Memory Latency and Bandwidth

We've never been able to look at some of the low level characteristics of Intel's Core architecture, and although we didn't have enough time to do a thorough run of low level benchmarks we were able to run ScienceMark 2.0 in order to get an idea of how the Core 2 Extreme stacked up against the FX-62 in terms of memory latency and bandwidth.

We had seen Conroe performance results that showed the new architecture being able to offer fairly competitive memory access latencies to AMD's architecture, without the need of an on-die memory controller. Our ScienceMark 2.0 results confirm just that:

ScienceMark 2.0 - Memory Latency (256-byte stride)

While AMD still offers lower memory latency, the Core 2 Extreme X6800 is very close in comparison - especially considering that it has no on-die memory controller. With lower clock speeds than its Pentium D siblings and a faster FSB, memory access latency is reduced tremendously with Conroe. On a larger scale, through a very effective cache subsystem as well as memory disambiguation, Conroe can offer significantly improved memory performance compared to its predecessors, including the Athlon 64 X2/FX.

ScienceMark 2.0 - Memory Bandwidth

ScienceMark's memory bandwidth results offer a very telling story, showing us the bandwidth limitations of Intel's FSB architecture. While the FX-62's peak theoretical bandwidth is not achieved in real world, you can see how AMD's Direct Connect architecture offers higher limits for chip-to-chip communication.

Index Business Application Performance
Comments Locked

134 Comments

View All Comments

  • thestain - Tuesday, June 6, 2006 - link

    Be nice to see the benchmarks with the settings at default, 2X and 4X default to see if the settings requiring more work from cpu and greater memory use will change the results any.

    Thoughts?
  • IntelUser2000 - Tuesday, June 6, 2006 - link

    quote:

    And based on these performance figures,

    a 4x4 board with TWO FX62 will vastly outperform a lonely Intel Conroe.

    And for heavy I/O ie beating network and disk to death, Intel has not been shown to have the performance headroom. AMD I/O will scale nicely.

    When Intel counter with quadcore, they will find their FSB even more limiting, at which point the wisdom of the Hypertransport approach will be evident.

    Depends how quickly 4x4 comes to market (but said to be 2H2006)


    Err, I DO hope for AMD a 2x core part would beat Intel's platform which would cost 1/4 of the cost wouldn't you??

    And then only in multi-threading. If you want to argue about MT, then Pentium D 805 is the absolute leader over AMD's here.
  • peternelson - Wednesday, June 7, 2006 - link

    Only if price is a major consideration.

    In a highend system you will be paying say $795 for a myrinet lan card, lots for an areca 1260 8 port sata, more for your TWO high end video cards.

    The processor(s) then become a smaller proportion of the total system cost, so the price differential between 805 and FX62 becomes less important.

    Plus the 805 doesn't support hardware virtualisation so is not a good choice.
  • mpeavid - Thursday, June 8, 2006 - link

    Uhm - cost is a cost.
    If one product cost significantly more, it is duely noted.

    Cost is still the PRIME factor in any enterprise endeavor. Business to make money.

  • neweggster - Tuesday, June 6, 2006 - link

    quote:

    And for heavy I/O ie beating network and disk to death, Intel has not been shown to have the performance headroom. AMD I/O will scale nicely.


    The two are comparable at limiting the max available I/O's on each cpu, thus both would be almost identical in comparison and AMD still would not have any lead in this area since both are similar here. Take a look at the communication bench to see for yourself on that.

    Just curious, where did you see Intel not having any performance headroom and where is it show AMD I/O scaling better? I guess your an AMD fan? At a glance one observation I took in was how AMD and the rest say how AM2 has a faster resolution to their current AM2 fastest lineup, saying how AM2 has a faster cpu in the works and it will eventually outperform conroe.

    Based on that argument I seen on many sites, it is my opinion that no multi billion dollar industry has a 1-man marketing team that would put together a product not worthy of such big changes knowing the community of people they are selling to are educated more on researching and buying the best performance bang for their buck. Knowing this they surely wouldn't have released AM2 currently as is if they had already a faster version capable of competing with its rival Intels top performing release.

