Coffeehouse Thread

27 posts

Test shows Windows 7 is 5x slower than XP in UI and apps load slower than in Vista

Back to Forum: Coffeehouse
  • User profile image
    androidi

    UI operations compared: http://www.youtube.com/watch?v=ay-gqx18UTM

    I did the services.msc resize test on latest NVIDIA driver on 8800 GT and it's similar in result. Slightly puzzling is that the YT tester uses also 8800 and gets better result on XP. To my knowledge the dedicated 2D acceleration chip is no longer in the 8800 series and with older GPU it would be much faster in XP still.

     

    Also I had a friend who plays games on a 8800+Q6600 upgraded to 7 from Vista month ago and even though his Vista was insanely full of crap and slow, he says in 7 the games load different levels slower. I told that I experienced similar result in early 7 testing and assume it's because of changes to SuperFetch.

     

    edit: delete some ranting about SuperFetch, I'll really need some hard data first. SF is really problematic to measure since it's supposed to "train itself" over time to know what the user wants. So lets wait a month (he only used 7 for month now) and if my friend STILL says after 2 months that 7 is slower than Vista then I'll come back with the rants.

     

    PS. When I used Vista I didn't feel any need whatsoever for having SSD. In 7 I feel like I need to buy one SOON! What changed? Is 7 a flop? Did SSD industry conspire with MS to tone down SuperFetch so that SSD's would fly off the shelves? 

  • User profile image
    Sven Groot

    You experiences are exactly contrary to my own. I never thought Vista was particularly slow, but 7 felt just way faster from the moment I installed the first beta. Of course, that's just anecdotal evidence, so make of it what you want.

  • User profile image
    androidi

    Yes it "feels faster" I agree with that, however 1) that UI test shows its slower 2) while it generally feels and no doubt is faster than Vista, in particular application load times tell another story - Sysinternals tools and listening the HDD seeks tells that it is working much more during application use in 7. In Vista HDD was working after boot and after exiting applications because SF was re-filling memory with previously loaded stuff. I claim they don't load as much app data in 7 and that obviously can have dramatic effect when apps load hundreds of megabytes or even gigs of data. RAM speed will be in effect in Vista often while in 7 more often you get the disk speed.

    Looking with Sysinternals tools my observation is that 7 SuperFetch caches mostly just small files. If apps load lot of large data Vista can be .. in one case by my estimation even 10 times faster due to loading all game data from RAM.

  • User profile image
    Sven Groot

    Yeah, SuperFetch was changed because many users complained about the excessive disk thrashing caused by Vista, and because it was found it would cache useless large files in some situations.

     

    Many people don't feel Windows is "ready for use" until the disk stops working after login, which in Vista could take many many minutes. It's more a perceptual thing than anything else, and trying to explain what it's doing and why it's not a bad thing to an average Joe user is not really feasible.

  • User profile image
    vesuvius

    I am running a pretty big application on Win 7 at the mo, and it complies and runs without lag on a little laptop.

     

    I am sure if I had XP and Visual Studio 2003 running on the same machine they would be blazing fast because they have fewer moving parts (or parts in general). I don't think we can get to the stage where a 1970s Golf GTI weighs half as much as it's modern day equivalent, and the modern day car is faster  and better all round. The older version will always take corners quicker, because, lets face it, it weights half as much.

     

    Not really sure if it is worth comparing the physical with the virtual, but I think you get my gist!

  • User profile image
    DCMonkey

    I'm not suprised he found UI performance issues in services.msc. MMC and its plugins have some of the crappiest UI in Windows, in both form and function. He should have used the Event viewer as an example if he wanter to see flicker.

  • User profile image
    W3bbo

    , DCMonkey wrote

    I'm not suprised he found UI performance issues in services.msc. MMC and its plugins have some of the crappiest UI in Windows, in both form and function. He should have used the Event viewer as an example if he wanter to see flicker.

    On my Windows 7 laptop, I do experience the black DWM artifacts, but that's kinda inherent in the DWM (but isn't anywhere near as bad as in that video), but absolutely no choppiness when it comes to 2D operations (but I can feel a bit of latency in some things).

    But yeah, the Event Viewer, Task Scheduler, Server Manager, and Active Directory Administrative Center MMC add-ins are the worst. There's really no excuse for their slow, laggy, UI-buggy user experience. The IIS MMC is managed code and I get a great smooth experience with it (although I would change a few part of how the UI is arranged, but that's irrelevant).

  • User profile image
    Minh

    So.... the test used:

    * Win7 RTM
    * On 1 hardware configuration
    * Unknown test software
    * On 1 app

    Anybody else has a problem w/ this? 

  • User profile image
    CyberPunk​News

     I've done side by side tests between os's on the same hardware and only once I thought to compare xp vs win 7, and results were the same, xp ran circles around 7, and vista, while vista ran 20% more resources than win 7 while idle.

  • User profile image
    AndyC

    , W3bbo wrote

    *snip*

    But yeah, the Event Viewer, Task Scheduler, Server Manager, and Active Directory Administrative Center MMC add-ins are the worst.

