Nehalem TPC-E results - 4S x86 is dead

By: Anders Jensen (.delete@this..), March 31, 2009 10:36 pm
Room: Moderated Discussions
anonymous (no@spam.com) on 3/31/09 wrote:
---------------------------
>>4S Nehalem should have a fair chance against 2S 12C 32nm Westmere due to FBD memory
>>interface. DDR3 has a real bandwidth*capacity issue for servers.
>
>You do know that Beckton will connect to Millbrook chips
>which in turn will connect to DDR3 DIMMs, right?
>
>http://pc.watch.impress.co.jp/docs/2007/1023/kaigai396_02.pdf
I am aware of that. The Millbrook compromise between pure FBD2 architecture and pure DDR3 architecture is quite OK although I'm not sure it is really worth it. It would actually made a lot more sense for DDR2 than for DDR3. Sure power and cost will go down a bit and so will capacity. The big deal however is to have a high BW/pin, easy to route and farther than DDR3 reaching interface going through the socket. If the buffer is on the mainboard, on a memory card or on the DIMM is not so important and can be tailored by the server vendor anyway.

Most important of all, it "solves" the problem as Wes pointed out.

One thing I haven't seen discussed though. Since FBD buffers can be chained together wouldn't it also be logical if Millbrook was designed the same way?
< Previous Post in Thread 
TopicPosted ByDate
Nehalem TPC-E results - 4S x86 is deadMichael S2009/03/30 09:52 AM
  SPECjbb and SPECpowerHenrik S2009/03/30 10:20 AM
    SPECjbb and SPECpoweranon2009/03/30 10:41 AM
      SPECjbb and SPECpowerHenrik S2009/03/30 11:12 AM
  Nehalem 2S Jbb2005 result - lower than 4S ShanghaiMichael S2009/03/30 10:43 AM
    3S Jbb2005 result, for funEduardoS2009/03/30 03:04 PM
      3S Jbb2005 result, for funMichael S2009/03/30 03:42 PM
        3S Jbb2005 result, for funmpx2009/03/31 03:15 AM
          3S Jbb2005 result, for funMichael S2009/03/31 10:12 AM
            3S Jbb2005 result, for funmpx2009/03/31 11:00 AM
  Nehalem TPC-E results - 4S x86 is deadLinus Torvalds2009/03/30 01:58 PM
    Nehalem TPC-E results - 4S x86 is deadPaul2009/03/30 03:40 PM
      Nehalem TPC-E results - 4S x86 is deadhpc guy2009/03/30 04:46 PM
        Nehalem SPECint/fpDavid Kanter2009/03/30 05:04 PM
  TPC-CMichael S2009/03/30 03:20 PM
  Nehalem TPC-E results - 4S x86 is deadAnders Jensen2009/03/30 11:12 PM
    Nehalem TPC-E results - 4S x86 is deadanonymous2009/03/31 08:16 AM
      Nehalem TPC-E results - 4S x86 is deadWes Felter2009/03/31 11:28 AM
      Nehalem TPC-E results - 4S x86 is deadAnders Jensen2009/03/31 10:36 PM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell avocado?