Windows 8 to boot in 8 seconds Microsoft is touting very fast boot times for Windows 8, thanks to the clever trick of writing the kernel state to disk at shutdown. Rather than write the whole contents of memory to Windows' hibernation file, Windows 8 just writes enough to be able to put the state of driver, services and such back into memory, ready to run, …
They've been optimizing boot performance for years - as with many of Windows' deficiencies it's primarily the result of poorly optimised drivers:
http://jvert.wordpress.com/2007/02/20/fast-resume-6-years-later/
By hibernating the kernel Windows 8 should be able to sidestep this problem with slow-loading drivers - quite a neat solution.
Then it's irritating marketing rather than a really good technical effort.
I don't consider the time of booting to be how long it takes the kernel to reinstate itself, but the total time between power on and usable desktop (excluding the time it takes me to enter my user name and password).
Considering that it seems to be the post-kernel boot process that's the real killer, that's where the effort needs to be applied.
This post has been deleted by its author
you got to the point where you had a command line from which you could write a 2 line goto loop fairly quickly. And strictly speaking that is the point and yes you're both correct.
However, the example I was trying to make (badly) is that modern, rival Os's often boot up to a usable state quicker than MS products and if that wasn't the case then MS wouldn't be striving to make their boot time quicker...
I remember ordering a ZX81 when I was a young ignorant programmer (and now of course, I'm an old ignorant programmer) but the delivery time was going to be phenomenal, which cancelled out the OS loading time, on average. And I too got a BBC Micro, which in those days was easy to learn inside out.
Then I sold it, and I so regretted selling it (the BBC).
With a decent SSD a full boot of windows 7 from the end-of-bios to the login prompt is not exactly a time consuming operation. The major issue is the crud which occurs after you log in (which is not helped here).
I think I'd rather have the full boot every time so that at least I know my current session is not contaminated with some undiagnosable lurgy inherited from a previous session.
i wasn't aware apple had dragged the rest of us down to their level.
seriously though, I have a brand spanking new imac at work and it still takes 5 minutes to warm up enough to give me a useable desktop in the morning - which I'm managing to overlook only because it turns out you can change the system options to put the control key where it *should* be.
some people seem to think they are somehow 'better' but tbh all I see is 'different'.
would you trust a Windows kernel session to live longer than a week anyway?
One thing I love about having a Windows VM is that I can pause execution. It's the only way to keep the system usable and mitigate the need for relentless rebooting.
I suppose saving the session is a similar concept, but having the benefit of a robust host OS and only minor need for Windows applications means that a typical session of mine only lasts half an hour a week.
And it still manages to spend most of that time fragmenting the disk and then pissing around trying to defragment it.
... if you can't keep a WinTel box running more than a week.
My boxes get rebooted on the monthly patch schedule and that's it. Our 2008R2 Core servers (no GUI) haven't been rebooted for nearly 6 months as there's sod all to patch that's actually in use / a service restart won't fix.
Just decomissioned a box that is out of scope of our patch runs that on 2008.... hasn't been booted for 2 years
So yeah - I'd trust the NT Kernel session to live for way more than a week. My Win7 laptop only gets a look in every couple of months unless there's a patch that needs to be urgently applied.
Correct me if I'm wrong but my XP laptop has been doing better than this for a LONG time. It's basically called hibernation or standby - this isn't improving BOOT times at all - it's hibernating. And hibernating with a MUCH faster disk that I'll ever put into a laptop to make it look fast. And all the problems and software-cooperation that comes with that, too (hope all your legacy drivers are perfect and now how to hibernate properly!).
Now my XP workstation can avoid BIOS boot, so long as I keep some very low battery power supplied, for 24 hours of more (it's called "bog standard standby"). Resume is pretty much instantaneous. All this is is an improvement on hibernation (where you write the standby memory to a disk instead). Granted that the BIOS would pop up but any half-decent BIOS can easily be written to do such a thing quickly - it's just the same as something like Coreboot speeding up the hard-disk into its fastest modes and THEN reading a file from disk and resuming from its state. It's not anything particularly clever, innovative, or new.
Computers have been doing this for literally DECADES and a one-off show on pre-chosen hardware is really nothing to crow about. In fact, in that case, 8 FECKING SECONDS?! That's ludicrously slow. You could have diddled the BIOS into being a "Windows 8 compatible" one and made it near-instantaneous with an SSD (which Windows is increasingly being designed towards so would barely raise an eyebrow).
Additionally - we have the age old problems with boot-time claims:
1) Nobody boots. Laptop users don't HAVE a boot time, only a suspend/resume time. Full boots are for when things go wrong.
2) Those who do boot don't notice the time compared to anything else (e.g. application load time, etc.)
3) Those who do boot and take ages in BIOS (i.e. servers) do so for a reason - stability, testing and predictability (not to mention that they only full-boot once a year, if that).
