Gearing up for Lynnfield and Win7

The advent of new mainstream CPUs from Intel and a major (and promising) new release of Windows has us reworking all of our CPU test rigs in Damage Labs, in preparation for a busy period of testing.  After using the same basic hardware and software on our CPU test systems for quite a while, this seems like an appropriate time to revamp them.  To that end, boxes have been arriving via UPS and FedEx for the past few days, resulting in this excellent pile of new gear in the corner:

Yep, that ought to do it.

On the left there are new 610W PC Power & Cooling Silencer PSUs that the folks at OCZ were kind enough to send out.  We’ve been using older GameXStream 700W power supplies for at least a couple of years now, and I figured it was time for an update.  The Silencers are some of our favorite PSUs, and these are noticeably quieter than GameXStreams, which weren’t bad for their day.  We’ve backed down on the wattage rating a little in hopes of getting more efficient PSU performance when the test rigs are at idle, while keeping the right connector payload for a powerful graphics card.

Speaking of which, those are Asus GeForce GTX 260 TOP cards right next door to the PSUs.  Switching to these GeForces should reduce power consumption by roughly 30W at idle versus our previous Radeon HD 4870s.  I also kind of like the idea of going with a third-party GPU vendor instead of going AMD-on-AMD for CPU test rigs, just on principle.  Thanks to Asus for sending its excellent TOP rendition of the GTX 260.  This is a higher-clocked card that shouldn’t be the cause of many GPU bottlenecks, to say the least.

Western Digital hard drives will be the storage engines powering our new test rigs.  Those are Caviar RE3 1TB drives stacked up there.  We briefly considered SSDs, but given the big SSD performance delta when going from a new to used state, that didn’t seem like a savvy choice for CPU test systems.  Too many issues.  Not to mention the capacity constraints.  These RE3s are very nice drives that should suit our needs perfectly.  Props to WD for helping out here.

On the far right of the picture is a pair of Corsair Dominator DDR3 DIMMs intended for Lynnfield processors. These puppies are rated for 1600MHz operation at a CAS latency of 8 with only 1.65V of juice. They even auto-tune themselves to those settings via built-in profiles, in concert with the right motherboards.

Several of the right motherboards are stacked up behind the DIMMs, including the Gigabyte microATX P55 board I mentioned the other day, the P55M-UD4.  Sitting on top of them are a couple of Lynnfield-ready CPU coolers.  The big dawg from Thermalright is already up and running here now.  It seems to be quiet and effective without being especially heavy, interestingly enough.

As you might imagine, I’m looking forward to testing with our new systems.  The next step is to get Windows 7 up and running.  After that, I’ll be spending as much time as I can, within limits, trying out new applications we may want to add to our CPU test suite.  If you have suggestions, now is the time to offer them.  We have limits to what we can include, especially since each new benchmark takes time to set up and confirm, but I would like to add a few new things time time around.  Remember: the best candidates are easily timed, repeatable, don’t have crazy DRM restraints, are CPU-bound, and are fairly widely used by consumers on desktop PCs.

Comments closed
    • Clint Torres
    • 10 years ago

    Please please please include 3DStudio Max rendering benchmarks…did I say “PLEASE”?

    • Jeff72
    • 10 years ago

    Please add Linux testing via live bootable media such as using the Phoronix Test Suite PTS Desktop Live media bootable test suite.

    §[<http://www.phoronix-test-suite.com/?k=pts_desktop_live<]§

    • apt1002
    • 10 years ago

    “If you have suggestions, now is the time to offer them.”

    I would like to see a benchmark based on a computer Go program. Computer Go is computationally very intensive and can be relatively easily parallelised.

    I suggest perhaps a version of Fuego (http://fuego.sourceforge.net/). A good test would be to fix the number of playouts and time it playing a game against itself up to (say) move 100. It’s not perfectly repeatable, but the error will be small.

    • indeego
    • 10 years ago

    Software suggestionsg{<:<}g 1. 7-zip built-in benchmark. This puppy really scales well with more coresg{<.<}g 2. WinSAT. This would allow us to compare our own desktops to your benchmarks at no additional configuration/expenseg{<.<}g

      • Convert
      • 10 years ago

      Everything would be maxed out I would think, except for the HDD.

