Windows 8 and the 8 second boot... for real, but with a catch

Discussion in 'Windows 8' started by JeanYuhs, Sep 16, 2011.

  1. JeanYuhs

    JeanYuhs MDL Junior Member

    Feb 16, 2010
    Note: This is a long post, with details you may or not give two farts about, I'm just reporting on my personal and very repeatable findings in the past 36 hours with Windows 8. Do your own testing and see if you note the same things I have that prompted me to create this post.

    Ok, been working on this all night and most of Thursday to present my findings. I wanted to have some nice charts, graphs, etc, but I don't - I've just got numbers that you folks can use as the basis (I suppose) for doing your own testing. What this posting is about is something I noticed when installing and testing the Windows 8 Developer Preview since it was released (DOH!). That something is explained below, but for those of you ADD/ADHD sufferers where tl;dr is a way of life, it's this simple:

    Based on my testing, the 32 bit build of 8102 and the 64 bit build of 8102 are effectively the same - but that 64 bit "extra" version (the 4.8GB ISO) that includes Visual Studio Express 2011 and the code samples and other things isn't the same thing as the standalone 64 bit build of 8102.

    There's something different going on, and I don't know what it is but here's what I found in testing. My primary machine is a Dell Latitude D630 laptop with the following basic specs:

    - Core 2 Duo T7500 CPU @ 2.2 GHz w/4MB of L2 cache, 800 MHz FSB
    - 4GB of DDR2 667 RAM, running at 5-5-5-12 timings according to CPU-Z
    - 80GB Hitachi SATA II 7200 rpm laptop drive (manufactured in October of 2008, factory drive with this laptop, no damage/bad sectors, 62MB/s max sustained read, 49.5MB/s average from start to finish)
    - 14.1" 1440x900 LCD panel
    - Intel GMA X3100 GPU (Windows 8 does not have drivers for this natively, nor does Windows Update provide them so I use the last version from their support site, posted in October of 2009 but for this testing they are NOT involved, read further for details)
    - Broadcom Gigabit NIC and 802.11abg wireless (also not really part of this)

    A solid working laptop that I have had for some time now and it's built like a tank. Now on to the meat and potatoes:

    This is about the boot times, cold boot (complete powered off state, not standby, not hibernation) and warm boots (standard reboot). What I've discovered is that the 32 bit and 64 bit standalone builds are slow, really slow, and with not much explaining it, while the 64 bit version that comes with the extra tools/VSE2011/etc will boot extremely fast on this machine.

    Here's the process:

    1) Install the OS for testing from a USB stick (I don't use actual CDs or DVDs much any more for such purposes)
    2) When the installation is completely done and at the first actual boot, remove the USB stick, shut down the laptop, then image the installation at that point - no updates installed (those are disabled by the options I choose - the laptop is not connected to the Internet at this point, not by the NIC or the wireless)
    3) Once the image is done, reboot back into the OS, install the appropriate build of Microsoft Security Essentials (32 bit or 64 bit) locally off another USB stick I have holding those sorts of tools for quick access
    4) Once MSE is properly installed, go to the Windows Update options in Windows 8 and make sure it's set to check for updates but not install them (for the entire OS). Connect the NIC (hardline to my 50Mbps cable modem connection) and update MSE ONLY - that's the only update that I get at that point, just MSE and not the few available updates for Windows 8 itself
    5) When MSE is "green" and updated, reboot the laptop 10 times in a row and measure (with a stopwatch in my hand, so don't complain this isn't perfect testing) the time where POST ends and Windows 8 actually starts loading to the perceived moment when the lock screen appears - that's the target: seeing the lock screen. When that appears, stop the stopwatch and record the time.

    When I'm done, I install the next OS, follow the same procedure, then do the last one since there are 3 versions of Windows 8 build 8102: 32 bit standalone (2.8GB ISO), 64 bit standalone (3.6GB ISO), and the 64 bit version with the extras/VSE2011/samples/etc (4.8GB ISO). I did this testing in that order, also.

    Now for the pretty shocking results:

    32 bit standalone:
    - Slowest time of 47.2 seconds and a fastest time of 25.7 seconds
    - Average of the 10 boots: 28.3 seconds from POST handoff to lock screen appearance

    64 bit standalone:
    - Slowest time of 43.1 seconds and a fastest time of 24.3 seconds
    - Average of the 10 boots: 27.5 seconds from POST handoff to lock screen appearance

    About what I expected since they're basically going to perform the same in terms of booting. But...

    64 bit extras/VSE2011/samples/etc:
    - Slowest time of 17.4 seconds and a fastest time of 7.1 seconds
    - Average of the 10 boots: 8.3 seconds from POST handoff to lock screen appearance

    I still don't believe it, but it's repeatable, practically every single time I do it. Cold boot, totally powered off, even with the hard drive pulled out and the battery, then put back in, hit power, POST takes all of 3 seconds at best, then the handoff and effectively ~8 seconds later I'm staring at the lock screen.

    Now, the issue here is this is the base OS starting up with Microsoft Security Essentials - you'll remember (if you're paying attention) that I did mention earlier that Windows 8 doesn't have drivers for my video chip, the GMA X3100, so these incredibly fast boot times to the lock screen are not accounting for me installing the video drivers which, as expected, hurt performance considerably:

    On each of the OSes, once the Intel GMA X3100 drivers are installed, there is an additional 10-14 seconds added to those times, on average. The incredibly fast 64 bit extras/VSE2011/samples/etc build hurts the worst because it basically doubles the time to the lock screen, sadly. I even went in and removed (not disabled) the 3 entries for the Intel hotkey/persistence/tray processes that start up, but it doesn't help: that only gives back about 1-2 seconds off the boot speed.

    This might not seem like such a big deal but, considering this is a 4 year old laptop with a 3 year old slow hard drive (current 7200 rpm laptop drives can push out 125MB/s sustained at the beginning of the drive, easily twice as fast as this one), and SSDs... good lord, if I had an SSD I would suspect I'd boot in about 4 seconds flat, or less. :D

    Anyway, I just wanted to share the info and my experience. Try it yourself, folks. Probably won't notice much if you're doing your testing in a VM with VMWare or VirtualBox - I find such testing to be a waste of time myself. All my installs are on the bare metal hardware, always.

    But just for the record: this old workhorse laptop with the slow spinning platters did hit the lock screen once in 7.1 seconds... approximately.

    I don't know why there's a difference in the 64 bit extras/VSE2011/samples/etc version, but the boot time isn't the only difference I've noted. There are some other small ones but that's not what this posting is about: this is about speed, and the standalone versions don't seem to be anywhere near this fast for booting - and I've noted the non-standalone version is somewhat snappier and more responsive overall too. IE10 in the standalone builds can do Sunspider in about 340 ms; the non-standalone version gets it done in about 317 ms, and that's consistently in several runs in a row. Not a lot but, nearly 10% difference.

    There's something special about that non-standalone version. You folks test it out yourself...

    Pretty amazing stuff, this Windows 8. :D
  2. Espionage724

    Espionage724 MDL Addicted

    Nov 7, 2009
    So... TLDR: WDB standard and SDK x64 have different boot times, with the SDK having faster boot?
  3. gamer765

    gamer765 MDL Novice

    Oct 24, 2009
    tl;dr. it's just you