So smartphones/laptops (hell, even my old Palm) already have it. 99% of servers wouldn't use it (not much point in a server being in hibernation - either it's on or not). The rest of the market don't really care about boot time anyway.
Don't get me wrong, the tech is wonderful. It was back when APM was first invented too, and even before that. But claiming that Windows 8 will be doing anything "special" as regards boot-time is ludicrous. If this is the first selling-point of Windows 8, that's a warning to people like me who have to decide whether or not to buy hundreds of units of it.
Yes Hibernation has been around for a long time, but the problem these days is that as powerful computers get more and more memory in them that means more and more data that needs dumping to disk every time. Hiberating an 8GB computer takes quite a long time.
OK, so this isn't revolutionary, but to my mind it is smart evolution. If you can't fix all the problems in one hit, at least fix some of the little things that you can control.
That was my first thought upon reading the main article: if one is going to claim "boot in X seconds", you have to define when "booting" is finished.
For instance, WinXP tries to give the impression that it "boots" quickly, by loading the desktop as soon as it can. However, you then have to wait a couple of minutes whilst all the remaining services load up, as the machine is just too slow to use until they've finished. (That's not a fanboi point: Mac OS X isn't much different in my subjective experience.)
Of course, a fast boot also depends on the specific services you're loading up. My Eee 701SD netbook (running Arch Linux) can boot to the login prompt in about ten seconds, but then I've tried to cut out the fatty bits (GNOME, KDE, etc.) and run Fluxbox, which helps speed things up.
So, maybe MS' claim is accurate, but I'd want to find out how the goalposts are spaced...
Boot times of brand new, empty PCs were never the issue. The problem starts when you install a few things. A Windows PC takes longer to start every day, until users dread turing the thing on. Many non-expert users just assume their PC is slow becuase it is old. They buy a new one and are delighted how fast it boots. Some of them use a clean-up service with similar effect.
Windows' long boot time is largely down to the loading of an ever lengthening list of "zombie craplets", and the necessaity to have malware protection in the background. The OS isn't stable enough to undergo many hibernate/wake up sessions, it will always need rebooting now and again, so long boot up times will remain a feature even with this clever mod.
Yep, the key to improving windows boot times, is to stop allowing applications to install loads of crap that runs at boot, especially without your explicit permission.
I really don't want my pc to take 20-30 seconds longer to boot, just so that office, or IE can be loaded then, and kept in memory and appear instantly when i click the icon. Not to mention the sheer volume of crud that iTunes installs that boots with windows.
I've managed to get it down to a fine art: "press power button, boil kettle, while boiling, enter password, return to tearoom, set tea steeping, and the return to machine", which by this point is all ready to go.
Although speaking of instant, you should smell (or heaven forbid, taste) the coffee we're given. Yuck!
It sounds like some of you would be much, much happier if Windows 8 took longer to boot.
At least then you'd have some valid reasons for criticisms rather than this usual tripe..
If they add features it's anti-trust baiting bloatware
If they take them away then it's a rip off
If they have different versions then it's "confusing"
If they don't have different versions, then it's a rip off paying for unused features..
It's too different from the previous version
It's too similar to the previous version
This doesn't just go for Microsoft, but in general critics don't mind if other critics have entirely contradictory reasons for complaining... so long as they're complaining.
Yes it remains to be seen (which makes the whole business of prejudgements silly), however it seems apparent they even if they announce features that would be welcomed on any other system, some will find a reason to complain.
And complain, not that it'll be bad, but shock horror, it might be good, possibly successful and pervasive OS on everything from desktops to tablets. That would be awful, if M$ ever bring out a decent O/S you guys will have to find something else to bitch about.
Your mum's cooking perhaps?
"That would be awful, if M$ ever bring out a decent O/S you guys will have to find something else to bitch about."
Any of us who have been reading El Reg for some time have noticed that a certain section of the commentariat *will* regard it as a disaster if Win8 is any good and the trans-platform strategy is competently implemented.
@AC - Bang on.
It never ceases to amaze me the amount of things that IT guys manage to whinge about. Often it boils down to change, which is ironic for such a fast moving field.
"I've learned XYZ OS and can't be arsed to learn another, therefore I'll just bitch about the others in the hope that I can keep going with my single OS and keep my cushy job." Seems to be a very common reason for bitching other OSes. This goes for a certain group of all OS admins/devs: zOS, iOS(not that one!), UNIXes, Linuxes, Windows, Mac OS, all have these people and they're all tedious.
Still, one thing that I've learned in 12 years of working in IT is that you don't get anywhere by whinging all the time. I know people still on the helpdesk at my first company, whinging about the fact that they've not been promoted, but not actually doing anything about it. There are other people who started at the same time now in senior management, because they put themselves forward, embraced new stuff and made things better rather than just complaining about their lot.
