This is why FB-DIMMs are a legitimate issue.

Discussion in 'Mac Pro' started by milozauckerman, Sep 14, 2006.

  1. suneohair macrumors 68020

    suneohair

    Joined:
    Aug 27, 2006
    #2
    Why would they not test the 2.66?

    Plus there are 2 variables

    1: FB-Dimms
    2: All apps arent optimized to use multiple cores.
     
  2. badmofo9000 macrumors regular

    Joined:
    Feb 14, 2005
    Location:
    Shores of Lake Michigan
    #3
    I am sorry but what a useless test. The imac has a faster cpu. All the cores in the world aren't going to make for apps that are not multi core aware.
     
  3. [G5]Hydra macrumors regular

    Joined:
    Jul 2, 2004
    #4
    What does it have to do with those tests? They tested a stripped down 2.0GHz Mac Pro against faster clocked iMacs. The Mac Pro was tied at second slowest of 7 computers as far as clock speed goes and still won 3 of 8 tests outright in the first series and was only 10fps slower in UT despite the fact it had the lowly 7300GT compared to the iMac's far superior 7600GT. Most of those test don't appear to be memory bandwidth limited as it is so FB-DIMMS wouldn't likely make a huge difference.

    -Jerry C.

    Oh and by the way, the way the iMac was configured it probably cost more than the Mac Pro they tested.
     
  4. suneohair macrumors 68020

    suneohair

    Joined:
    Aug 27, 2006
    #5
    I just hacked the 2.66 MP benches into the iMac C2D to show the difference.
     

    Attached Files:

  5. milozauckerman thread starter macrumors 6502

    Joined:
    Jun 25, 2005
    #6
    The stock 2.0GHz vs. upgraded 2.33 leans to the 2.33 by about $150 (unless you'd like to use something more than the crippling 1GB stock - in which case they come just about even). Obviously this is skewed given the LCD factor.

    All of those tests - most especially the Photoshop and Aperture (where an Apple-optimized program still barely beats the iMac) tests - are memory-reliant, which (to beat the drum again) make up a significant portion of the Apple user base.

    Multi-core optimization is a factor, but that's not really relevant to questions of real-world performance - theoretically a 32-core SuperMacPro would have more processing power, but if that doesn't translate into usability, who cares?

    (ironic to see the clock speed argument raised here - a bit of a backhanded way to recognize that FB-DIMMs all but eliminate the 100% FSB advantage enjoyed by the Mac Pro)
     
  6. ipoddin macrumors 6502a

    ipoddin

    Joined:
    Jan 6, 2004
    Location:
    Los Angeles
    #7
    The iMac costs $2374
    The Mac Pro they tested costs $2124 (assuming 160gb hard drive option)

    While the iMac was $250 more, it also comes with a 24" screen, airport express and bt built in, as well as a 250gb hard drive standard.

    Looking at suneohair's hack of the benchmarks, it's surprising that even the 2.66ghz MP doesn't destroy the iMac either. Then there's also this from the article:

     
  7. suneohair macrumors 68020

    suneohair

    Joined:
    Aug 27, 2006
    #8
    Ok. Macworld is no longer credible. You can not just add all the cores together and magically have 8Ghz. It doesnt work that way. That is by far the stupidest thing I have heard in awhile.

    Moving along, the problem is not FB-DIMMs, the problem is multi-core optimization in applications. Let me see some FCP or Aperture benchmarks.

    That is where the improvements will lie, and people who use these apps are the ones buying the Mac Pro. I personally am not buying a Mac Pro to run random benchmarks and make zip files.

    Hoepfully CS3 will also be optimized for more than 2 core operations.

    To think that the FB-DIMMs are dragging down a system with 4 cores is asinine.
     
  8. milozauckerman thread starter macrumors 6502

    Joined:
    Jun 25, 2005
    #9
    You mean like the Aperture benchmark they included?

    The closest 'benchmark' you're going to find for FCP is iMovie.
     
  9. milozauckerman thread starter macrumors 6502

    Joined:
    Jun 25, 2005
    #10
    To argue that FB-DIMMs don't hamper memory-intensive programs is remarkably asinine.
     
  10. suneohair macrumors 68020

    suneohair

    Joined:
    Aug 27, 2006
    #11
    You mean the Aperture benchmark they failed to do for the 2.66 Mac Pro?
     
  11. suneohair macrumors 68020

    suneohair

    Joined:
    Aug 27, 2006
    #12
    I am not saying they don't. But your post does not prove that only FB-DIMMs are causing the problem.
     