     

    In the 'real world' people don't sit around constantly resizing the Server Manager UI, so it's not really a big issue. Any lag is usually because you're connected to a remote machine rather than fraction of a second differences in drawing performance.

     

    I'm not entirely sure the video is comparing like for like either, if I'm not mistaken XP SP3 still had the older MMC didn't it? And MMC 3 always was a more flickery beast than MMC 2.

     

    All in all though, Windows 7 feels faster and more performant and that is what actually matters. Not whether some micro benchmark somewhere implies that a combobox takes a couple of nanoseconds longer to draw.

  • User profile image
    W3bbo

    In the 'real world' people don't sit around constantly resizing the Server Manager UI, so it's not really a big issue. Any lag is usually because you're connected to a remote machine rather than fraction of a second differences in drawing performance.

    That's not what I'm talking about.

    The Server Manager on Windows Server 2008 R2 takes....15.4 seconds (I timed it) from when I click the icon to load it and before it becomes responsive. This is because whilst it's "Collecting Data" the tree view on the left is unresponsive.

    The Event Viewer on 2008R2 takes 7.9 seconds before its usable, again because whilst the main pane is loading it blocks the navigation tree.

    Let's not overlook the fact it takes it 4.4 seconds (over 7 seconds on the first try) to sort 5018 events in the System log by Date/Time. On my XP machine with 10,304 events it sorts it by Date in just under 2 seconds.

    Both of these tests were done locally.

  • User profile image
    AndyC

    , W3bbo wrote

    *snip*

    Both of these tests were done locally.

    And in the real world, they rarely are. Having a more effective tool, which things like the new Event Manager and Server Manager certainly are is less important than a few seconds of raw performance. You'll only ever notice that locally in practice, wheras the tools see far more usage in a remote management scenario, at which point network latency is a bigger deal.

  • User profile image
    felix9

    it looks like you are testing different code on different os, its hard to tell whose fault it is.

  • User profile image
    BitFlipper

    I didn't follow the link because this test sounds too stoopid to me (hey, let's create some more MS FUD) so it isn't worth wasting my time on, but wouldn't a more valid test be to run XP on an XP era machine and Win 7 on a Win 7 era machine, to see what the real world experience between the two would be?

    Windows 7 is way more complex that XP, but the important point is (which is completely ignored if you run the test on the same hardware), is how is today's Windows 7 experience compared to what it was like when XP was the latest and greatest?

  • User profile image
    elmer

    , BitFlipper wrote

    I didn't follow the link because this test sounds too stoopid to me (hey, let's create some more MS FUD)

     

    <sarcasm>

    Micro$ucks Windoze-7 runs like a dog on my Pentium-III 650 with 512MB ram and 10GB hdd... I've tested at least once, so M$ programmers must all be idiots.

    </sarcasm>

  • User profile image
    W3bbo

    , BitFlipper wrote

    I didn't follow the link because this test sounds too stoopid to me (hey, let's create some more MS FUD) so it isn't worth wasting my time on, but wouldn't a more valid test be to run XP on an XP era machine and Win 7 on a Win 7 era machine, to see what the real world experience between the two would be?

    Windows 7 is way more complex that XP, but the important point is (which is completely ignored if you run the test on the same hardware), is how is today's Windows 7 experience compared to what it was like when XP was the latest and greatest?

    I disagree with the premise. You're implying that it's acceptable for programs to become slow and bloated as long as they stay under Moore's curve. That a simple text editor today should take 5 seconds to open because that's how long it took to load from tape in the mid-1980s.

    The whole point of upgrading hardware is to allow older programs to run faster. New programs should take advantage of faster hardware, yes, but should do so in accordance with any added functionality. In this case the Event Viewer in Windows 7 is considerably slower than in XP yet adds very few features that would seem like they should slow it down so much. It's even regressive in that respect: performing the same operations within the program as in XP are slower on Windows 7 (such as sorting and filtering an event log).

  • User profile image
    JoshRoss

    <rant>

    IE9 seems fairly snappy in Windows 7. It doesn't seem worth the time to investigate
    the possible advantages that exist in an alternative universe where the OS was
    XP. The windows might do this, or that, faster in XP, but that is the extent of
    what is advantageous in XP.

    Windows XP is a virus honey pot. I’m not sure if I can prove it, but using it also seems
    to make you stupider or at least more willing to bicker about crap that doesn’t
    matter. I bet you that MSDOS 5.0 can do a directory listing 10x faster that XP.
    But you know what? It doesn’t matter! It would be more productive for people to
    polish coprolites, than to benchmark the compare the crappy bug ridden software
    of yesterday.

    </rant>

  • User profile image
    JoshRoss

    @JoshRoss: Speaking of coprolites, did any one else see the MythBusters episode where they polished turds? At least that was informative.

    -Josh

Comments closed

Comments have been closed since this content was published more than 30 days ago, but if you'd like to continue the conversation, please create a new thread in our Forums, or Contact Us and let us know.