OWC Sandforce SSD vs Intel X25-M G2

Discussion in 'Mac Pro' started by Tom Sawyer, May 5, 2010.

  1. Tom Sawyer macrumors 6502a

    Tom Sawyer

    Joined:
    Aug 29, 2007
    #1
    For anyone considering the OWC Sandforce based Mercury Extreme SSD, here's a quick comparison benchmark between it and a 160gb Intel X25-M G2. Nothing extensive, just a quick Xbench run, but FWIW:
     

    Attached Files:

  2. alphaod macrumors Core

    alphaod

    Joined:
    Feb 9, 2008
    Location:
    NYC
    #2
    Yes the OWC drive is faster; the Intel drive is kind of old.

    But in terms of value, the Intel is still better.

    And XBench is a terrible tool.
     
  3. frimple macrumors 6502

    Joined:
    Nov 18, 2008
    Location:
    Denver, CO
  4. Tom Sawyer thread starter macrumors 6502a

    Tom Sawyer

    Joined:
    Aug 29, 2007
    #4
    Yep, Xbench is hardly the best tool for benching these days, just thought I'd post a quick & dirty result after first testing. But, seat of the pants I feel a considerable difference in speed between the two drives. The Intel definitely gives more space for the $ of course, but so far the Sandforce drive is an amazing performer.

    My machine boots so fast that the boot animation circle never even has a chance to display!
     
  5. Salavat23 macrumors 6502

    Joined:
    Feb 7, 2008
    #5
    The only issue is that Sandforce controllers don't hold a candle to Intel in terms of reliability. Sandforce and Illindix have some over the worst reliability reputations on the market. Just look around on other forums and review sites. Intel's drives; however, have showed to be pretty much bulletproof.
     
  6. Tom Sawyer thread starter macrumors 6502a

    Tom Sawyer

    Joined:
    Aug 29, 2007
    #6
    No argument there... Fortunately owc gave a 5yr warranty on it and we know it will be quite obsolete in 3. For now I image my drive and use time machine as well so feel pretty good about the setup. So far so good.....
     
  7. VirtualRain macrumors 603

    VirtualRain

    Joined:
    Aug 1, 2008
    Location:
    Vancouver, BC
    #7
    Is this OWC drive equivalent to an OCZ Vertex 2?
     
  8. Tom Sawyer thread starter macrumors 6502a

    Tom Sawyer

    Joined:
    Aug 29, 2007
    #8
    Very similar.... The vertex 2 uses a slightly different controller the sf-1200 where this drive uses the sf-1500 which equates to more "enterprise" type features fwiw. There are some great articles/reviews on anandtech.com that detail the differences. From what I've seen they have the same performance.
     
  9. drayon macrumors 6502

    Joined:
    Aug 20, 2007
    #9
    Tom, I don't think the results are accurate form xbench. Check out the results of the uncached random read for 4k blocks and compare them to the uncached 'write' of 4k blocks. Anything look weird to you with those results? haha.

    Why not do a real world test, like clone and OS to an SSD then clone that to the other SSD, erase then reverse. Copy some large video files. a large pictures folder etc, **** that we all do every day that annoys the piss out of us how slow it is on magnetic media ;-)
     
  10. Loa macrumors 65832

    Loa

    Joined:
    May 5, 2003
    Location:
    Québec
    #10
    I'd really like to see this as well. Theoretical and benchmark throuputs are informative, but we all know how real life situations can mess these up!

    Loa
     
  11. bzshutter macrumors newbie

    Joined:
    Mar 25, 2009
    #11
    This is actually no longer true. OWC recently switched to sf-1200 controller too. There may have been some problem they found with the 1500 firmware. Anandtech had a report about that, I can't find the link at the moment. But have a look at OWC's spec sheet here: http://eshop.macsales.com/shop/internal_storage/Mercury_Extreme_SSD_Sandforce
    It's been updated to 1200 .
     