    • eitje
    • 10 years ago

    Talking about upgrading your test platform, and then thinking about looking back at older platforms, I got to thinking…

    What about taking us back 8 years, and running the benchmarks from 2001 on hardware from 2009?

    §[<http://www.techreport.com/articles.x/2975<]§ §[<http://www.techreport.com/articles.x/2784<]§ As a benchmark geek, I want to see current-gen hardware compared to hardware from long ago in the darkest recesses of the past, so that I can get a real feeling for how far we've come in the last 8 years.

      • derFunkenstein
      • 10 years ago

      well you can always download some of that stuff yourself. Everyone should have a Quake III Arena CD kicking around, and 3DMark 2001 is still out there for download.

        • eitje
        • 10 years ago

        I am completely aware of that fact, evidenced by the benchmarks I’ve run in the forums against numerous VIA products.

        What I’m missing is how I can get the hardware which Damage has for free.

    • Anomymous Gerbil
    • 10 years ago

    Work out your requirements, then work out what software meets those requirements.

    So you might list out major critera and minor criteria (for example, are the things that Tarx/#6 brings up important to you?) in a table, and assess each application against which criteria it tests for.

    This process will let you pick the smallest set of applications that test everything that’s important to you, and it will provide a transparent basis for a (hopefully) sensible discussion of the test suite.

      • Anomymous Gerbil
      • 10 years ago

      Was this such a silly idea?

        • OneArmedScissor
        • 10 years ago

        What, you mean use benchmarks that are an actual representation of reality for real use, for real people?

        But then how would they be benchmarks? We need questionable tests that show the $999 CPU is actually faster than the $200 one! :p

    • flip-mode
    • 10 years ago

    7-zip for file compression.

    • Krogoth
    • 10 years ago

    Will they run Crysis?

      • shaq_mobile
      • 10 years ago

      an excellent question! quick, robin, to the damagemobile!

    • Dent
    • 10 years ago

    I would be interested to hear how your build goes, and details of any problems you find (drivers etc). Many of your readers will no doubt be trying Windows 7 in the near future.

    I recently purchased a WD greenpower 1TB drive, and had some dramas with my BIOS detecting it as 32mb (Gigabyte P35DS3R 1.0 Bios F13).

    • SomeOtherGeek
    • 10 years ago

    Scott,

    I have to tell you this: You have too much hardware! Way too much! It is killing you. You need to offload it, so here is my address…

    Software to test on: Maybe something that uses all the cores and threads, like MS SQL Server backup or Restore. They use a ton of CPU power. Of course, it would cost money to get the software. Maybe running a script that opens a virtual box for every core and thread. Do you have a application recorder? That can create a script of what you are doing and then you can just copy the script to every machine and run it… Or maybe I can just shut up?

    Second warning, you need to offload the hardware now! Again, my address is: see Jeffry55.

      • moshpit
      • 10 years ago

      Folding@Home in VM’s is a great way to measure CPU power as well. And since that’s one of the best ways of achieving optimum scaling in folding for the number of cores you have present, it would also serve well to show advantages or disadvantages to HT being on or off for the Core i5 versus i7.

        • SomeOtherGeek
        • 10 years ago

        This guy is on point!!

        • OneArmedScissor
        • 10 years ago

        It would be great, except it serves no functional purpose to most people. I’m a lot more interested in getting work done than causing my electricity use to skyrocket.

        Sorry, but I think folding scores are the equivalent of wPrime benchmarks. It just doesn’t help me.

          • SomeOtherGeek
          • 10 years ago

          But it shows you the full power of the system. What is the use of a 4 core – 8 thread if you are not going to use them. So, why not test them?

            • flip-mode
            • 10 years ago

            q[

    • Igor_Kavinski
    • 10 years ago

    Winrar 3.90 and Windows Live Movie Maker benchmarks on Windows 7 x64 please.

      • eitje
      • 10 years ago

      use 7-zip instead of WinRAR! 😛

        • cygnus1
        • 10 years ago

        7-Zip = FAIL
        WinRAR FTW!!!

          • SuperSpy
          • 10 years ago

          Prove it or get out.

