SD Card speeds

Discussion in 'Accessories' started by jlabate, Feb 12, 2012.

  1. jlabate

    jlabate Mu-43 Regular

    132
    Nov 22, 2010
    Why does my SanDisk Ultra SDHC card say (printed on the card) it's Class 4 and 15MB/s? I thought class 4 is 4MB/s (and class 6, 6MB/s, class 10, 10MB/s, etc.)
     
    • Like Like x 1
  2. lenshoarder

    lenshoarder Mu-43 All-Pro

    Nov 7, 2010
    The class rating is the minimum. I have some class 6's that hit 19MB/s.
     
    • Like Like x 1
  3. jlabate

    jlabate Mu-43 Regular

    132
    Nov 22, 2010
    Thanks. And by the way, how do you measure your card speeds?

     
  4. lenshoarder

    lenshoarder Mu-43 All-Pro

    Nov 7, 2010
    Crystalmark is good.
     
  5. addieleman

    addieleman Mu-43 All-Pro

    Aug 5, 2010
    The Netherlands
    Ad
    A simple approach is to measure the time it takes to copy a 1 GB file to the card (write speed) and copy it back (read speed, presumed that your PC's hard disk is fast enough).
     
  6. lenshoarder

    lenshoarder Mu-43 All-Pro

    Nov 7, 2010
    Something like Crystalmark would give you much more comprehensive results. Copying a 1 GB file would give you the write speed for large sequential files, what the class rating measures, but it wouldn't tell you what the random I/O is. It's not uncommon that a 20MB/s sequential card will give you 0.006MB/s random speeds. Why is this important for this forum? A high sequential card is great if you are recording video, but if that same card has low random writes then it may not be good for jpeg stills that are small.
     
    • Like Like x 1
  7. punkman

    punkman Mu-43 Regular

    151
    Dec 30, 2011
    Europe
    I bought some of the fastest SanDisks (45mb/s and 95mb/s) but then I realized I probably won't have a device that can take advantage of them for a few years. Might as well have bought some slower ones, but Amazon did have a good price for the fast stuff.
     
  8. meyerweb

    meyerweb Mu-43 Hall of Famer

    Sep 5, 2011
    I believe the "class" rating is based on write speed, while manufacturer numbers might be anything, since there's no real standard. But I think it's safe to assume that manufacturers will tout read speed, which is always faster than write speed.
     
  9. lenshoarder

    lenshoarder Mu-43 All-Pro

    Nov 7, 2010
    Class rating is write speed. There's no rating for read speed. If you flash doesn't get 15-20MB/s read then it's seriously defective. Any flash should do that.
     
  10. lenshoarder

    lenshoarder Mu-43 All-Pro

    Nov 7, 2010
    A hacked GH1 will do that for video.
     
  11. punkman

    punkman Mu-43 Regular

    151
    Dec 30, 2011
    Europe
    Good thing I bought a hackable GH1 then. :)
     
  12. adelorenzo

    adelorenzo Mu-43 Regular

    113
    Nov 9, 2011
    Whitehorse, Yukon
    Class ratings and manufacturer's claimed speeds seem to vary widely, but they don't tell you how fast a card is. You can check all the benchmark ratings on a site like Tom's Hardware.

    Benchmarks 2011 SD Cards
     
  13. lenshoarder

    lenshoarder Mu-43 All-Pro

    Nov 7, 2010
    Dude, the fastest card I've ever bought was one of the cheapest. A $9 8GB special at Target. 120mb/s. It's rated a "class 6".
     
  14. punkman

    punkman Mu-43 Regular

    151
    Dec 30, 2011
    Europe
    That might be because a few workers stayed late at the factory, pulled out the rejects from the trash and sold them very cheaply to someone. Or a mistake during the binning process. Or just a matter of wrong labeling. Or who knows what.
     
  15. oldracer

    oldracer Mu-43 All-Pro

    Oct 1, 2010
    USA
    Probably not. Speed markings are not determined solely by how fast the part is. Here is how it usually works:

    At final test, the parts are binned according to speed. If the company can sell all the parts in the "fastest" bin at full price, then that's what happens. If not, then the excess is marked at a slower speed -- one where the market can absorb the parts. This is regardless of the actual speed of the parts. It is a market strategy decision designed to extract the largest number of margin dollars from the range of parts being produced. So a very fast part might be marked as very slow if that's the only speed level where it can find a market. To offer it as a fast part at a lower price (to move it) would drive down the price of all the fast-marked parts and may not result in margin dollar maximization.

    During the life of a product, it is common to do "shrinks." This is just what it sounds like; the part is made smaller by some percentage. This increases leverage (the number of dice per wafer) and it improves performance. It is mainly done as a cost-reduction, since the cost to process a wafer is roughly the same regardless of how many dice are on it. But each shrink results in more and more parts ending up in the "fast" bin at final test. This also increases the likelihood that some, maybe all, of the parts marked with slow speeds (and sold at low prices) really being faster than marked. It is all about marketing and maximizing margin dollars.
     
  16. D@ne

    D@ne Mu-43 Top Veteran

    593
    Feb 23, 2012
    Toronto
    Gah...why didi I not know about this sooner!! :mad:
     
  17. ean10775

    ean10775 Mu-43 All-Pro

    Jan 31, 2011
    Cleveland, Ohio
    Eric
    Right now I'm using a Transcend 8GB Class 6 card and the write time from the camera to the card seems ridiculously slow - sometimes its quick but sometimes its between 5-10 seconds for a single image. That said, I am shooting Large JPEG+RAW. I haven't tested any other cards so I'm not sure if its the card I'm using or just a limitation of the camera. From other E-PL1 users, is what I'm experiencing typical or should I be looking for a faster card?
     
  18. c5karl

    c5karl Mu-43 Regular

    144
    May 31, 2011
    Fairfax, Va., USA
    There's a real world question implicit here that I haven't seen clearly answered anywhere. Right now I'm using store-brand Class 10 cards I bought at Micro Center because the price was right and I've had good experience with reliability of these cards over the years.

    When shooting RAW in burst mode on a G2, I'd love to be able to get more than 4-5 frames before shooting slows to a crawl. I haven't seen anything that clearly states whether I could do better with another memory card.

    Two areas that aren't clear to me:

    1) There's no shortage of available benchmarks. Which one matters for this use case?

    2) Once I do identify the fastest possible card, will it matter? Or is there a bottleneck upstream of the memory card that's limiting write speed on a G2?

    I've come close to saying, what the heck, I'll just get one of those snazzy Sandisk Ultras and see what happens, but it seems like there are seven or eight different kinds of Sandisk Ultras to choose from. So I just throw up my hands and stick with what I've got.

    Any real-world experience about whether better burst performance is really possible?
     
  19. fredlong

    fredlong Just this guy...

    Apr 18, 2011
    Massachusetts USA
    Fred
    If you really mean 120Mb/s, that's 15MB/s, about normal for a class 4 card.
     
  20. lenshoarder

    lenshoarder Mu-43 All-Pro

    Nov 7, 2010
    15MB/s is not normal for a class 4 card. I think you are confusing read speed with write speed. All cards, regardless of class, should get between 15-20 MB/s reads. Writes are a different thing altogether. Normally, a class 4 card does 4 MB/s writes.