Back in the olden days, my Windows System booted up from DOS to Windows in 2 seconds.
Then most linux-based X-Servers boot up in that timeframe, even when running on a Pentium 90.
The big problem is that this time doesn't reflect the true boot time. Those 8 seconds won't give you the services you usually need, like an SQL-Server or a webserver.
Around 1999 to 2000 I worked for the local council installing PCs and servers into council offices, schools and that sort of thing. Because Councils never permit anything new until they have tested it to death we were installing Windows 3.11 onto Desktops with a PII processor running at something like 450-500MHz
Boot time from cold to the logon screen was about as long as it would take you to read the following sentence out loud:
post, post, post, post, doswindows
As long as you read the last bit really quickly.
Your PC wasn't designed to run 24/7, the hard disk probably has a duty cycle of about 8 hours at a time. This doesn't mean to say that it won't run all day, but you'll reduce it's lifespan.
Having said that, I leave my Fedora netbook booted all the time and just close the lid when I'm finished with it, it does run on an SSD though. Last reboot was after a month of "uptime" becuase I accidentally let the battery run out.
Sounds like they are adding complexity at shutdown time, and whenever a real shutdown is needed, which I'm guessing will be a familiar scenario. I've seen some Windows systems take longer to shutdown than boot already...
What users would really appreciate is a system so stable it can be hibernated or suspended endlessly, apps and all, instead of being shut down. Some non-Microsoft systems already do this :)
You must have a really fast macbook air then, my wifes one isn't that fast. It is fast, but not 8 seconds fast. I have just put a new crucial 256GB SSD in my macbook pro, so now it is 28 seconds from power to useable desktop with apps and documents loaded (down from 67 seconds on a monumentus XT 500GB). I am quite happy with that, although I don't tend to restart.
However, I think it is good that people are trying to improve. My windows 7 machine only takes a few seconds to logon screen, but from start it takes a few minutes before it is in the state that you could call usable. I don't quite understand the linux / mac comments anyway - windows starting faster doesn't take anything away from you.
After all - it may take 2 minutes to get to a usable desktop in windows, but in linux, it has been 30 years and we are still waiting :) <runs>
Or they should be, but I don't see much evidence.
PC architecture dictates these ridiculous, non-threaded ways to initialise hardware. We may have progressed from ISA through PCI, PCI-x and (now) PCIe, but the model has remained the same for a long time.
Devices are dumb and rely on drivers to make them do anything. This is evident so clearly when your OS first starts the long crawl to usability, after power-on. Those lovely low-res graphics reminding you which OS you have installed. No greater reminder is there that we are relying on ridiculously outdated technology.
Where is the standard graphics interface? It shouldn't be difficult, what with all the muscle your average GPU carries (more than most PC's, if stories are to be believed). Windows should be interfacing with something resembling a fast framebuffer, but with obvious additional capabilities. Instead, Windows interfaces with a driver, which abstracts all of the fast stuff behind a lot of code. Sometimes the code doesn't do much and just passes and translates API calls, sometimes it does a *lot* of work to fool Windows into thinking it is dealing with native capabilities.
Until we morph the 'driver' model into something else, like a smart device with a teensy bit of glue code, those boot times are only ever going to get longer.
Mashing kernel images into a compressed file is just putting lipstick on a pig.
The CPU does NOT access a frame-buffer for rendering. The CPU might access textures to write to using aperture memory. After that it uses vertices, vertex shaders and pixel shaders (assuming a low-endish GPU) to cause a completely different co-processor to render the result to the screen.
The driver is responsible for handling the fact that all of these GPUs are incredibly different. They run different ISAs on their streaming units (basically a SIMD instruction set). You supply an instruction set that is JIT compiled to the correct architecture by the driver. Some GPUs run video on a general compute framework, other GPUs run video on very specific piexe of silicon. Lets not even talk about the fact that the memory ordering of all of the texture memory is specific to the GPU too to optimize cache locality.
In an ideal world all of this crazy capability would be standardized into an ISA and you could abandon the driver, but we are nowhere near having enough conformity in the hardware to allow this. But if you look at what AMD is implying by their roadmaps, seems to me like we'll get there.
This post has been deleted by its author
As soon as you turn on the damn PC, every installed program (including the OS) piles onto the Internet looking for updates. If they could simply block all the 'check for updates' app panic for the first twenty minutes after boot, then maybe I could actually get to the emergency situation information I require before it's too late.
In summary: Let the Meat Machine use the Internet FIRST, please and thank you.
Damn, I can't believe that it's 2011 and we still have ask for this basic, common sense sh!t. On most versions of Windows, even moving the cursor from the Start Button directly towards the Control Panel pops up the All Programs list, blocking the desired destination. Sigh...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
