Results 1 to 14 of 14
  1. #1
    Join Date
    May 2004
    Location
    Atlanta, GA
    Posts
    3,872

    SQL benchmarks: Xeon Woodcrest vs Opteron Socket F

    here you have it:
    http://tweakers.net/reviews/646
    woodcrest vs dempsey vs opteron socket F
    1-core vs 2-core vs 4-core
    MySQL & PostgreSQL
    C.W. LEE, Apaq Digital Systems
    http://www.apaqdigital.com
    sales@apaqdigital.com

  2. #2
    Join Date
    Jun 2004
    Location
    Oregon
    Posts
    1,315
    Thanks.
    Great performance of Woodcrest.

  3. #3
    Join Date
    Jun 2002
    Location
    PA, USA
    Posts
    5,143
    Nice article too. Thanks!
    Fluid Hosting, LLC - Enterprise Cloud Infrastructure: Cloud Shared and Reseller, Cloud VPS, and Cloud Hybrid Server

  4. #4
    Join Date
    May 2004
    Location
    Atlanta, GA
    Posts
    3,872
    in fact, I am more interested in:
    performance per watt- woodcrest 5150 is a good 43% ahead of opteron 2160

    performance per cost - this is truly a shocker:
    single woodcrest 5150 2-core has a higher MySQL performance than dual Opteron 2160 4-core! it also means you can use $700 1x woodcrest 5150 (65w) to beat up $1600 worth of Opteron 2160HE (68w+68w=136w) running MySQL

    the cost of ECC registered DDR2 modules is going thru the roof right now, so the cost disavantage of ECC FBDIMM is minimizing fast!
    Last edited by cwl@apaqdigital; 09-22-2006 at 06:41 PM.
    C.W. LEE, Apaq Digital Systems
    http://www.apaqdigital.com
    sales@apaqdigital.com

  5. #5
    Join Date
    Jun 2002
    Location
    PA, USA
    Posts
    5,143
    Good observation, Chong.

    BTW, has anybody tried 5148LV? Not much review yet. Between 5148LV and 5150, which one will you prefer? Myself, I will prefer 5148LV due to the lower wattage it consumes (40W vs 65W or 38.5% power saving). Performance per watt, the 5148LV should give a better yield than 5150 compres to Opteron 2160.

    And with power pricing increasing in major data centers now, it may become an important saving. Assuming Dual CPUs (4 cores) and everything else being equal, that will be a saving of 0.45 Amps per server.

    I am glad we have been sticking with Intel all this time, at least now we will have something to boast back to the AMD crowds
    Fluid Hosting, LLC - Enterprise Cloud Infrastructure: Cloud Shared and Reseller, Cloud VPS, and Cloud Hybrid Server

  6. #6
    Join Date
    May 2004
    Location
    Atlanta, GA
    Posts
    3,872

    Unhappy

    Quote Originally Posted by cwl@apaqdigital
    in fact, I am more interested in:
    performance per watt- woodcrest 5150 is a good 43% ahead of opteron 2160

    performance per cost - this is truly a shocker:
    single woodcrest 5150 2-core has a higher MySQL performance than dual Opteron 2160 4-core! it also means you can use $700 1x woodcrest 5150 (65w) to beat up $1600 worth of Opteron 2160HE (68w+68w=136w) running MySQL

    the cost of ECC registered DDR2 modules is going thru the roof right now, so the cost disavantage of ECC FBDIMM is minimizing fast!
    must correct myself: I do mean to say Opteron 2216 or 2216HE, not the "2160" referrenced above. getting old, must be!
    C.W. LEE, Apaq Digital Systems
    http://www.apaqdigital.com
    sales@apaqdigital.com

  7. #7
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,957
    Quote Originally Posted by FHDave
    Good observation, Chong.

    BTW, has anybody tried 5148LV? Not much review yet. Between 5148LV and 5150, which one will you prefer? Myself, I will prefer 5148LV due to the lower wattage it consumes (40W vs 65W or 38.5% power saving). Performance per watt, the 5148LV should give a better yield than 5150 compres to Opteron 2160.

    And with power pricing increasing in major data centers now, it may become an important saving. Assuming Dual CPUs (4 cores) and everything else being equal, that will be a saving of 0.45 Amps per server.

    I am glad we have been sticking with Intel all this time, at least now we will have something to boast back to the AMD crowds
    In the same vein, has anyone been able to find the lower power usage AMD AM2 processors? AMD has had them listed on their site for a bit now, and I just can't seem to find them.

    As a note, AMD still wins the performance per watt battle by a long shot on the low end systems, single core, etc. and the Core 2 Solo still seems to be a decent way off.
    Karl Zimmerman - Founder & CEO of Steadfast
    VMware Virtual Data Center Platform

    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation

  8. #8
    Join Date
    May 2004
    Location
    Atlanta, GA
    Posts
    3,872
    Quote Originally Posted by KarlZimmer
    In the same vein, has anyone been able to find the lower power usage AMD AM2 processors? AMD has had them listed on their site for a bit now, and I just can't seem to find them.