            • MadManOriginal
            • 10 years ago

            One has RAR in the name and that’s all the proof that’s needed.

            • Usacomp2k3
            • 10 years ago

            I very much prefer the simplicity and free-ness of 7-zip.

            • maxxcool
            • 10 years ago

            +1 for 7Zip… i have never used another zip utility since then…

          • Meadows
          • 10 years ago

          I say WinZip.

            • Farting Bob
            • 10 years ago

            I say Windows built in .zip unpacker!

    • Game_boy
    • 10 years ago

    Have you looked at the Phoronix Test Suite? It runs benchmarks on a variety of common applications and is easy to set up. And it’s free/open source.

    §[<http://www.phoronix-test-suite.com/<]§

    • glynor
    • 10 years ago

    This is probably too specialty, but I’ll throw it out there anyway…

    I’d love to see encoder tests using Telestream Episode and Flash-compatible H264 video encodes in HD. Something like: 1080p 4:2:2 (or 4:4:4) source video in a high-quality editing codec (Blackmagic Design codecs, Apple ProRes, or Avid DNxHD) and encoding to approx 1mbps 720p H264 MP4s would be ideal.

    While probably not common on your average gamer’s desktop, this is a very common work scenario, and one that would actually make use of a new high-end multi-core CPU. Telestream’s Episode is a common (and well loved) encoder used by video pros.

      • wibeasley
      • 10 years ago

      Would those results be similar to the existing tests, _[http://www.techreport.com/articles.x/17023/8<]§

    • SecretMaster
    • 10 years ago

    Are you using Antec Skeletons for all of these testing rigs or some other case?

      • d0g_p00p
      • 10 years ago

      I would hope a open bench. Those Antec’s are full of issues for takedown/rebuilds. Just google for the issues.

        • SecretMaster
        • 10 years ago

        Oh. I know Scott showed interest in them when they were first announced, I was wondering if he actually got some.

          • MadManOriginal
          • 10 years ago

          I would think that those things would be more trouble than they’re worth versus a simple testbench made of a mobo tray with expansion card holders.

            • Damage
            • 10 years ago

            Yeah, Skeletons are pretty big and would kind of get in the way. I like the idea, but it’s not ideal for our purposes. I’m just using an open test bench, with bare systems sitting on mobo boxes. 🙂

            • wibeasley
            • 10 years ago

            What’s the advantage of setting them on cardboard boxes? Does the table conduct electricity?

            • Damage
            • 10 years ago

            Expansion slot covers in back stick down. Have to have the mobo propped up on something.

            • MadManOriginal
            • 10 years ago

            I was actually thinking of a cheapo-style ‘testbench’ like what I have although I rarely use it – a pullout mobo tray with the expansion slot section from a late 90s Inwin ATX case. It’s nice to have screw down mobo risers and a way to secure cards. Then again when I do use it it’s with stuff that matters to me not freebies 😉

            If something like that sounds useful just pick up an old free case from where ever and go to town with a dremel if it hasn’t got a slideout tray.

    • rUmX
    • 10 years ago

    That P55-UD6 box is huge!! Twice as big as my EX58-UD4P.

    • eitje
    • 10 years ago

    Windows Live Movie Maker 14 just came out. I could see using it to re-encode WTV/DVR-MS files into WMV.

    This, for whatever reason, is a laborious process on a P4-M. 😛

      • MadManOriginal
      • 10 years ago

      I think it’s laborious because it’s a P4-M 😉

        • eitje
        • 10 years ago

        The illusion that “for whatever reason” provided allowed me to feel better about myself! 😛

    • Tarx
    • 10 years ago

    1) Turboboost vs Threads
    As the difference between Turboboost and stock speed is quite a bit more significant than with the current i7s…
    I wonder if some games (and apps) that spawn multiple threads will have a negative impact due to turboboost. (i.e. 2 threads running at a significantly faster speed turbo boost speed versus 4 threads that have the CPU running at stock speed). Of course this probably doesn’t apply if all 4 threads are heavily used. But perhaps a game like FarCry2 that heavily uses 1 thread and lightly/moderately uses other threads might show more threads = worse performance? If the number of threads can be controlled, it would be interesting to see if this can be tested.

