Announcement

Collapse
No announcement yet.

uSD cards (probably applies to other, similar media, as well)

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    uSD cards (probably applies to other, similar media, as well)

    I have been unable to make any sense of the speed/performance ratings of uSD cards -- esp wrt their selling prices. What appear to be identical products from the same vendor often have different prices. Or, different markings.

    Is there a general rule-of-thumb regarding what you should be willing to pay (i.e., what level of performance you should strive for)?

    E.g., I recently purchased some 32G cards for my eBook readers with the thought of moving my entire library onto uSD cards and swapping them in/out as needed (I think 32G is the largest size they support). I am assuming performance in this application is insignificant -- read a few KB from the card, wait for user to flip the page (I don't watch movies on ereaders).

    In fact, I can't imagine how performance would be of any concern short of trying to display video, regardless of the device (e.g., why would a phone need any particularly advanced level of performance?)

    Or, is this just "marketeering" -- gold plated speaker cables, etc.?

    #2
    Re: uSD cards (probably applies to other, similar media, as well)

    performance is critical for realtime writing - recording stills or video being the main thing.

    you cant use a card thats slower than the video-stream in your car-cam for example.
    i'm sure the majority of cards are going into fones for this function.
    Last edited by stj; 10-27-2018, 06:04 AM.

    Comment


      #3
      Re: uSD cards (probably applies to other, similar media, as well)

      the transfer speed only matters if u're using it to transfer stuff from one device to another. e.g. from your computer to mobile device via a usb card reader.

      sandisk and samsung also recently came up with high write endurance micro sd cards. obviously, these cards are targeted at video recording and also cost more. i have the sandisk 32gb high endurance micro sd card for recording hdtv programmes off my digital tv set top box via a usb card reader. the card is only class 10 rated tho so only good for recording full hd resolution stuff. if u want to record 4k resolution or higher, u need uhs-1 u1 rated cards or faster.

      Comment


        #4
        Re: uSD cards (probably applies to other, similar media, as well)

        interesting, what type of price difference?

        Comment


          #5
          Re: uSD cards (probably applies to other, similar media, as well)

          Originally posted by stj View Post
          performance is critical for realtime writing - recording stills or video being the main thing.
          I wouldn't imagine it to be a problem when writing snapshots -- as I would expect the software to simply wait for each block to be written (there is no timeliness constraint involved -- beyond what the user is willing to wait).

          you cant use a card thats slower than the video-stream in your car-cam for example.
          Yes, in this case, there is a timeliness constraint -- the first video frame has to be completely written (barring any elastic store in the camera) before the second frame is available to be written. Any slower (on average) and data gets lost.

          i'm sure the majority of cards are going into fones for this function.
          So, in the usage case that I've mentioned -- loading apps, contacts, snapshots (taking and viewing), ebooks, etc. paying for speed is foolhardy; "hurry up and wait" (for the user).

          OTOH, if I design an app with an RDBMS embodied within it, then there will likely be more hammering on the medium "behind the scenes" and I may perceive a (small?) delay in the app's performance (?)

          Comment


            #6
            Re: uSD cards (probably applies to other, similar media, as well)

            Originally posted by ChaosLegionnaire View Post
            the transfer speed only matters if u're using it to transfer stuff from one device to another. e.g. from your computer to mobile device via a usb card reader.
            Well, that's where you are likely to perceive the delay in its most "isolated" form. But, it will also affect how quickly objects are loaded/stored from/to the medium.

            sandisk and samsung also recently came up with high write endurance micro sd cards. obviously, these cards are targeted at video recording and also cost more. i have the sandisk 32gb high endurance micro sd card for recording hdtv programmes off my digital tv set top box via a usb card reader. the card is only class 10 rated tho so only good for recording full hd resolution stuff. if u want to record 4k resolution or higher, u need uhs-1 u1 rated cards or faster.
            Presumably, (all of) these are MLM technology? How is the read-disturb performance reflected in their overall durability? I.e., even in a read-only application (ebooks), the FTL will still need some level of write performance.

            Said another way, how does the "lower" write endurance (of "nomal" media) affect data retention for read-only applications?

