Hacker News Reader

Comment by blinkingled | Hacker News Reader

Windows and Linux are general purpose OSes. The same kernel runs on high throughput servers and low power tablets. The apps are written with little focus on battery life. The drivers too. Microsoft doesn't write their own drivers and then there's firmware too. It's the one area where Apple has really taken advantage of the vertical integration - they do everything from firmware to most apps people use.

But both Windows and Linux are more than capable to get this all sorted. Like Google showed with the Nexus 7 - focus is all that's needed. It's just harder for Microsoft considering everything they need to juggle.

Edit: Fun fact: Apple's own Boot Camp drivers disable USB selective suspend on the 2013 Air! Check our powercfg /energy for more fun :)

Edit 2: Surface is Tegra 4 SoC isn't it? Microsoft still is limited by Tegra's power characteristics as far as what I can tell from Anand's review. So the integration story is better but still no match to Apple.

View on HN

Perhaps you missed the part about Mac OS X (also a general purpose OS) using significantly less battery than windows on the exact same hardware. And that's not even considering OS 10.9 which is due to be released soon which has a bunch of battery life improvements (rumored to be around a 10% to 15% increase).

There's nothing general purpose about Mac OS X as far as hardware support is concerned. Just like iOS, it only has to target a very small set of hardware configurations.

Edit: as pointed out elsewhere, this doesn't excuse MS from not taking advantage of the same vertical integration on their own hardware...

As far as I know they still use the Darwin kernel in iOS. Which means it scales from desktops and servers all the way down to phones... That seems pretty general purpose to me.

Additionally its folly to think servers can't gain from power management.

Most servers are idle, or running maybe 40%, might as well let those other cores sit idle if they aren't used. Example is at idle with power management our latest xeon draws 300W versus 390 without. 900W at full tilt and all power management disabled.

The excuse that windows is a "general purpose" OS doesn't fly in my boat. Granted I don't use Linux for laptops generally but power management should be something any OS does somewhat sanely. We're moving to Amdahls law now away from Moore, also is there a law yet on power requirements for appplications? So power consumption is only going to be more of a basic feature than bolted on. That and doing things like using avx or gpu's for array multiplication because the power requirements are cheaper etc...

See especially people who used to have performance problems realising that the server was running a screensaver.

Terminal Server running serverside, not clientside, screensaver is poor for performance (https://groups.google.com/forum/?hl=en-GB#!search/windows$20...)

Q: My server has poor performance. (https://groups.google.com/forum/?hl=en-GB#!search/windows$20...)

I genuinely don't understand the Google groups interface. So I might not have linked to the correct messages. Some scrolling might be needed.

The Darwin kernel scales down as far as an HDMI adapter SoC[1]

[1] https://www.panic.com/blog/2013/03/the-lightning-digital-av-...

Thanks for linking to that. I read it before the anon comment about the kernel so it was nice to catch up.

No one uses Darwin for servers, not even Apple.

No one except a fair amount of people: http://www.macminicolo.net/

You can turn old laptops into "servers", or raspberry pi's, or hell even a commodore 64 for that matter. As long as it serves something of value over a network it qualifies.

Whether its a good idea is another matter.

I think the fact you felt the need to put the word servers in quotes does a lot to show that you and justincormack are talking about computers serving a similar role, but with such laughably different requirements that it's obvious what point he was trying to make, and your statement, while technically correct, doesn't really change his point.

You are correct though, people do use commodity and low-end hardware for extremely light server needs, and even in data centers.

Apple took a shot at traditional datacenter rack servers with their Xserve product line. I don't think it failed because OS X couldn't scale.

I put servers in quotes to denote something that serves that wasn't purpose built for the role. Nothing more, the context from justincormack wasn't lost but there are a lot of uses for osx servers, for example compile farms.

What I was referring to, which I thought was apparent but failed to convey, is that this perception of nobody doing it is a fallacy.

We are saying the same thing. My post wasn't meant to contradict yours, just clarify.

> What I was referring to, which I thought was apparent but failed to convey, is that this perception of nobody doing it is a fallacy.

I just think that in this context the "nobody" you are referring to and the "nobody" he is referring to are not equal sets. Yours is a superset of his, which is why both statements are true (in spirit), while yours is true in fact.

True enough, think its been a long day for me then. Sorry for the confusion on my end.

I do believe that means it is beer o clock time. Cheers and no worries mate.

I'm pretty sure some of these [1] are still floating around - they were only discontinued two years ago.

[1] https://en.wikipedia.org/wiki/Xserve

What's this then? http://www.apple.com/ca/mac-mini/server/

I know of a lot of businesses that run their servers on OS X.