    Then does the Nvidia’s GPU also spawns another active thread by itself? If so, then would be interesting to compare with ATI cards…

    2) Turboboost vs Heatsink (& cooling)
    IIRC, Turboboost depends on power use & cooling. Just wondering about how well Turboboost will work with stock cooling versus the aftermarket cooling to be used in the test systems.

    3) Impact of DDR3 RAM (e.g. cheap DDR3-1333 C9 versus the higher-end DDR3-1600 C8 that the test system has). For building a “budget” system with i5-750 with a $120 MB, the price difference between the cheap and more expensive 8GB DDR3 RAM would be significant. Is there a big difference in performance to be worth paying the extra $$?

      • MadManOriginal
      • 10 years ago

      Yeah Turboboost needs some investigating if not in the initial article then shortly afterward. My particular question about it is how an i5/i7 using Turboboost compares to an overclocked Core 2, or alternatively how they compare at roughly similar stock speeds with Turboboost disabled.

      Turboboost is going to throw a fair amount of uncertainty in to not only testing but also ‘how much faster is _blank_’ types of questions.

        • OneArmedScissor
        • 10 years ago

        I don’t know if it really needs any “investigating.” It’s going to change from test it test, and most people are going to leave it on, so I’d prefer a representation of reality.

        What they need to do is just list the darn clock speed that it ran during the test. I can’t stand it when sites list things like “Core i7 920 2.66 GHz” in their system specs, when a huge focus of the article is benchmarks of games that undoubtedly ran at the max turbo boost.

          • MadManOriginal
          • 10 years ago

          Yes I certainly wouldn’t want it left out entirely since it’s there and is a useful feature and wasn’t suggesting that. I guess what I want to know is how an oc’d Core 2 or Phenom II compares to a CPU with Turboboost because it’s not like comparing two non-Turbo CPUs with oc’s. Even the ‘oc them both’ argument doesn’t hold because you aren’t going to get as much of a Turboboost on an oc’d i5/i7.

          Showing what speed the CPU runs a benchmark would be a simple thing to do and is a good suggestion.

            • wibeasley
            • 10 years ago

            I was under the impression that the clock speed has variation during a long-running benchmark -kinda like cpu% levels.

            • OneArmedScissor
            • 10 years ago

            You have a valid point, but my thinking is that you could piece it together yourself if the actual clock speed was listed. The turbo boost clock speeds and clock speeds of various Core 2 Quads and Phenom IIs all coincide.

            If they were overclocked as far as they go, in an attempt to determine the “true winner,” they’d all end up being overkill for most things, and they wouldn’t be fit for use in anything serious enough to call for that sort of processing power. I doubt it would tell you anything worthwhile.

            I’m more interested in the possibility of getting away from overclocking, while still getting as much out of the CPU as you’d ever really need.

            For most truly CPU intensive things, you don’t want to be overclocking and risking instability, so turbo boost is potentially a god send. No website I’ve seen does a very good job of representing the facts and allowing you to judge for yourself, though.

            wibeasley, you are undoubtedly correct, but just as speedstep and cool ‘n quiet can cause the multiplier to fly up and down, it still sticks to one speed for most of the time.

    • scribly
    • 10 years ago

    “don’t have crazy DRM restraints”
    that’s gonna be a tough one

    • Obsidian
    • 10 years ago

    If I wasn’t already spending all my money on new peripherals I might be a little jealous of you Scott. What are you going to do with all the old hardware I wonder?

      • SomeOtherGeek
      • 10 years ago

      I hope he is farming them…?

    • wibeasley
    • 10 years ago

    On my wishlist for a new TR cpu benchmark: Something that runs with data that’s almost entirely contained in cache and linearly scales to mutiple threads.

    • StashTheVampede
    • 10 years ago

    First test: does it blend?
    Second test: oh wait

      • KyleSTL
      • 10 years ago

      I think you mean “Will it blend?”

        • _Sigma
        • 10 years ago

        internet meme fail

    • FubbHead
    • 10 years ago

    Drool.. I want new hardware too! Oh, I wish I wasn’t a poor student… 🙂

Pin It on Pinterest

Share This