  12. Tom Sawyer thread starter macrumors 6502a

    Tom Sawyer

    Joined:
    Aug 29, 2007
    #12
    Hmm... very true, the results are a bit misleading as I look at them closer. Easy to focus on the main number, but we all agree that Xbench is worth very little these days.

    I'll try to do some more 'real world' tests and see what kind of results I get. I like the idea of timing a large video file copy, etc.

    This particular drive is one of the early ones after it was released and does have the SF-1500 controller. They did switch recently, truly the 'features' of the 1500 are nothing to get excited about really. I suspect the cost was a factor too on switching to the 1200.
     
  13. Tom Sawyer thread starter macrumors 6502a

    Tom Sawyer

    Joined:
    Aug 29, 2007
    #13
    Wow.. they have a 400gb version coming. For $1599!!! OUCH!
     
  14. TheStrudel macrumors 65816

    TheStrudel

    Joined:
    Jan 5, 2008
    #14
    How can you make that claim at this point? Indilinx, sure, but Sandforce drives are pretty new - there hasn't been time for any kind of real-world longevity testing...
     
  15. wader macrumors newbie

    Joined:
    May 14, 2007
    #15
    No problems with my Crucial F200.... 338.23 Xbench (yeah yeah, Xbench isn't reliable I hear.. can anyone recommend a better benchmarking tool for mac?)

    Really enjoying the load times of VSL Libraries & Ableton Live...
     
  16. wader macrumors newbie

    Joined:
    May 14, 2007
    #16
    ... incidentally, & for perspective, my Mac Pro with 4x 1TB Hitachis & Apple Raid card in Raid 5 scored a measly 95 for Xbench disk test.
     
  17. nanofrog macrumors G4

    Joined:
    May 6, 2008
    #17
    Unfortunately, Apple's RAID Pro card is an expensive pile of junk. :eek: :(
     
  18. NoManIsland macrumors regular

    Joined:
    Feb 17, 2010
    #18
    Wader, I run VSL and Logic, and have been debating between a RAM upgrade or an SSD to improve my VSL performance, and would like your feedback as to your experience. I do a lot of switching between two Logic projects which requires reloading of the VSL instruments every switch, which is a major annoyance. Do you think this would be better improved by more RAM or an SSD? One of the issues I would bring up about the SSD option is that my VSL library is 270+ GB, which would be damn hard to fit on an SSD. I haven't hit a RAM wall in terms of limiting my loaded instruments. My budget is $400 or less at this point. Thoughts?
     
  19. wader macrumors newbie

    Joined:
    May 14, 2007
    #19
    I'm with you on the major annoyance of load times. I have a feeling the software could be better written, although it probably has to do with the samples being compressed, and presumably uncompressed on-the-fly. I can definitely say that an SSD will speed up the load times in much more than some extra RAM.. I had 9GB RAM in my Mac Pro (sold due to relocation - currently using mid 2009 MBP) and it didn't seem to help at all. Big samples are all about disk performance, and whilst it's still not instant, it's definitely faster with an SSD.

    The problem you have is budget vs storage. I suppose using the VSL Directory manager to point at your most used samples on the SSD would help a lot, but if you're using all 270 GB, then maybe just a faster HDD. Either way, an SSD for the OS will improve the whole experience a lot!
     
  20. nanofrog macrumors G4

    Joined:
    May 6, 2008
    #20
    Your applications require random access is to load your libraries, so an SSD is the better choice.

    Given your capacity requirements, it can actually be less expensive to create a stripe set out of a pair of SSD's in order to increase the capacity. It also happens to double the sustained performance, and should give a slight improvement in random access as well (ideally this area is doubled as well, but realistically, that's very rare). Just check around, and see what the current prices are (I'm thinking you may need to increase the budget in order to create a set with that much capacity). Of course, you could also take a hard look at what's in there, and dump whatever you're not using regularly to mechanical media (put what you're using most on the SSD).