Plus, consumers don't really care. They just want the better product. "Yes, Mac's get twice the battery life but Microsoft has to support more hardware" doesn't really cut it. Sure, I appreciate it, but when it comes time to spend $2000, I simply want to know that I'm getting my money's worth.

To some people, the one that supports more hardware is the better product. Especially when they are spending $2000.

He was talking about one $2k laptop. Once you buy one windows version of that, knowing that the software installed on it can support other laptop is not going to make you less sad about your short battery life...

Yes, Linux is very cool in that respect. I really home Linux continues its long and relentless march.

Even with extra hardware support, I'd like to know the details of exactly what OS X is doing that Windows doesn't.

"Supports extra hardware" sounds suspiciously like the "unoptimised 3rd party drivers" excuse that Microsoft used to throw around.

I find it very difficult to believe that supporting extra hardware somehow costs Windows 50% of the battery life of a Macbook Air, but somehow doesn't have the same costs in terms of application performance.

>extra hardware somehow costs Windows 50% of the battery life of a Macbook Air

See mjg59's reply and my note on USB suspend elsewhere in the thread. (OS X turns off lot of things using non standard firmware implementation and uses out of band notifications that Windows and Linux aren't made aware of. Apple's Boot Camp drivers also turn off USB suspend on Windows. Those things matter a lot for power management.)

On the Air Apple disables USB suspend. Can you blame Microsoft for that? Replace Spoke with Asus/Lenovo/Dell/HP and you've got quite the problems on your way to"Windows Gets Great Battery Life!" goal post.

For surface - yes they're being a bit sloppy. They should work closer with their hardware manufacturers to fix things such as SoC current leaks, display power draw etc. But they are still limited by say NVidia Tegra 4 power characteristics. Microsoft doesn't design their own SoC like Apple does with A7 for example.

Mac OS X is much better than Windows 8 on the same Apple hardware. I think the key to Max OS X's efficiency is clearly the optimised drivers. Nevertheless, MS has to step up, good for us!

OSX is not a general-purpose OS.

That might be true, but doesn't the whole "vertical integration" argument get thrown out the window when he describes the problem on Microsofts own hardware, the Surface Pro?

Certainly Microsoft should have the ability or money to optimize those drivers for maximum battery efficiency.

Quite simply, no.

Performance optimizations are all about trade-offs, and what might be a good trade-off for Microsoft's own hardware could be a bad trade-off for 95% of the other hardware on the market.

In theory, Microsoft could ship a second set of bits for their own hardware. But now you start to have nightmares about the antitrust lawsuits in the 90s. Why is Microsoft keeping a secret faster Windows that only it gets to use? That's the headline anyway.

The only way for Microsoft to achieve the kind of vertical integration that Apple has would be if it was willing to destroy its existing OEM partnerships. And risk a lawsuit from the DOJ. And this is a crazy strategy, because Apple has all the experience manufacturing hardware, and Microsoft has very little.

>Performance optimizations are all about trade-offs, and what might be a good trade-off for Microsoft's own hardware could be a bad trade-off for 95% of the other hardware on the market.

Why not have an "Extended-Battery profile" that you can turn on and off, that will have Windows be more aggressive with respect to power conservation? It'll be toggled on automatically on Surface-type devices, and turned off for desktops and servers (though servers may also benefit).

So give me a break. Where there's a will there's a way.

>The only way for Microsoft to achieve the kind of vertical integration that Apple has would be if it was willing to destroy its existing OEM partnerships.

Hyperbole much?

(I can't reply to your child comment so replying here.)

>Any move to give preferential treatment to MS's own hardware is the push the OEMs need to convert their Windows Tablet initiative into an Android Tablet one.

I hope Microsoft's OEM partners don't think this way for the sake of progress. It is a scary thought that MSFT/Google would not optimize their reference hardware for the sake of pleasing OEM partners. I have a Nexus 4 phone. I certainly hope when they write an update for mako, they optimize it for mako.

It doesn't mean OEM partners can't use similar optimization. However, to my naive mind it is absurd to cripple reference hardware for the sake of pleasing OEM partners.

I already told you why not. Because anything that gives the Surface an advantage has that antitrust lawsuit smell.

> Hyperbole much?

Again, no. Any move to give preferential treatment to MS's own hardware is the push the OEMs need to convert their Windows Tablet initiative into an Android Tablet one. Not to mention having potential ripple effects in unrelated businesses lines like spooking HP into doing more Linux servers.

For better or for worse, Microsoft's most valuable asset is its OEM partnerships. It's not as cool a trick as it used to be, but it's the only card they have to play.

>I already told you why not. Because anything that gives the Surface an advantage has that antitrust lawsuit smell.

Who says it needs to be Surface only? Who says it needs to be limited to Microsoft hardware?