  12. Spanky Deluxe macrumors 601

    Spanky Deluxe

    Joined:
    Mar 17, 2005
    Location:
    London, UK
    #13
    This is an absolute joke of a comparison. Of course I would expect a 2.33Ghz Core 2 Duo to beat a quad core 2.00Ghz Woodcrest in most applications. They are essentially the same chips except for the number of cores and the speed. Most applications aren't optimised for more than two cores if that. They should try these tests again but run two of the tests at the same time. Combining processor speeds like that is what a complete newbee would do, its not something I would expect from a reputable site.
    The iMac 24" machines are very impressive to be sure and great value for money but this is an unfair comparison. Its like comparing an ueber tuned hatchback to a base model BMW 3 series.
     
  13. milozauckerman thread starter macrumors 6502

    Joined:
    Jun 25, 2005
    #14
    Nor did I say it was. They are, however, a rather large factor - and most importantly, a highly expensive factor - in the equation.

    It is simply a huge mistake (as far as consumers are concerned) for Apple to have used only Xeon in their sole desktop.
     
  14. suneohair macrumors 68020

    suneohair

    Joined:
    Aug 27, 2006
    #15
    You may very well get your little xMac. However, I don't understand why you are taking out a personal assault on the Mac Pro. It is a massive waste of time. If you dont want it, shut up about it. Most people here can read just fine, we have read that FB-DIMMs cause a hit in performance. Yet we still choose to buy it because it suits our needs now and the computer will become faster as more applications can take advantage of its power.

    Do you honestly think that ranting here is going to change the fact that Apple is using Xeon, Xeon is made by Intel, and Intel requires FB-DIMMs? I don't think so.

    At this point I wish Apple would release a G4 Cube-esque computer so you people would shut up about it.
     
  15. Spanky Deluxe macrumors 601

    Spanky Deluxe

    Joined:
    Mar 17, 2005
    Location:
    London, UK
    #16
    Well said. How long do you expect to hold out for an xMac anyway? If anything, the article you linked to has shown why there is no need performance wise for Apple to release a mid sized mac.
     
  16. macenforcer macrumors 65816

    macenforcer

    Joined:
    Jun 9, 2004
    Location:
    Colorado
    #17
    Not a damn thing can be done about this FB-DIMM problem. Its intels doing and won't change until they do.

    Anyone notice FB-DIMM prices going through the roof? I got the 2gb Dimms from newegg at $349 and now they are like $400.
     
  17. [G5]Hydra macrumors regular

    Joined:
    Jul 2, 2004
    #18

    I don't understand your argument. The iMac does pretty well against a similarly clocked Mac Pro. The top iMac is slightly more expensive than the Mac Pro but comes with a better graphics card and the 24" display. What do you expect, a 4 core MP w/a display to be cheaper? So if you don't like the Mac Pro get the iMac. If you dropped a 4 core Core Duo 2 into an iMac and tested it in many of those tests you'd probably see hardly any change in the results. The Mac Pro is a machine intended to be used by professionals to do work. This is not to say that "regular" people don't have a need or desire for one but they are not the target demographic for the product. I can see you are quite upset they didn't release the X-Mac consumer tower but I doubt Apple is going to make it. It would be nice if they did but tearing down the current pro tower isn't gonna make that happen. If you really want a single CPU tower just buy a Mac Pro of the speed you like and pull the other CPU out and sell it on ebay.

    The Anandtech article that many people seem to be using as dogma wrt FB-DIMMs leaves out some very important tests that would show that gaming, a fairly intense memory bandwidth test, at higher than 1024x768 reses will significantly diminish the gap between a Core Duo2 system using regular DDR2 vs. the Xeon using the FB-DIMMS. How many people are going to buy a 1900XT for a $2500 computer to play at 1024x768? If Anandtech wanted to actually bench performance and not advance the theory that was postulated in their original article about the Mac Pro wrt FB-DIMMs being so bad they would have shown more than the 1024x768 reses after all.


    -Jerry C.
     
  18. Demoman macrumors regular

    Joined:
    Mar 29, 2005
    Location:
    Issaquah, WA
    #19
    I must have missed the FB-Dimm discussions; what are the issues?
     
  19. [G5]Hydra macrumors regular

    Joined:
    Jul 2, 2004
    #20
    Part of the problem is availability and profiteering. I got 2x1GB from Crucial for $399, yes with the right heat-sinks and all, and the next day they were up to $499 and even hit as high as $530. OWC has the ones with the red heat-sinks that started at $550 for the 2x1GB then dropped to $500 and now stands at $480. The prices are coming down and should become better as they become more common.

    -Jerry C.
     
  20. suneohair macrumors 68020

    suneohair

    Joined:
    Aug 27, 2006
    #21
    Yeah I am going to hold out as long as I can with the 1GB. Hopefully as the end of the year gets closer the price will come down.
     
  21. amin macrumors 6502a

    amin

    Joined:
    Aug 17, 2003
    Location:
    Boston, MA
    #22
    Agreed. AnandTech seems to be really pushing this "issue," without any solid data. In their last installment, they compare a Mac Pro with only two RAM slots occupied, which while the default arrangement is well known to them to be suboptimal for the dual Xeon processors, to a Conroe setup where two RAM chips is optimal. They should compare a Mac Pro with 4x512MB to a Conroe setup with 2GB (any arrangement), and I bet those minute "latency issues" would melt away in a real world comparison.
     
  22. [G5]Hydra macrumors regular

    Joined:
    Jul 2, 2004
    #23
    Actually a 4x1GB with 4 dual rank DIMMs is optimal right now but the problems I have with Anandtech's review are:

    A) They state how much RAM the Core 2 Extreme X6800 system has at 2x1GB but show nothing for the Mac Pro and there is an issue with the Mac Pro currently under XP as it only sees 2GB of RAM and no more.

    B) They appear to be using 2x1900XT Crossfire but the Mac Pro's motherboard only could do 8 lanes on each - it can only do 16 lanes on one card at a time - and potentially limit bandwidth

    C) The Mac Pro, and all intel Macs for that matter, using XP is a work in progress as the fact they got display corruption with a 256MB 1900XT for no know reason. You can't make definitive judgments when you are dealing with what amounts to a beta. If they really wanted to test their hypothesis they would have bought a Dell 690 dual Xeon.

    -Jerry C.
     
  23. JayRx1981 macrumors newbie

    Joined:
    Sep 9, 2006
    #24
    I'm a consumer. I'm not the least bit concerned with Apple's choice to use only Xeons. In fact, my experience with this particular mac pro on which I'm typing my response is better for the SMP environment enabled by the Xeons. I've yet to max the processing capabilities of this machine out as a typical (well, ok, semi-typical) consumer. If I've gone above 75% cpu utilization yet I'd be surprised. This machine has worked phenomenally for me so far. In fact, the only hiccups I've had have occurred when using so many programs at once that I ran out of RAM (an extra gig on order and set to arrive tomorrow). As a consumer, I'm incredibly happy with my choice. That I *may* suffer a slight performance hit due to FB-DIMMs in a single-task environment really doesn't bother me in the slightest since, well, I'm almost *NEVER* in a single-task environment.

    I'd LOVE to see a set of benchmarks run with multitasking taken into account. I'd love to see the performance of a top of the line iMac 24" vs even the base 2x2GHz Mac Pro (with 2GB ram to level the amount) with both of them running a stock Aperture benchmark + FCP benchmark + Handbrake/FFMPEGX encode benchmark running simultaneously. Somehow I think that all of a sudden the stock mac pro would wipe the floor with the iMac, even with the FB-DIMM performance hit.

    Now, I get why you're not happy with the Mac Pros. I get that many people here want a headless upgradeable Mac (non-pro) desktop. But complaining about the Mac Pro using FB-DIMMs in a non-apple owned forum is, well, a waste of time. It's your time to waste, to be sure, but it's not doing you much good. If you're upset about the choices that went into making the Mac Pro--simply don't buy one. Wait for your headless consumer desktop model or bite the bullet and buy/build a winblows box. The Mac pro may not be right for you, but for many of us non-professional consumers, it WAS the right choice for ourselves.
     
  24. MacPhil5 macrumors newbie

    Joined:
    Sep 3, 2006
    Location:
    New York
    #25
    just to show that multiprocessor optimization IS an issue, here is my experience using Handbrake.

    It is supposed to be a multi-threaded wideo encoding program, and yet it didn't load the 4 cores of my MP. I assumed it was a memory latency problem.

    But then I ran this test: I duplicated the Handbrake app and ran both applications at the same time encoding two videos (MPEG4) at the same time and I found out that both encoding processes were running at 115 fps, whereas one alone ran at 120 fps. The video processing power was now 230 fps, three times that of my PowerMac G5 (dual 2.5) and all four cores were loaded. So I assume it was a multiprocessor issue…
     

Share This Page