windows host OS had contracted a keylogger, for example.

And then Steam are pushing a Linux-based OS for gaming performance reasons. Since these days many people have a second internet device to hand (phone, tablet), then restarting your main computer to boot into a gaming OS isn't the inconvenience it used to be (especially if booting from an SSD).

0
0

Re: Battery life under Linux?

It's really personal preference. For me, as the device comes with a licensed W8.1 (and I don't have any spare licenses) I'd want to leave it where it is, especially as I know booting into that partition will give me 100% hardware utilisation with 100% of the memory (with the Asus T100, that's only 2GB RAM). The same kinda goes for Linux - I want it to have access to all of the hardware with minimal bloat behind the scenes, so that rules out virtualising Linux within Windows. Truth be told, I probably will run VirtualBox for some small VMs, things like terminal only CentOS that's only allocated 256MB RAM and resides on the microSD card.

Why triple boot with Android? Mainly so I can have access to a traditional sandboxed tablet for testing and certain Android only apps that are linked to my Google account.

0
0

Ditch Windows, save the planet?

So, quite apart from the extended battery life that we can all enjoy by ditching Windows for some flavour of *nix, think of the financial and environmental benefits. If the world ditched Windows today, how many power stations could we turn off?

10
1
Silver badge

Re: Ditch Windows, save the planet?

If the world ditched Windows today, how many power stations could we turn off?

Or, more worryingly, if we ditched Windows today, how many power stations would we have to turn off, because their monitoring and/or control systems are Windows-based?

7
4
Bronze badge

Re: Ditch Windows, save the planet?

Probably them all, since all the software used to run businesses runs on Windows.

2
5
Silver badge

Re: Ditch Windows, save the planet?

Hehe, do you think I'm getting downvoted for saying that: 1) power stations' monitoring and/or control systems might be Windows-based; or that: 2) it might be worrying if 1) were the case.

Maybe I'm collecting downvotes from both sides. :-)

2
0
Bronze badge
Coat

Re: Ditch Windows, save the planet?

Shush, you'll get the Tory party advocating linux as a means to cut energy bills...

0
0

Re: Ditch Windows, save the planet?

No, Power stations both Nuclear and other tech. have fail safe monitoring/shutdown and a lot of day to day control that is implemented in pure HARDWARE (big contactors, multi channel hardware interlocks with cross monitoring etc. etc.). Even normal control system elements with software (PLC's etc.) are EXTREMELY reliable if left to do their thing (although they may have crap security on external interfaces). A power station is not a P.C peripheral, so can all those scaremongers out there stop with the 'Sellafield is run by Windows 95 on a 486DX' crap.

0
0

Re: Ditch Windows, save the planet?

@45RPM

Would be interesting to see your figures on how much better versions of Linux perform on the same hardware for these benchmarks. Couldn't see numbers on Anandtech or codinghorror. What sort of improvements are you seeing?

0
1
Silver badge
Meh

Re: Ditch Windows, save the planet?

"Probably them all..."

That explains the state of the economy in a nutshell then...

0
0
Nya

Nothing to see here but flame bait

Come on El Reg, surely on a quick look. But the devices with longer battery life are smaller (7" or the nVidia handheld), or the only real exception is the Ipad which simply has a bigger battery since (same with the Asus) only real machine to compare it to on that list is the (old) Nexus 10 which its comparative with.

4
2

Re: Nothing to see here but flame bait

What about the MBA 11" with smaller battery?

2
0
Anonymous Coward

Re: Nothing to see here but flame bait

But as noted in the article, they took a 15-inch 2009 MacBook Pro running OS X 10.5.7, Windows Vista x64 SP1, and Windows 7 RC1 and the battery life was much better ON THE SAME MACHINE when running OS X.

No flame bait going on here.

4
1
Anonymous Coward

As soon as your back is turned

Windows seems to "catch up" on all that stuff like indexing and fragmenting the disk (intentional use of the word).

Antivirus too, you'll need that of course, face recognition software checking your family snaps for bearded men etc..

I expect they will get there one day, as long as the corporate battery does not run out first.

6
0
Gold badge
Unhappy

All those *automatically* started "services" perhaps?

And that PoS "indexing" service,

WTF that about?

Do you get the feeling that MS work really hard to stop people figuring out what does not need loading/starting/running ?

2
0
Bronze badge

Sample size of one

Fixed a PC the other day that the owner had replaced because it was so slow, there was so much HD activity it sounded like an angle grinder and ran like a dog with no legs.

Switching out Norton 360 for AVG made a difference, but the major culprit was MS's readyboost service which appears to achieve it's reason for existence best when disabled.

Maybe this is wrong, but I get the impression that all those services that are supposed to 'run when idle' simply check for low CPU usage, which is generally always true these days, rather than checking for low HD activity before they start grinding the disks.

6
0
Thumb Up

Re: Sample size of one

@jubtastic1 - full agreement - the best way to speed up a Windows laptop seems to be to turn off all of Microsofts performance and productivity "enhancements".

6
0
Happy

It's the spooks

Could MS be so firmly embedded with the security services that windows eats battery sending logs of all that you do back to the mothership?

3
1
Anonymous Coward

No fair, you say.

perfectly fair. Windows 8 only really works for fondling (web/email/youtubecats/music), it fails abysmally for real worth.

Both an iPad and a Nexus will also do the aforementioned fondling tasks equally well at a quarter of the price (or an 8th of the price in the case of the Nexus 7).

When Microsoft decided to gimp Windows, they didn't understand they were cutting out the key area the PC was being used for, and sending them deep into the realms of what the £99 tablet was doing.

Total bozos.

4
5

Page:

This topic is closed for new posts.