            Comment


              #7
              Re: uSD cards (probably applies to other, similar media, as well)

              what actually goes on in the card will never be known, they contain a microcontroller that handles bad-block management and raw sector allocation etc.
              your data may be neatly stacked at the start of the flash, or randomly scattered to reduce cell wear.

              Comment


                #8
                Re: uSD cards (probably applies to other, similar media, as well)

                Originally posted by stj View Post
                what actually goes on in the card will never be known, they contain a microcontroller that handles bad-block management and raw sector allocation etc.
                your data may be neatly stacked at the start of the flash, or randomly scattered to reduce cell wear.
                Yes, but the read-disturb phenomenon causes new writes to occur (by the FTL) to "salvage" data that is in danger of degrading. As the number of erase/program cycles has a relatively fixed limit, even media that is only "written" (by a user) ONCE can "wear out" as the FTL tries to compensate for this "read wear" phenomenon, over time.

                While a medium may fail after N user-writes (and be noticed by the user as having failed, at that point), I'm interested in how long the media retains data in the absence of user-writes. I.e., is the extra write endurance mentioned, above, by chaos of any real value in "read-only" applications? (because it would let the FTL recover read-disturbed data over a longer period of time)

                In the designs that I create, the (NAND) flash is only read once -- to load the application into RAM. So, read-wear is not a real issue. (and write-wear is only an issue if you are updating the firmware thousands of times!)

                Comment


                  #9
                  Re: uSD cards (probably applies to other, similar media, as well)

                  well another issue with flash is cell entrophy, and whether the controller has a way to check for it and refresh the cells!!!

                  Comment


                    #10
                    Re: uSD cards (probably applies to other, similar media, as well)

                    Originally posted by stj View Post
                    well another issue with flash is cell entrophy, and whether the controller has a way to check for it and refresh the cells!!!
                    I think, of necessity, controllers are getting cleverer at monitoring the state of flash cells. With MLC and TLC (and QLC!) technologies, its just too damn easy for data to degrade -- when you're counting HANDFULS (<< 100) of electrons to differentiate between the (various!) "states" of the cell.

                    Apparently, there is a lot of BFM involved in flash technology. If you monitor the ongoing research, folks are really delving into the technology in ways that are surprising (e.g., recovering "erased" data).

                    As "consumer" kit tends to bottom feed, I'm leary as to how much hand-waving happens with the "specs" on those products. (e.g., they're talking as few as 100 (just two zeroes) program/erase cycles for QLC (vs. 100,000 for SLC)!
                    Last edited by Curious.George; 10-27-2018, 06:45 PM.

                    Comment


                      #11
                      Re: uSD cards (probably applies to other, similar media, as well)

                      beware of fakes ..

                      Comment


                        #12
                        Re: uSD cards (probably applies to other, similar media, as well)

                        Originally posted by petehall347 View Post
                        beware of fakes ..
                        Yeah. Unfortunately, there's no way to truly GUARANTEE you're getting The Real Deal.

                        To date, I've only had one (consumer) flash device fail (a Sandisk thumb drive that suddenly "write protected" itself). But, that was replaced (Costco) at no charge. And, because it was a loss of writability (not READability), I was able to recover its contents.

                        Comment


                          #13
                          Re: uSD cards (probably applies to other, similar media, as well)

                          one test is fill will data of a known value then see if its intact on the card . many years ago before fakes were rife i went and bought a fake for lots of money .. lost some real nice pictures .. 2GB back then was massive .. £30 was a lot of money too ..

                          Comment


                            #14
                            Re: uSD cards (probably applies to other, similar media, as well)

                            Originally posted by petehall347 View Post
                            one test is fill will data of a known value then see if its intact on the card .
                            Well, you can do that with any "content". E.g., copy some files to the card; then compare them to the originals.

                            But, you can only perform that "test" after you've purchased the item.

                            So, "knowing your vendor" becomes the important question. I always felt jittery buying from DX for exactly that reason.

                            OTOH, Costco never quibbles about returns (if they notice lots of folks returning defective thumb drives, I suspect THEY will contact THEIR supplier). I assumed Amazon would behave similarly, hence made the most recent purchase through them (hard to find cards as SMALL as 32G locally).