    At least you might have an option to do what you want with a little compromise or entirely with an increase in budget.

    Mechanical is far slower than SSD in terms of random access. For sustained throughputs, you can create a RAID set of mechanical drives that can outperform SSD's for less money, but it's not the best solution to the particular usage.
     
  21. macuserx86 macrumors 6502a

    macuserx86

    Joined:
    Jun 12, 2006
    #21
    Anyone care to shed some light on how OS X deals with SSDs and page outs?

    I'm going to replace my 640GB WD Black boot drive with a 100GB OCZ Vertex LE (it's the exact same thing as the OWC, but $70 cheaper) but if I do that, my Applications and system stuff take up about ~78GB, which leaves only ~22GB for VM. Is that enough space?

    Also, any other thoughts about SSDs are welcome.
     
  22. Loa macrumors 65832

    Loa

    Joined:
    May 5, 2003
    Location:
    Québec
    #22
    Hello,

    Nano will probably pop by with a more detailed explanation, but one of he weaknesses of SSDs is their relatively high write wear. They're not (yet) designed to be written to continuously like mechanical HDs.

    One of the tricks they can use to overcome that limitation is to write all over the SSD instead of always writing on the first available free space like mechanical HDs.

    While the performance of your SSD won't decrease until fill 99% of its available space, it's the long term wear problem that's going to catch up to you. How soon? I don't think that anybody knows, precisely because SSDs are still relatively young.

    The rule of thumb is that filling up a SSD is not the best long-term strategy.

    Loa
     
  23. nanofrog macrumors G4

    Joined:
    May 6, 2008
    #23
    Simple is good. :D

    OS X hasn't yet been optimized for SSD's, so they're treated the same way as mechanical disks. The differences in operation are then handled by the SSD itself (firmware and controller take over).

    So for now at least, SSD's make excellent OS/Application disks (high read environments), and mechanical should be used for high write environments (scratch and data that changes often).

    As per page outs, I presume you're referencing scratch. As Loa mentioned, SSD's aren't good for high write environments, particularly MLC based drives (SLC was designed for it, but it's quite a bit more expensive; compare the Intel X25-E (64GB) vs X25-M (80GB) to see what I mean about cost).

    So if you're using an application that requires scratch space, put it on mechanical drives (the location can be set).

    You'd want to go bigger, as you'd want to keep ~20% (10% on Intel, as it already has a hidden 10%) for wear leveling.

    One way to do this is a single larger drive, but it's also possible to make a stripe set (RAID0) of a pair (i.e. 64 - 80GB SSD's for example). You get increased capacity and double the speed.

    Just note that if you plan to run Windows, you'd need a separate card to get it to work (Disk Utility makes a change in the firmware that kills both Boot Camp and separate disk operation with the SATA ports on the logic board).

    The cheapest way to do this, is via a SATA/eSATA card. There are both 3.0 and 6.0 Gb/s cards that can boot Windows, as you only need a BIOS based card. Some even have drivers that will work for OS X, but only work once booted from another location, which wouldn't be a problem in your case.

    Hope this helps. :)
     
  24. macuserx86 macrumors 6502a

    macuserx86

    Joined:
    Jun 12, 2006
    #24
    I run windows on a separate disk, so there's no problem there.
    I was under the impression that the SSD itself keeps a certain % of the drive free (the 100GB OCZ actually has 128GB worth of NAND) for wear leveling.
    maybe I don't fully understand what wear leveling is.
     
  25. barefeats macrumors 65816

    barefeats

    Joined:
    Jul 6, 2000
    #25
    Lloyd Chambers has come up with a way to simulate a year of usage on an SSD. Turns out that all Intel and Crucial models slow down significantly. The OWC Merc Extreme, OCZ Vertex 2 and Agility 2 do NOT slow down thanks to overprovisioning.
     

Share This Page