A "beta BIOS update" broke compatibility with ESX, so we had to postpone our virtualization testing on our quad CPU AMD 8384 System.
So we started an in depth comparison of the 45 nm Opterons, Xeons and Core i7 CPUs. One of our benchmarks, the famous LINPACK (you can read all about it here) painted a pretty interesting performance picture. We had to test with a matrix size of 18000 (2.5 GB of RAM necessary), as we only had 3 GB of DDR-3 on the Core i7 platform. That should not be a huge problem as we tested with only one CPU. We normally need about 4 GB for each quadcore CPU to reach the best performance.
We also used the 9.1 version of Intel's LINPACK, as we wanted the same binary on both platforms. As we have show before, this version of LINPACK performs best on both AMD and Intel platforms when the matrix size is low. The current 10.1 version does not work on AMD CPUs unfortunately.
We don't pretend that the comparison is completely fair: the Nehalem platform uses unbuffered RAM which has slightly lower latency and higher bandwidth than the Xeon "Nehalem" will get. But we had to satisfy our curiousity: how does the new "Shanghai" core  compare to "Nehalem"?


Quite interesting, don't you think? Hyperthreading (SMT) gives the Nehalem core a significant advantage in most multi-threaded applications, but not in Linpack: it slows the CPU down by 10%. May we have found the first multi-threaded application that is slowed down by Hyperthreading on Nehalem? That should not spoil the fun for Intel though, as many other HPC benchmarks show a larger gap. AMD has the advantage of being first to the market, Nehalem based Xeons are still a few months away.
Also, the impact of the memory subsystem is limited, as a 50% increase in memory speed results in a meager 6% performance increase. The Math Kernel Libraries are so well optimized that the effect of memory speed is minimized. This in great contrast to other HPC applications where the tripple channel DDR-3 memory system of Nehalem really pays off. More later...
Comments Locked


View All Comments

  • joshuamora - Saturday, November 29, 2008 - link

    You can make the math yourself.
    Take N*N*8 and that will be the number of bytes.
    Divide by 1024^3 and you'll get the GBytes.
    N=43008 for 8 cores is about 85% of 16GB, ie. 2 GB per core which is reasonable for a 2 socket quad core system.
    I understand the test is done with small amount of memory: 18K /4 cores is ~ 600MB. For HPL you want to have as much memory as possible but I am not doing 4GB per core, just something reasonable.
    3GB for a system with 4 cores is a bit low.
    Now all depends what you do with the system.

    Best regards,
  • BlueBlazer - Saturday, November 29, 2008 - link

    Another thing, which operating system was your test run on?
  • joshuamora - Wednesday, December 3, 2008 - link

    Sorry for the late replay.
    The runs were done on SLES10sp2. No special configuration whatsoever, just default packages, init 5.

  • Trisagion - Friday, November 28, 2008 - link

    I'll say. I don't want to say that there's a considerable amount of Intel bias here. You've had a full Nehalem review BEFORE the chip launched and three further reviews discussing everything from the L2 cache to QPI, but not a single Shanghai review.

    What gives?

    Don't tell me you haven't completed it yet or you have to get motherboard or BIOS updates...
  • befair - Friday, November 28, 2008 - link

    hey guys ... its clear and crisp! Anandtech has been continually favoring Intel products. They should rename is AnIntelTech.com

    Wow .. product after product after product ... Intel is the king, Intel wow, Intel rocks ... wow! cant believe a site that built itself on being an objective reporter of products now wants to be goody goody with Intel across the board.

    AMD comes up with a product, not even a mention. Intel says "hmm .maybe after 10yrs, we will release a new cpu" and the site goes "wow! this is salvation to the human race as we know it!"
  • MamiyaOtaru - Saturday, November 29, 2008 - link

    You know why? Because Intel products are better now. End of story. If you have some sort of weird loyalty to a company that doesn't know you exist and doesn't care, feel free to buy inferior products. But you're really delusional if you expect everyone to share your misplaced loyalty.
  • Griswold - Saturday, November 29, 2008 - link

    Not in the IT segment. They havent been hands down better in *every* aspect for the past 5 years and now with the arrival of Shanghai, you'll have to look even closer to justify buying a Xeon based system in many more situations (shanghais efficiency blows many of Intels Xeon systems out of the water) - but surely not in all of them. It will be a while before Intel rectifies the situation with nehalem based systems and even longer in the multi-socket arena - unless Intel changes their plans drastically due to Shanghais qualities to take back marketshare AMD lost recently.

    Yes, we're not talking about gaming systems here, in case you missed it.

    With that said, its truly high time AT lives up to their standards and presents an exhaustive review.
  • BlueBlazer - Saturday, November 29, 2008 - link

    Tool, you must be referring to AMD's own published Xeon results compared to Shanghai's. However that wasn't even the top score for the Xeon system.

    Others take it with a huge grain of salt...
  • formulav8 - Sunday, November 30, 2008 - link

    Are you anandtech's personal body guard? It seems that negative or even slightly negative posts you start calling people Tools or Retards like a 12 year old. There isn't anything wrong with their posts either. I'm sure anandtech can take care of themselves and don't need a fanny publicists.
  • BlueBlazer - Monday, December 1, 2008 - link

    Here's another review http://techreport.com/articles.x/15905/6">http://techreport.com/articles.x/15905/6 which proves that Shanghai didn't "blow all Xeon systems out of the water".

Log in

Don't have an account? Sign up now