Vista “Enable Advanced Performance” Benchmark

I was doing some upgrade work on my Vista machine last weekend. Once I was done popping in some more ram and updating all my drivers, I decided to check out some of the Vista performance tip guides available on the web. They all highlighted roughly the same features, but one that stood out was a little checkbox in the hard disk management properties that let you “further improve disk performance” at the expense of “increasing the risk of data loss if the disk loses power.” No problem, that’s what my battery backup device is for.

I started looking for benchmarks, and found none. My search string was pretty similar to the title of this post. That’s on purpose, because I want to help anyone else looking for a benchmark on this feature.

THIS FEATURE DOES NOTHING EXCEPT RE-INTRODUCE AN OLD WINDOWS BUG.

That’s right! There’s no performance increase available through this checkbox. It simply reintroduces an old bug that some old software relies on, so that said old software can regain the performance it lost when the bug was fixed. If you’re just an everyday average Joe like me trying to tweak your computer for the best performance, this checkbox isn’t going to help you. In fact, it’s going to hurt you, because if the disk loses power you could lose or corrupt your files.

You can read the full story, courtesy of Raymond Chen, at this link.
Windows Confidential: The Power of Bugs

Special thanks to the members of DriverHeaven for helping me find this article and the real answer to what this Vista feature is all about.

12 thoughts on “Vista “Enable Advanced Performance” Benchmark

  1. Chris

    Nice little blog! Was wondering what that was and googled it. Thanks for doing the research for me. Talk about a misleading “feature” description.

    Reply
  2. leloch

    I am afraid you are not right. This feature actually makes disk to operation in write-back mode – which means system is not waiting for data to be written from cache to disk.

    This gives substantial performance gain in write operation, but of course you don’t have any guarantee that your data really has been written to hdd when you PC is running.

    In this mode OS is syncing data when the disk is idle, but if power outage occurs before this happen – they are lost.

    Reply
  3. Daniel Vogel

    Hmmm, I just fiddled with the option and I saw a measurable difference. I’ll probably run some of the tests again later to be sure.

    The benchmark I used is compiling a binary with a custom build tool on a 8 core machine. We kick off 12 simultaneous tasks and the performance is very reliable.

    With it disabled I get 3:20 min and with it enabled I get 3:15 min for a full rebuild using the Visual Studio 2005 SP1 compiler invoked from the command line.

    Time to read the article to see why I might be seeing the placebo effect 😉

    Reply
  4. RobertD

    Oracle custom written stored procedure, mostly write operation, striped across four disks, using Oracle ASM.

    I would never risk data corruption in a real environment, but for testing, it helps to speed things up, especially as my test machine is rather lightweight.

    14 minute run, now 6 minute run. I reset the database and tried it again, 6 minute run.

    Seems like it does something to me.

    Reply
  5. Andrew Pellerano Post author

    http://technet.microsoft.com/en-us/magazine/2007.04.windowsconfidential.aspx

    Please read the source article. There are specific programs that will be sped up because they take advantage of the Windows bug this check box toggles.

    Maybe the internet’s opinion has changed in the year since I’ve written this article, but originally this check box was being touted as a “Vista tweak” by every performance enhancement guide around. My intent with this article was to reverse those feelings by showing people the truth behind the check box.

    If you think this feature is going to increase the average user’s average Windows speed, it isn’t.

    Reply
  6. Olive White

    Really useful informations, i make photographs and i do have to optimize my hdd reliability not to get corrupted data. I have performed a reliable configuration using several HDD with one for OS, one for data (2 striped hdd), one for backup (external e-sata). I do recommend this way to do instead of shadow copies.

    Reply
  7. maxregistrycleaner.com

    Funny! I can imagine opening a screen and seeing that description next to a check box. If only MS had the guts and sense of humor to actually do it, eh? I worked in software development and we used to come up with weird remarks in error messages. Always tricky in case you forget to replace them with the real message when releasing the software! 🙂

    Reply
  8. Winders

    Great post Andrew. This feature is evil.

    However, the Oracle guy (RobertD) is right. The feature DOES do something. It disables write cache flushing. This will especially show up on databases. In a benchmark, it would be reflected in a higher IOPS count.

    Daniel Vogel is probably also right, for the reason this feature is evil. Disabling all write flushing also means that when NTFS needs to make sure something is written to disk or else, it can’t. That can lead to corruption.

    If you need the speed, buy a RAID controller with on-board battery.

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *