Latest News Posts

Social
Latest Forum Posts

Windows Vista System Performance Reports
Bookmark and Share

vista_performance_0606_logo.gif
Print
by Rob Williams on January 25, 2007 in Windows

Does performance suck on Vista when compared to XP? That’s what I was set out to find out. I was worried at first, since the performance in Beta 2 was quite bad. While there is indeed a performance decrease, it’s quite minimal as you’ll find out.


Last summer, I wrote an article entitled “Windows Vista Beta 2 Performance Reports” which attracted a lot of attention. In fact, it proved to be our second most popular article of 2006, so it’s clear that people who are looking to purchase Vista truly care about whether or not they will lose a good deal of performance when making the move.

Well, that article was more for fun since it was indeed based on a beta operating system. I noted there that when the RTM became available, I would revisit all of the benchmarks mentioned there and come up with a more appropriate conclusion. I received a lot of criticism for that article, but most people didn’t seem to see the “Beta 2″ in the title. I was completely aware that the performance seen there did not reflect the actual performance differences in the final release.

I will admit one thing, the RTM is -far- superior to Beta 2… there is no comparison. I have been using Vista Ultimate RTM full time on my Windows rig for over two months and haven’t run into any serious problems, except for a few of my favorite applications that still do not function. I’ve found the system to be much more stable overall and speedier as well. It still has lingering issues, but I mention most of those in my rant article from this past Tuesday.

Back to performance. This article will not go into depth like the other one did. In fact, it will be split into two articles. There is a gaming performance article on the way, but we are still awaiting final drivers from NVIDIA before proceeding with it.

Testing Overview & Methodology

For our performance testing today, I will be using my Intel dual-core rig. The goal of the article is to show the direct performance decrease you will experience when making the move. Take note though, that the difference on your system may vary from my results, especially if you are using an older machine. I chose to use a “midstream” PC to benchmark with, as most people reading this may have a similar setup. If you use a computer that’s more than 2 years old, with a small GPU or one that has only 1GB of ram, you can expect Vista to not run that seamlessly.

I will be comparing the performance from a freshly formatted machine with both an XP Professional and Vista Ultimate installation. I will be using a variety of synthetic benchmarks and also real world benchmarks that touch on multi-media tasks. All tests are performed on machines that are -not- altered for the best performance. Normally when we benchmark memory or a CPU, we disable many services and close needless programs before doing so. This will not be done here, as realistically, this is not what people do prior to crunching a video file or playing a game. No post-install third party applications will be left running though… just default system services.

    Testing System

  • Intel E6300 Dual Core @ 1.866GHz
  • ASUS P5N-E SLI 0307 BIOS
  • Corsair 2GB PC2-9136 @ DDR2-1066 4-4-4-12
  • ASUS EN7900GT 256MB
  • Seagate Barracuda 7200.10 320GB
  • Corsair HX620W Power Supply
  • Corsair Nautilus 500 Water Cooling
  • LG 18x DVD-Burner
  • D-Link WiFi Card
  • NZXT Apollo Case

Prior to benchmarking, the system is “cleaned” up. Needless scattered files on the desktop are removed, and post-install applications are closed down if they are running in the background.

Throughout all of these charts, you will see many deltas. These symbolize the advantage or disadvantage Vista Ultimate has over XP.

SANDRA 2007

First up is SANDRA, one of my favorite benchmarks.

    XP Professional

      Sandra Arithmetic

    • MIPS: 17116
    • MFLOPS: 11906
      Sandra Multi-Media

    • Int: 102632
    • Float: 55427
      Sandra File Systems

    • Avg. Read: 63MB/s
    • Latency: 8ms
      Sandra Memory

    • Int: 5773
    • Float: 5775
    Vista Ultimate

      Sandra Arithmetic

    • MIPS: 17037
    • MFLOPS: 11905
      Sandra Multi-Media

    • Int: 102281
    • Float: 55338
      Sandra File Systems

    • Avg. Read: 61MB/s
    • Latency: 7ms
      Sandra Memory

    • Int: 5782
    • Float: 5717
    .

      Delta
      - 0.46%
      - 0.008%
      Delta
      - 0.34%
      - 0.16%
      Delta
      - 3.28%
      + 14.8%
      Delta
      + 0.16%
      - 0.66%

There’s very little difference here so far… less than 1% all around.

Cinebench, Sciencemark, 3D Mark 06

The same goes for here with Cinebench, but I started to see even larger differences with Sciencemark where the delta was closer to 3.5%. Super Pi also proved 19 seconds slower under Vista.

    XP Professional

      Cinebench 2003

    • Single Thread: 312
    • Multi-Thread: 581
      Sciencemark 2 (Lower is Better)

    • Mol Dyn: 83.72s
    • Cipher: 15.39s
      3D Mark 06

    • 4907
      Super Pi

    • 32 Million: 25m 0s
    Vista Ultimate

      Cinebench 2003

    • Single Thread: 312
    • Multi-Thread: 579
      Sciencemark 2 (Lower is Better)

    • Mol Dyn: 86.58s
    • Cipher: 16.06s
      3D Mark 06

    • 4693
      Super Pi

    • 32 Million: 25m 19s
    .

      Delta
      - 0.00%
      - 0.34%
      Delta
      - 3.41%
      - 4.35%
      Delta
      - 4.56%
      Delta
      - 1.27%

The 3D Mark score is about 300 points lower on Vista, but please don’t take that result too seriously at this point. NVIDIA is working around the clock on final Vista drivers, so the performance may improve when they are finally out the door.

Hard Drive Benchmarks

Overall with the disk access benchmarks, the only tests that had a rather large differences was with the burst and minimum speed tests. In both HD Tach and HD Tune, the Avg. MB/s proved no different at all, which is good to see.

    XP Professional

      HD Tach

    • Burst Read: 134.5MB/s
    • Avg. Read: 66.2MB/s
    • Latency: 13.4ms
      HD Tune

    • Min: 30.6MB/s
    • Avg: 62.3MB/s
    • Max: 76.3MB/s
    • Latency: 13.4ms
    • CPU Usage: 2.6%
    Vista Ultimate

      HD Tach

    • Burst Read: 120.7MB/s
    • Avg. Read: 66.2MB/s
    • Latency: 13.4ms
      HD Tune

    • Min: 36.6MB/s
    • Avg: 62.3MB/s
    • Max: 76.5MB/s
    • Latency: 13.5ms
    • CPU Usage: 7.7%
    .

      Delta
      - 11.43%
      - 0.00%
      - 0.00%
      Delta
      + 19.6%
      - 0.00%
      + 0.26%
      - 0.74%
      N/A

The only thing that stands out to me here is the fact that HD Tune required over 5% more CPU power during usage. This could be the result of Vista forcing a Windows XP program to function. By default, it did not function in Vista, but adjusting the compatibility fixed everything.

Real World Benchmarks

For some light real world testing, I chose a few programs to conduct multi-media recoding. For Sony Vegas, I took a 50MB .divx file, recoded it to a high-definition profile and added a watermark. For Nero Record, I took a DVD from Volume 4 of the Family Guy collection and recoded it to fit on a backup sized DVD (4.5GB). Lastly, for LAME, I first ripped Voodoo & Serano: Cold Blood to a single .WAV file. Then I ran than through a “lame -b 320 vns.wav” command.

    XP Professional

      Multi-Media Recode (Lower is Better)

    • Sony Vegas: 18m 38s
    • Nero Recode: 9m 56s
    • LAME Recode: 3m 59s
    Vista Ultimate

      Multi-Media Recode (Lower is Better)

    • Sony Vegas: 18m 21s
    • Nero Recode: 10m 31s
    • LAME Recode: 4m 07s
    .

      Delta
      + 1.5%
      - 5.87%
      - 3.34%

Interestingly enough, Vegas proved a touch better under Vista, but it fell short with the other two tests.

Final Thoughts

I don’t necessarily recommend Vista to anyone at this point in time, simply because it’s not needed. In the two months that I’ve been using it full time on my Windows machine, I’ve run into too many weird bugs that I would have liked to have avoided. This ranges from applications not working, quirky UI glitches or hardware drivers not working as well as they should.

That aside though, it’s no secret that many out there are dying to install it on their system, which is fine. As we have seen from the performance results above, should you stick with XP to save on performance? Hardly… there was not much of a difference between XP and Vista overall, I’m actually quite impressed.

As I mentioned earlier though, if you run an older machine, you probably will see a greater slowdown than I did. My test rig consisted of a nice processor, midstream GPU and 2GB of ram. If you run a last-gen CPU, low-end GPU and 1GB of ram, chances are good that you will have a sluggish running system.

What should you take away from this article? The fact that Vista actually isn’t -that- bad performance wise, if you have a decent machine. It’s leaps and bounds above the performance I saw in Beta 2, so I have no major complaints. In the months to come, companies should be continually releasing more Vista-friendly drivers, so the stability/performance side of things should get even better.

This is the fourth article in our Vista series this week. The fifth article was to be published tomorrow, but that’s pending on NVIDIAs release of an updated driver. As it stands, the current Vista drivers floating around still have some bigger issues. Once we get our hands on the final drivers, you can expect an in-depth look at gaming performance shortly afterwards. In the meantime, be sure to read through our other Vista content from earlier this week.

Discuss in our forums!

If you have a comment you wish to make on this review, feel free to head on into our forums! There is no need to register in order to reply to such threads.