If the changes need to be in the driver or hardware layer, all Microsoft needs to do is release specs and then setup a certifying service for OEMs. Something like "Battery Optimized" cert that the partner can then communicate to the end-user. However, I have a feeling that lots of improvements can be done at the OS level to boost battery life before you start thinking about changing your driver model.

Bottom line this can be done. Windows tablets can and should be able to match Apple hardware in battery life and I have no clue why you're arguing they can't.

> all Microsoft needs to do is release specs and then setup a certifying service for OEMs

So this exists. It's called the Windows Logo Program for Hardware. The reason we are reading this article is that it failed to do the job.

What you seem not to be understanding is that in order to "match Apple hardware in battery life" one also has to "match Apple hardware" in the small number of distinct hardware configurations sold per year.

Taking that hardware configuration and calling it "a standard" doesn't actually do anything. If your standard is broad enough to support 100 hardware configurations, battery life will be poor. If your standard is narrow enough to only support 2 hardware configurations, OEMs will not build it, because Dell has no way to differentiate itself from Toshiba.

To match Apple, Microsoft needs to cut back to maybe 3 models of Windows tablets per year. You can make that into a standard if you like, but no OEM will ever adopt it. They will all (correctly) point out that such a standard is designed for Microsoft's hardware team to implement and to hell with all the third parties.

Google's Nexus 4 is a useful example here. It's in pretty much the same position as the Surface, and it also suffers from low battery life.

>Windows and Linux are general purpose OSes. The same kernel runs on high throughput servers and low power tablets. The apps are written with little focus on battery life.

What are you saying? That Microsoft can't build a consumer OS with great power-management because the kernel also needs to work on high throughput servers?

Come on.

>But both Windows and Linux are more than capable to get this all sorted.

Bingo. So the questions stands, given that the consumer market is a huge priority for Microsoft and that iPads and iPhones and Android devices represent some of the biggest competition for Microsoft, why haven't they done this already?

> Microsoft can't build a consumer OS with great power-management because the kernel also needs to work on high throughput servers?

Not only servers but they need the same kernel design to work well across a ton of different hardware/firmware combinations, form factors (Phone, Tablets, Laptops) and they do an OK job. But to excel they need to put in a lot more effort (working with OEMs, working with SoC makers is another challenge - everyone optimizes for Android now a days including Qualcomm and Nvidia). So yes, they can do it but it is exponentially more challenging for them due to their legacy and current market position.

Funny, the Linux kernel performs excellently on my $35, 800MHz, 400mA ARM RaspberryPi embedded machine, my x86 desktop, and some of the world's largest supercomputers with who knows what architecture.

For a company with hundreds of millions to spend on R&D, you'd think Microsoft could match a bunch of hobbyists.

I am as big a Linux fanboy as anyone - but Linux doesn't do power management across different set of hardware consistently and when it does it is with mediocre results. In fact you'd be hard pressed to find a PC Laptop running Linux that gives you great battery life. It's largely a miss. So performance yes, great power characteristics no - that's too hard due to sheer number of variables.

My bet: backwards compatibility. Better power management requires changes in behaviour of the OS and it is very hard to change behaviour without breaking things.

Suppose the USB sleep state handling code is suboptimal. Is that because programmers paid less attention to power usage a decade ago, or because some devices implement a power state incorrectly? If the latter, which devices? If Microsoft isn't 100% sure they know of all devices for which it is needed, they may choose to leave the less efficient code in place regardless of whether a deficient device is plugged in.

An example might be a pet peeve I have with Windows: its tendency to 'discover new hardware' when one plugs a bog standard mouse or keyboard into a USB port for the umpteenth time. That in itself is baffling, but it can take Windows half a minute or so to figure out that it has a driver for the device and start it. I do not have the faintest idea what it does in that time, but it must be unavoidable within the constraints microsoft sets itself, since it hasn't been fixed in a decade or so (but it has been improved quite a bit) and I bet it must have itched quite a few engineers.

Whatever the cause, plugging in a USB mouse must use way more power on Windiws than on a Mac.

It's not baffling, it's manufacturers implementing USB standards improperly, and Windows doing the least-bad thing.

http://blogs.msdn.com/b/oldnewthing/archive/2004/11/10/25504...

Suppose Microsoft rewrites Windows to optimize for the Surface platform. How long would it take for Microsoft to stand accused of unfair business practices in the press? How much risk would they run of being hauled into court in the EU or US?

They just paid a billion dollars for failing to provide browser options in the EU last year.

They paid a billion dollars for disobeying a direct court order. You phrase that as if they paid a fine for randomly not doing something.

Your suggestion that they are equally vulnerable to the same thing for solving a difficult, market driven technical challenge seems like a stretch to me.

