Vista SP1 will deliver big network speed boost

Vista SP1 will deliver big network speed boost

Summary: I was prepared to wait till the public debut of Vista Service Pack 1 release candidate next week before writing about it. But after upgrading two machines here and doing some tests, I changed my mind. If Microsoft's decision to ditch the WGA kill switch in SP1 didn't convince you, would you be interested in tripling your network file transfer speeds?


[Update 6-Dec 14:30 PST: Corrected several errors in reporting percentage increases. Thanks to Jeff for pointing out the error in a Talkback comment.]  

I downloaded the release candidate of Vista Service Pack 1 yesterday and was prepared to wait till its public debut next week before writing about it. But after upgrading a few machines here and doing some tests, I changed my mind. If Microsoft's decision to ditch the WGA kill switch in SP1 didn't convince you, would you be interested in a 300% increase in tripling your network file transfer speeds?

Forget the reports you might have read about SP1 resulting in no performance boost. That story was based on a silly artificial benchmark involving scripting of Office applications. Back here in the real world, where gigabit network connections are now commonplace, you'll see at least one huge improvement when transferring files over network connections.

In its original release, Vista had some design problems with its networking stack, resulting in slow file transfers, especially when connecting to computers running Windows XP, Windows Server 2003, or Windows Home Server (all three of these products share a great deal of their code base, including core networking components). In Vista SP1, file transfer speeds are dramatically improved. In this post, I'll describe what I saw.

I did two sets of file-transfer tests using two separate systems configured to dual boot between Vista RTM and the new Vista SP1 release candidate. Both systems have dual- or quad-core processors (both in the Intel Core 2 Duo family) The first group of files consisted of two large DVD images in ISO format, totaling 4.2 GB. The second group of files was a folder filled with more than 3,000 files of all types, in 299 subfolders, totaling roughly 6.5 GB.

For the first test, I transferred the two groups of files from a shared folder on an HP MediaSmart Windows Home Server to the two test systems running Windows Vista RTM, recording the total transfer time for each one. Then I rebooted the two systems into an SP1 installation and repeated the test. I converted the times into throughput rates; here's the result (note that bigger bars equal higher throughput and thus better performance):

Throughput for file transfer from Windows Home Server

As you can see, the file transfers under Vista SP1 were dramatically faster than the Vista RTM times. For the directory full of many small files, the performance increase throughput was more than 300%; for the large files, the speed increase was roughly 260%. Note that you can expect similar results when transferring files from Vista to systems running Windows XP or Windows Server 2003.

For the second set of tests, I performed transfers between the two machines running equivalent versions of Windows Vista: RTM to RTM, SP1 to SP1. Here, the results were less dramatic. For the folder full of small files, the throughput rate increased by about 50% under SP1, and the large files transferred slightly slower, although still faster than the transfer from Windows Home Server.

File transfer between machines running Windows Vista

When I spoke with Microsoft about this phenomenon a few months ago, they explained that the issue was caused by a design change in Vista that eliminated the buffering used by XP and its siblings when transferring files over a network. Bypassing the cache read-aheads and deferred writes makes for better disk-to-disk performance and provides better control over how much data you're pushing over the network, but the mismatch slowed down transfer speeds in Vista RTM. That's been addressed effectively in SP1, as these results show.

Unfortunately, the other big Vista networking issue doesn't appear to be addressed in SP1. If you run an application that uses the Multimedia Class Scheduler (such as Windows Media Player), you'll continue to see a performance hit when transferring files over gigabit network connections. For more details, see this explanation from Microsoft's Mark Russinovich and earlier test results from ZDNet's Adrian Kingsley-Hughes.

But when I fired up Windows Media Player while a network file transfer was in operation I saw a sharp drop in throughput when the music began playing and then saw throughput pick back up when WMP was closed.

Even with the "release candidate" label, this is still a beta, so I can't recommend SP1 yet unless you're willing to assume the risks that come with beta software. But so far, the results I'm seeing are extremely encouraging.

Topics: Networking, Microsoft, Operating Systems, Software, Windows

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Other Performance Test

    Will you be publishing other test on any performance changes between RTM and SP1? The file transfer performance increase is a very encouraging sign that some one at Redmond is listening to us Vista users along with removing the kill switch.
    • What else would you like to see tested?

      Ed Bott
      • Here are a few things

        Some encoding test, simple gaming benchmarks and office productivity test. Just a few basic real world test to see if there is any difference in performance for every day task.
        • I forgot, Thank you for all your hard work (NT)

          • What is 'NT' supposed to mean?

            Forgive my ignorance here, but I see this quite a bit and thought I would figure it out from context by now, but I haven't.

            Thing is, I've only noticed it coming from snarky posters and flamers until now and assumed it was some kind of 'burn'.

            Wait a minute...let's utilize the power of Google:

            Narrower term: A term that is one level below this term in the hierarchy

            Have I found the answer? A shorthand way of saying 'my comment is only relevant to the post directly above it'?
            Your Mom 2.0
          • NT means no text. The reply is in the header.(nt)

          • No text

            The comment form won't allow you to submit a comment with no text, so convention is to use (nt) if your full comment is in the subject line.
            Ed Bott
          • You're supposed to put the nt in the subject

            line though so that people don't have to open the message. :)
          • Waht is NT supposed to mean?

            NT means NEW TECHNOLOGY. It does NOT mean No Text as some people are saying. This was a change in technology from Windows 3.11 and from Lan Manager
          • Context, its all about the context (nt)

          • Yeah, if you read this on a FAT32 system

            you can't see the reply text. Hence, the alternative text "nt". :)
          • 'No text' Here is your text VISTA vs. XP

            Comparisons of these two operating systems with the latest updates would be the most benificial. Even better yet do it using machines that the majority of people own. I only have 2 friends I know running a computer with the dual core processors and none with a quad. Plus of the ones who do have the 2 cores one only has 512mb of memory and the other 1gig. Which would make even them poor for Vista. The rest of my friends and other aquaintences are stuck with single core processors and usually a gig or less of memory. So using an average computer Vista won't run.
            Victory XP
            The new computers at work all came with Vista. I have noticed that with out the top of the line tech inside they are extremely slow. It takes them forever to do even simple tasks. The seriously bloated and resource tasking vista just demands to much. I wouldn't be suprised to find out that the computers, that my work just bought, are not up to the demands that Vista needs. Trouble is I have also been to many retail outlets and have seen that there are quite a few computers selling that don't meet this requirement.
            So a comparison with average computers that most consumers own running both would be benificial. Then use the average computer bought today and run both Vista and XP to see the effects. I was seriously looking forward to Vista but after the few experianes I have had with it. I scream all the way back to my XP. PLus the fancy new design is Ok but heck it only lasts until I do something. I don't see all that fancy work 90+% of the time. Heck once I go online or load a game or program no more fancy Vista.
          • 512 MB of memory

            There's your answer right there. 1GB of RAM costs $20. Those "seriously slow" machines would improve instantly with a $20 upgrade.

            Sounds like whoever buys your computers for work needs to be replaced. It's not like the system requirements for Vista are some sort of classified secret. It's been out for over a year, for heaven's sake.

            I've shown my test results on a five (nearly six) year old computer running Vista. It's impressively fast on most actions, as long as it has a least 1GB of memory. Now, if you throw a six-year-old Celeron at it, that might be a different story.
            Ed Bott
          • 512 MB?

            Why would someone who's spending 100-150 bucks on an OS going to worry about spending 20 bucks on ram?

            You can EASILY buy 2gb of ram for under $50.00. I believe I've actually seen it as low as $20.00 for 2GB!

            If that's too much for your friends to spend, then they can't afford the OS.

            I ran RC2 on an Athlon 64 3000 with 2GB and a then 2 year old GPU and it ran fine, so any semi modern Graphics card can handle it (if your friends bought a low end computer 3 or 4 years ago, then yeah, you'll have problems....welcome to the world of computers and software).

            I just don't get this comparing it with these so called average computers. The average computer user doesn't upgrade an OS. They buy new computers and get whatever the current version of Windows is.

            And your friends below average computers -- 512mb? My 70 year old moms 2.25 year old laptop has 1GB of ram -- are probably poor candidates for Vista. This is no different than win 9x machines that were poor candidates for XP (though for some reason everyone now seems to think XP was loved by all from day one).

            The tests aren't needed. If your friend's computer is over 3 years old and the GPU wasn't at least a 6600gt, he/she probably shouldn't upgrade. Their computer is probably used mostly for web browsing, and XP is perfectly capable of doing that.

            They should wait till they buy a new system and let the builder install Vista for them.
        • More details on office productivity?

          Most office productivity stuff to me is pretty much dictated by usability and not by any hardware or software speed limits. Can you suggest some areas where performance might come into play? I can definitely do some encoding and decoding tests. Might have to leave gaming for others, though. I am not a gamer!
          Ed Bott
          • Here is a link to Cnet's testing method for Microsoft Office


            This measures how fast some one can complete a number of task using Office.

            If this is too time consuming or beyond what you want to get into, I understand if you decline. Some one will eventually do all these test and post the results.
          • Not a question of complexity

            I'm skeptical of the benchmark for Office. All it's testing is macro execution speed, which can be affected by any number of things. The point of a macro is to automate a process that would normally have to be done manually. If hypothetically I have a 50-step process that would take me 15 minutes to do manually, I could do it in 1 minute with a macro, but more importantly, i can do it with one click. So if it takes one minute or 2 minutes, I really don't care that much; it's the automation that matters.

            This is my objection to the Devil's Mountain Software OfficeBench suite as well, which simply simulates a bunch of clicks in a non-real-world way.

            Anyway, I don't have their test macros so the question is moot. If you (or anyone reading this) can think of a test that will really simulate a real-world experience i'm game.
            Ed Bott
          • Good Point (NT)

      • XP!

        Can you run XP in the same enviornment, so we can compare the SP1 Vista results against XP SP2 results?

        • Vista SP1 vs XP SP2

          I agree...if Vista (either version) can't compete with XP, then there really isn't anything to worry about, is there?