SQL Server Performance

Server Compare

Discussion in 'SQL Server 2005 Performance Tuning for Hardware' started by rhunt, Feb 1, 2007.

  1. rhunt New Member

    I am about to buy a new box and I am trying to decide between two base configurations that are essentially the same price:

    HP ProLiant DL380 G5 - 3.00GHz Rack Server - 1P
    HP ProLiant DL380 G5 3.00GHz Server
    Dual Core Intel® Xeon® 5160 (3.00GHz, 1333 FSB) Processor

    HP ProLiant DL380 G5 - 2.00GHz Rack Server - 1P
    HP ProLiant DL380 G5 2.00GHz Server
    Quad-Core Intel® Xeon® E5335 (2.00GHz, 1333 FSB) Processor

    4 cores might be really nice considering I would still only have to license SQL 2005 for one processor. This is doubly true since this box will be mixed use and will have a small degree of SSRS load.

    However, I'm still skeptical about 4 cores on one chip and that 3GHz clock speed is calling my name.

    Anyone have any thoughts or real word compares?
  2. bradmcgehee New Member

    My personal preference is to go with the Quad-Core. But hopefully Joe Chang will respond to this, as he is the expert in this area on this website.

    Brad M. McGehee, SQL Server MVP
  3. joechang New Member

    it depends on the application
    most are not properly tuned for >2 cores,
    that is, many apps show decent gain from 1 to 2 (1.7X), moderate gain from 2 to 4 (1.5X), then negligle thereafter

    so it depends on whether your app was properly designed for a high number of cores
    if not, giving up too much frequency is not desired
  4. rhunt New Member

    So, since we are talking about SQL Server here - which is pretty darn good at using SMP, I would think the 4 core might be desirable. Again, this is going to be a mixed use box... I would say 80% straight SQL load via ODBC/OLEDB, 15% SSRS, 5% SSIS.

    Despite what I've said above, I am still leaning toward the 3ghz dual core...
  5. rhunt New Member

  6. joechang New Member

    the SQL Server engine is capable of getting good SMP scaling
    but the SQL code in your app may not
    writing SQL blind to the characteristics of how the engine handle each operation does not give any certainty, in fact, for some reason, people usually chose the SQL constructs that results in poor SMP scaling

    otherwise i would not have mentioned this
    be carefully about applying benchmark scaling results and your application
    the benchmark was tuned by the best experts in the world,
    for that matter, the SQL engine itself was also built to run the benchmark
    the same does not apply to your app

    per what i said in my post on this matter,
    i do not think i would drop from dual core 3.0GHz to quad 2.00GHz even though benchmarks show better performance for the quad at the same price
    if you want to go quad, spend the extra money for the 2.66
  7. rhunt New Member

    Great. Thanks Joe.

    This server is not going to have an I/O subsystem commensurate with the quad core 2.66Ghz chip...so no need to spend money on processors I won't be able to max out. I am going to stick with the dual core 3.0 Ghz.

    Thanks again. RH

Share This Page