Who says that Windows need to optimize only for Microsoft hardware?

> What are you saying? That Microsoft can't build a consumer OS with great power-management because the kernel also needs to work on high throughput servers?

It's more the drivers than the kernel. Apple has a lot more control over OS X drivers.

Microsoft picks all the hardware, and corresponding drivers, that goes into the surface.

Right, but the original post said the same about Linux.

With very few exceptions, Linux drivers live in the kernel tree, so the Linux community has absolute control over them. The claim doesn't really seem to make sense. It's not like power management is unimportant in a server rack.

In fact, there's a really nice talk somewhere by (I think) the LWN guy, about how running the same kernel on mobile and on big iron gives huge benefits to development -- it turns out that the needs are really not so different after all, and power management work intended for the mobile space helps big iron (and vice versa, cf. PowerTop from Intel).

Nothing is free - everything is a trade off. Hardware isn't magical. There is cost to how fast you can switch between various C states on Intel CPUs for example. There is hardware that explodes if you play too hard for another example. Servers need maximum performance and low latency - guess what aggressive power management will do to that?

You talk about drivers but what about firmware? Have you looked at how kludgy and unreliable ACPI implementations across hardware are? What about the in kernel drivers being mostly reverse engineered - how many hardware vendors care to provide power management docs for their hardware?

Everything is possible of course - it's just that it is way harder to do when you support everything from toasters to big iron in case of Linux and Phones to gazillion varying PC hardware combinations in case of Microsoft.

> But both Windows and Linux are more than capable to get this all sorted. Like Google showed with the Nexus 7 - focus is all that's needed. It's just harder for Microsoft considering everything they need to juggle.

Couldn't they just hire someone and say "Fix this. Identify the problem and we will support your findings to get the fix implemented."

It's not like they lack the financial means. I just don't think there's anyone at the company that gives a shit any more (or their is no culture to support the few people who do give a shit).

Microsoft in 2013 (actually, for a very long time now) focuses too much on the sizzle and not enough on the steak imho.

Yup, no one there cares about power consumption or the core OS development. /s

http://blogs.msdn.com/b/b8/archive/2012/02/09/10260956.aspx

http://blogs.msdn.com/b/b8/

And yet, we have this article.

It's not their problem most of the time, it's 3rd party apps and drivers. They do however have various levels of tests and certifications for both apps and drivers, some of which focus on energy efficiency.

Microsoft has a certfication programs. Sadly, those programs are notoriously weak (see the outrage about "Vista certified" machines not actually being able to run the hardware accelerated desktop). This was done to please hardware manifacturers and allow them to put new hardware out quickly. Applying pressure here and denying certification to drivers not implementing power features properly would have been the way to go.

Microsoft does have leverage, they just didn't use it for a long time to allow for a quantity over quality approach.

Microsoft doesn't have much leverage because the US Justice Dept sued Microsoft to enable OEMs to do almost anything they like. The "Vista certified" debacle showed just how little leverage Microsoft had.

And with OEMs shipping flaky, crapware-loaded PCs that took several minutes to boot, I don't expect power saving was Microsoft's biggest concern...

>> Couldn't they just hire someone and say "Fix this.

The article says that "if you take the latest and greatest 13" MacBook Air, and install Windows 8 on it", the battery life goes from 14:25 to 7:40.

It's unlikely a single fix. Probably everything from drivers to GUIs would have to be made more efficient to close the gap.

>Couldn't they just hire someone and say "Fix this. Identify the problem and we will support your findings to get the fix implemented.

Power conservation is more of a mantra, rather than a feature. If you decide that this is your priority, it should infuse every feature and every aspect of the OS. This can absolutely be done, and Microsoft is more than capable to cater to this need. I'd like to know why they haven't done so already. I mean, Intel finally got it with Haswell.

The infusion of Nokia engineers long familiar with engineering for battery life should help

I would presume with the Surface Pro Microsoft would be involved in, or at least work with hardware manufacturers to write the drivers. The integration is the same as Apple has on OS X. Microsoft is involved with the hardware, the OS and the app (IE), right?

Sort of but not quite. Tegra 4 that goes in the new Surface 2 - The AnandTech review states the power draw for Tegra 4 is not that impressive for example. So Microsoft would have to be involved in the design and manufacture of the SoC to cure that problem as just one example of how hard it can be.

> Windows and Linux are general purpose OSes. The same kernel runs on high throughput servers and low power tablets.

Since we haven't figured out free-energy yet, using less power on servers is just as important as on tablets - if not more, considering most servers sit idle.

Linux already has it sorted, at least on certain hardware. I have a 13-inch Retina MBP, and I get the same battery life with Linux and OSX on it.

Reply on news.ycombinator.com