PCI express cards & performance

Discussion in 'Mac Pro' started by spunkysayswoof, Sep 25, 2010.

  1. spunkysayswoof macrumors newbie

    Joined:
    Sep 11, 2010
    #1
    I've tried 2 different PCI express cards in a 2010 Mac Pro and have the following observations:

    Boot time increases substantially... my guess it's taking longer to identify boot disks.
    Disk performance via PCI express is worse than just plugging straight into a SATA slot: even with using a 6G PCIe eSata card to a RAID 0 enclosure.

    Moving a PCI card from the mac to a pc & performance on the pc is much higher than that on the mac/SATA slot.

    This is an expensive trial & error approach & I really don't want to spend more $'s on boards that exhibit the same problems.

    Is there a decent PCI express board that will allow for improved disk (ssd) I/O performance or is there a known issue?
     
  2. diazj3 macrumors 6502a

    Joined:
    Jan 19, 2008
    #2
    Exactly which two PCIe cards did you try? As far as I know, not any PCIe eSATA card will work with OS X as it would on a PC....
     
  3. mjsmke macrumors 6502a

    Joined:
    Mar 2, 2010
    Location:
    UK
    #3
    Ive installed a PCIe card to add 5 extra USB ports. It was a cheap £20 carp from Maplins. Works fine.

    Maybe there is a way to check the PCIe cards are working correctly?
     
  4. spunkysayswoof thread starter macrumors newbie

    Joined:
    Sep 11, 2010
    #4
    https://eshop.macsales.com/item/Newer Technology/MXPCIE6GS2/

    maxpower esata 6g pcie 2.0 controller

    http://g-technology.com/products/g-raid.cfm

    g-raid 2TB to which I've voided the warranty to open up to test & compare performance of 2 OWC Mercury pro SSD's in hardware RAID 0 through PCI express eSATA against the same 2 SSD's in software RAID 0 direct attached to internal built-in SATA connectors as well as a single SSD similarly connected.

    http://www.ocztechnology.com/produc...revodrive/ocz-revodrive-pci-express-ssd-.html

    ocz revodrive pcie ssd

    My benchmark test should be CPU bound but involves a lot of random I/O with an average of 4k units. Bottlenecks in the I/O system are easily identified by seeing the CPU utilization fall below 100%.
    A RAM disk demonstrates this perfectly.
    The SSD connected through built-in SATA approach shows some early latency & can't sustain 100% cpu utilization for more than 2-3 seconds.

    Using the PCIe SSD Revodrive CPU utilization is very erratic, never 100%, and often falls to 10%, never more than 80-85%.
    Using the MaxPower eSATA 6g PCIe controller, CPU utilization is also erratic, never over 40% when using the same SSD's, and never over 5% when using the provided Hitachi HDDs.

    Also using ATTO & CRD through Windows on the Mac Pro, or Digitools through OS X, which independently verify performance of any PCI express configuration tested is poor.
     

Share This Page