    As a note, AMD still wins the performance per watt battle by a long shot on the low end systems, single core, etc. and the Core 2 Solo still seems to be a decent way off.
    well, the "regular" AM2 chips still have 89w TPD, just about the same with s939. you need to pay extra for the "Energy Efficient" low-power version, then you can get 65w TPD. Core2 Duo (conroe) starts with 65w standard, you don't have to pay extra for it.

    granted, there is no server app benchmarks to put Conroe against AM2 because both are considered as "desktop" processors by the industry, but all the reviews I've seen do put core2 duo ahead of AM2 on desktop/workstation/gaming applications:
    http://www.anandtech.com/cpuchipsets...spx?i=2795&p=8
    http://www.tomshardware.com/2006/08/...est/page8.html
    http://www.gamepc.com/labs/view_cont...d=e6600&page=5
    I will say that the chances are slim that AM2 will atop Conroe on server apps.

    I also don't believe these heavy weight review sites have bias against AMD because they all had published very, very harsh words against Intel in the past when Intel was still in "dark" age just short while ago!
    C.W. LEE, Apaq Digital Systems
    http://www.apaqdigital.com
    sales@apaqdigital.com

  9. #9
    Join Date
    Nov 2001
    Location
    Atlanta, GA
    Posts
    633
    Interesting benchmark because they chose the chips where the Opteron starts getting it's *** kicked. Look at these results from GamePC, which show basically the same thing: http://www.gamepc.com/labs/view_cont...eon5160&page=9

    Notice that if you step down just one notch to a 5140, the Xeon starts losing like crazy. In particular, the 50,000 use benchmark show that the Opteron scales better and a 5140 would almost get beaten by a 265!

    It's that kind of stuff that indicates that more exhaustive benchmarks need to be run. I want to see the full suite of Opteron's against the full suite of Xeons. I'm not going to buy the 5160's or 5150's, so I want to see what the "common man's" CPU will do in a battery of tests. I'd also like to see something that I can reproduce myself.
    Former owner of A Small Orange
    New owner of <COMING SOON>

  10. #10
    Join Date
    May 2004
    Location
    Atlanta, GA
    Posts
    3,872
    Quote Originally Posted by timdorr
    Interesting benchmark because they chose the chips where the Opteron starts getting it's *** kicked. Look at these results from GamePC, which show basically the same thing: http://www.gamepc.com/labs/view_cont...eon5160&page=9

    Notice that if you step down just one notch to a 5140, the Xeon starts losing like crazy. In particular, the 50,000 use benchmark show that the Opteron scales better and a 5140 would almost get beaten by a 265!

    It's that kind of stuff that indicates that more exhaustive benchmarks need to be run. I want to see the full suite of Opteron's against the full suite of Xeons. I'm not going to buy the 5160's or 5150's, so I want to see what the "common man's" CPU will do in a battery of tests. I'd also like to see something that I can reproduce myself.
    historically, Opteron performs on Apache web serving exceptionally well against Xeon, but most hosts don't really use high-end Opteron/Xeon for web serving nodes, you prolly are one of those exceptions SQL or database are more likly to be running on these type servers.

    VPS is certainly another area of interests of these high-power servers. any feedback of dual Woodcrest 2.0Ghz 4-core VPS node vs dual Opteron 2.0Ghz 4-core VPS node? I know you have a few of them on both
    Last edited by cwl@apaqdigital; 09-23-2006 at 12:32 PM.
    C.W. LEE, Apaq Digital Systems
    http://www.apaqdigital.com
    sales@apaqdigital.com

  11. #11
    Join Date
    Nov 2001
    Location
    Atlanta, GA
    Posts
    633
    Haven't yet filled one of those Xeon VPS nodes yet. But the Opteron one is so underloaded (it's 0.15 now!) I'm considering buying an upgraded license to fit more VPS's on that machine.
    Former owner of A Small Orange
    New owner of <COMING SOON>

  12. #12
    Join Date
    Nov 2005
    Posts
    3,944
    On those tests the Opteron systems only have 4GB of RAM compared to the Xeons having 7GB...take that in to note too.

    but still, woodcrest looks like a good investment.

  13. #13
    Join Date
    Nov 2001
    Location
    Atlanta, GA
    Posts
    633
    Quote Originally Posted by devonblzx
    On those tests the Opteron systems only have 4GB of RAM compared to the Xeons having 7GB...take that in to note too.

    but still, woodcrest looks like a good investment.
    Good call. Why on earth did they do that, though?
    Former owner of A Small Orange
    New owner of <COMING SOON>

  14. #14
    Join Date
    May 2004
    Location
    Atlanta, GA
    Posts
    3,872
    Quote Originally Posted by Tweakers.net
    Unfortunately, stability problems with our preproduction hardware made it impossible to put more than four modules to work. We ended up testing with only 4GB of RAM, less than the other systems had at their disposal. The few results that we managed to obtain with the 8GB Socket F system allow for the conclusion that performance loss due to the lower memory capacity was about two percent, which needs to be taken into consideration when reading the diagrams.
    tweakers.net included above statement to explain why they were forced to use 4-Gig DDR2 only.

    in fact, I ran into the same issue too! on Tyan S3970G2NR board, loads of major brand (Kingston, Micron, Corsair...) DDR2 ECC reg. modules I tried all failed left-n-right. initially, I thought the issue is caused by the BIOS for new board is just too 'green'! however, it turned out to be RAM compatibility issue: thus far, only Infineon HYB18T512800AF37 1G modules can work with Opteron 22xx on-die DDR2 memory controller flawlessly and did allow me to install 8x 1G with good reliability.

    Tweaker.net also stated that 8-Gig DDR2 on Opteron F platform performs only 2% better than 4-Gig server.
    C.W. LEE, Apaq Digital Systems
    http://www.apaqdigital.com
    sales@apaqdigital.com

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •