The Notebook Review forums were hosted by TechTarget, who shut down them down on January 31, 2022. This static read-only archive was pulled by NBR forum users between January 20 and January 31, 2022, in an effort to make sure that the valuable technical information that had been posted on the forums is preserved. For current discussions, many NBR forum users moved over to NotebookTalk.net after the shutdown.
Problems? See this thread at archive.org.

    This test makes Dual-Channel look awesome

    Discussion in 'Hardware Components and Aftermarket Upgrades' started by jtodd939, Nov 9, 2006.

  1. jtodd939

    jtodd939 Notebook Enthusiast

    Reputations:
    0
    Messages:
    49
    Likes Received:
    0
    Trophy Points:
    15
    http://www.tcmagazine.info/articles.php?action=show&id=128&perpage=1&pagenum=4

    Oddly though, even the authors say in the conclustion that dual-channel doesn't give that much advantage. Yeah, it doesn't always have the fastest FPS in games, but look at the advantage it gives in HDD, XP Bootup, etc. It looks like it gives noticeable speed improvement. It's not a be-all-end-all but it does look like it gives a boost to functioning.
     
  2. Jalf

    Jalf Comrade Santa

    Reputations:
    2,883
    Messages:
    3,468
    Likes Received:
    0
    Trophy Points:
    105
    Of course dual-channel *can* be awesome. It depends on the rest of the system as well.

    The article you linked to is 1.5 year old. Back then, the reference system was an Athlon 64 running DDR at 400 MHz.
    That gives you a lot less bandwidth than DDR2 at 667, which is common today. The A64 has a smaller cache as well, which means data more frequently has to be retrieved from RAM.
    Dualchannel basically just doubles the RAM bandwidth available. And of course that makes a difference in cases where you actually need more bandwidth.

    A C2D needs less RAM bandwidth (because of its bigger cache), and when running on a DDR2 system, you tend to have more bandwidth available. So those factors mean that a single RAM channel can usually offer just about enough bandwidth to keep the CPU fed. In some cases you might still benefit from more bandwidth, but most of the time, the C2D is happy with one 553 or 667 mhz DDR2 channel.

    1.5 years ago, the situation was different, and you often needed more bandwidth than a single 400 MHz DDR channel could supply, so dualchannel gave a bigger performance boost.
     
  3. jtodd939

    jtodd939 Notebook Enthusiast

    Reputations:
    0
    Messages:
    49
    Likes Received:
    0
    Trophy Points:
    15
    Thx Jalf. That made a lot of sense. Thanks for taking the time to educate me about that. Much appreciated.
     
  4. matt.modica

    matt.modica Notebook Consultant

    Reputations:
    7
    Messages:
    248
    Likes Received:
    0
    Trophy Points:
    30
    Not necessarily true. DDR667 won't me that much faster than DDR400, in fact, the difference will be barely noticable because of the higher latencies. In fact, AMD was probably better off not introducing DDR2 into socket AM2, the only reason they did it was for marketing. They were losing money to Intel, because it went on with its "faster processors and faster RAM" advertising campaign.

    Also, the Athlon doesn't need a larger cache because of the memory controller. The AMD chips since the K7 Athlon XP have had these, and basicly, they don't need to process memory instructions. Thats why an Athlon 64 running at 2 GHz with a 1 MB cache will benchmark higher than a P4 (Prescott) running at 3 GHz with a 2 MB cache. An Athlon will have much less need for cache and speed, and will actually use the cache less often and won't have to access the RAM as much because of this.

    The Core 2 Duo (Conroe), while being able to process many more instructions per clock cycle than the Pentium D (Pressler), still needs the 2 or 4 MB cache because it still lacks the memory controller that made AMD famous. And it still needs the faster RAM because it takes longer to access the RAM. Why? Because rather than just "asking" the on-die memory controller, it has to go though the FSB, which has a relitively slow bandwidth, and it has to "ask" the memory controller, which is located on the motherboard.

    Bottom line: AMDs don't need DDR2, they use effeciency to acomplish what intel used speed to (up untill the Conroe, of course). And DDR2 will give a very small performance increase over DDR.

    And if you read the "final verdict" at the end of the series of articles, it said that "on today’s systems, the advantages of Dual Channel memory setups are negligible".
     
  5. Gator

    Gator Go Gators!

    Reputations:
    890
    Messages:
    1,889
    Likes Received:
    0
    Trophy Points:
    55
    Wait the K7's have a memory controller on die? Wow I need to research this...
     
  6. sionyboy

    sionyboy Notebook Evangelist

    Reputations:
    100
    Messages:
    535
    Likes Received:
    0
    Trophy Points:
    30
    No the on die memory controller first appeared on the Athlon64/Opterons. AthlonXPs had memory controller in the northbridge.
     
  7. Gator

    Gator Go Gators!

    Reputations:
    890
    Messages:
    1,889
    Likes Received:
    0
    Trophy Points:
    55
    Well something's inconsistent here, if the K7 Athlon XP's have the mem controller in the Northbridge then it still goes thru the FSB and you'll have the same constraints as the Intel counterparts and the same need for cache and speed.
     
  8. sionyboy

    sionyboy Notebook Evangelist

    Reputations:
    100
    Messages:
    535
    Likes Received:
    0
    Trophy Points:
    30
    They do, the K7s didn't have the on die memory controller. The K8s (Athlon64, Turion, X2) do have on die memory controllers.
     
  9. Jalf

    Jalf Comrade Santa

    Reputations:
    2,883
    Messages:
    3,468
    Likes Received:
    0
    Trophy Points:
    105
    That's why I only talked about bandwidth, not performance. On a bandwidth-starved system, the extra bandwidth comes in handy, even if latencies also go up.

    Not exactly true. Or rather, it depends on what "need" we're talking about. The onboard memory controller lowers memory latency, yes, so it can get more performance with a smaller cache, but it doesn't reduce the number of memory accesses needed. So the A64 needs more memory accesses than the Core 2 Duo, because it has a smaller cache to store frequently used data in. (Because of the memory controller it can execute those accesses with lower latency, yes, but that doesn't affect the amount of bandwidth required)

    Note that I wasn't talking about whether AMD or Intel CPU's perform best, or why that is. I'm only trying to explain why dualchannel RAM might give better results on one system than the other.

    And as said above, the K7's did *not* have an on die memory controller. All K8-derived chips do.