    If AMD is struggling this bad on a new core blah blah blah, then wouldn't you think in terms of processor technology and advancement that they couldn't possibly turn out 5+ years of research in a single year to outperform conroe. I know conroe is been in the works for a really long time now and all im saying is if AM2 was so bad and only gave a 2-4% increase in performance over its 939 top counterpart then why release at all? Just hold off and wait till you have something that can actually perform closer to its rival Intel. Yeah I know its been a war back and forth for years but in the past till now you can assume how bad AMD must be with Intels conroe on top with so much headroom.
  • bob661 - Tuesday, June 6, 2006 - link

    quote:

    I know conroe is been in the works for a really long time now and all im saying is if AM2 was so bad and only gave a 2-4% increase in performance over its 939 top counterpart then why release at all?
    Because you and I, the enthusiast, don't make AMD nor Intel any money. It's the J6P's that buy all of this stuff and keep those guys in business. AMD needed to transition to DDR2 and they did it now vs later. Why didn't Intel wait till Conroe to transition to DDR2 and 65nm when they knew they had a winner?
  • neweggster - Tuesday, June 6, 2006 - link

    quote:

    Because you and I, the enthusiast, don't make AMD nor Intel any money. It's the J6P's that buy all of this stuff and keep those guys in business. AMD needed to transition to DDR2 and they did it now vs later. Why didn't Intel wait till Conroe to transition to DDR2 and 65nm when they knew they had a winner?


    I understand that. Mainly Intel can take advantage of DDR2 on its core design early on and stretch new technology while keeping a clear indication on their doorstep that they had conroe to push them ahead and keep sales and the people happy.

    AMD didn't have anything more advanced then its FX60 939 to shove in the mainstream with AM2 only being what 2-4% faster, in this sense its only fair to question my original thought, would AMD be better to not launch AM2 for the sake of keeping face with the people and having it look like they need to get something that can beat this new conroe; furthermore, then releasing a small gain that appears to be a flop and having people say no to future AMD.

    For this purpose I think it would of been wise decision by AMD to let this conroe come out and have a winning solution launched to beat it or compete with it when the time was right, rather then launching a marketing flop and causing people like us to lean towards the opposition and loss of trust towards AMD. =)
  • bob661 - Tuesday, June 6, 2006 - link

    quote:

    For this purpose I think it would of been wise decision by AMD to let this conroe come out and have a winning solution launched to beat it or compete with it when the time was right, rather then launching a marketing flop and causing people like us to lean towards the opposition and loss of trust towards AMD. =)
    I think AMD performed as they should've by getting off the DDR bus and letting DDR2 fully take the market. K8L won't be here till next year and I'm sure the memory manufacturers would like to phase out DDR. Especially since we ALL knew AM2 wasn't going to be any more of a performer than 939.
  • Gary Key - Thursday, June 8, 2006 - link

    quote:

    K8L won't be here till next year and I'm sure the memory manufacturers would like to phase out DDR.
    The last volume orders for DDR have been sent to the FABs already, DDR production at the FAB side will be less than 10% by the end of Q3.
  • neweggster - Tuesday, June 6, 2006 - link

    quote:

    I think AMD performed as they should've by getting off the DDR bus and letting DDR2 fully take the market. K8L won't be here till next year and I'm sure the memory manufacturers would like to phase out DDR. Especially since we ALL knew AM2 wasn't going to be any more of a performer than 939.



    I agree in some ways. I never said that it was a something bad in context here but that AMD should of waited till they had something to show for.

    And DDR2 is not fully taken the market yet, not for a long time by any standards. DDR still remains top sales and DDR2 isn't going to really start seeing numbers matching DDR sales for another 1-2 years because all the major sales are exclusively business, schools, and computers bought on a basis by people to do small tasks where many P4's and regular AMD based machines are still the biggest sales for DDR, and DDR2 being more expensive to build with while keeping the cost down, granted sales are on a decline since the induction of AM2.

    From my other statement it reflects that no matter what AMD has done, rather waiting or releasing AM2 like they did, it does not open DDR2 sales that much yet. If you think about it DDR2 sales are still in my estimate a 80% or more Intel sale, because AM2 is proven to be something that enthusiast would rather not go to till it matures more. And then all the major retailers will not promote AM2 so hard to sell just because of the baring they have seen, its not yet cost effective for them. So by this it would make no difference on DDR2 making it more mainstream.

    The move to DDR2 is still very slow and will be so for a time. Trust me on that.

Log in

Don't have an account? Sign up now