                            many years ago before fakes were rife i went and bought a fake for lots of money .. lost some real nice pictures .. 2GB back then was massive .. £30 was a lot of money too ..
                            In hindsight, do you think YOU were scammed? Or, were the folks who sold the drive to you?

                            [We purchased some semiconductors many years ago -- through reputable disti channels. Devices didn't work. Discovered there was NO silicon inside the encapsulant! Disti embarassingly refunded our money and had to sort out the problem with THEIR source. I doubt they ever "lived it down", afterwards!]

                            Comment


                              #15
                              Re: uSD cards (probably applies to other, similar media, as well)

                              just thought i'd post some comments about both the sandisk and samsung high endurance micro sd cards since i recently got both of 'em in 32gb capacity.

                              the samsung micro sd cards seem to have more hours of write endurance than the sandisk ones and they also have a faster read & write speed rating but also seem to cost more if u refer to the pics of the brochures.

                              the samsung cards are class 10 with uhs-1 u1 support with 100MB/s read speeds and 30MB/s write speeds. they're also made in the philippines, surprisingly. i'd thought they'd be made in china. i guess china is getting too expensive for manufacturing now or it could also be geopolitics at play or both but this isnt the vip room so i wont comment further on that. 2 year warranty for the 32gb capacity is also disappointing but i'll explain later why. storage review also has a review of the samsung pro endurance 128gb.

                              the sandisk cards are class 10 only with 20MB/s read and write speeds and they're made in china as expected. like the samsung, the 32gb only has a 2 year warranty.

                              if u refer to the attached pics, the samsung one claims 17520 hours while the sandisk claims 5000 hours. if u read the fine print on the back, both manufacturers base their claims on full hd video recorded at 26 mbps (lucky for us) so based on this, we can calculate the tbw or terabytes written write endurance of the cards like on ssds.

                              so the video bitrate of 26 mbps equals a write speed of 3.25 MB/s (26÷8=3.25). since there are 3600 secs in an hour, so the write endurance rating of the sandisk would be 3.25x3600x5000=58,500,000 MBW or 58.5 TBW. the write endurance of the samsung would be 3.25x3600x17520=204,984,000 MBW or nearly 205 TBW.

                              if we look at the samsung 860 pro ssd write endurance rating, the lowest capacity of 256gb has a write endurance rating of 300 TBW. so if we divide the write endurance rating by the capacity, we can get the write endurance per gigabyte which tells us how many write-erase cycles the flash can tolerate; so for the samsung 860 pro 256gb ssd which is mlc based, 300÷256=1.171875 TBW/GB; for the samsung micro sd 32 gb, 205÷32=6.40625 TBW/GB; for the sandisk micro sd 32 gb, 58.5÷32=1.828125 TBW/GB.

                              so wow, it seems the samsung ones have an extremely high write durability per gigabyte compared to its peers. with such a high write durability, i'm tempted to wonder if the samsung pro endurance uses slc flash?

                              also, if u look at the write endurance of the other capacities of samsung 860 ssds, the write endurance per gigabyte all works out to be the same. both the 32 gb and 64 gb versions of the sandisk high endurance micro sd cards also work out to the same write endurance per gigabyte. the samsung pro endurance micro sd cards seem to have "diminishing returns" with regards to increasing the capacity but having a lower write endurance per gigabyte as a tradeoff of the higher capacity.

                              so from the above, we can surmise these about current flash memory technology. firstly, even though flash mem manufactured on a smaller process node has lower write-erase cycles, the lower write-erase cycles are offset by the higher capacities permissible by the smaller process node, thus higher capacities give higher write endurance (as in having a higher TBW limit) as well. so it is for this reason, the lower capacity micro sd cards have a shorter warranty length.

                              what im unable to find out is whether these high endurance cards use flash mem on a larger process node thereby allowing higher write-erase cycles which gives it high write endurance at a hardware level. or if they simply use highly advanced wear-leveling and special writing algorithms to maximise the write-erase cycles on the flash. in that case, u are at the mercy of how well the flash controller and its firmware are written by the manufacturer. i suspect the best solution would be both, so only time will tell just how enduring those high write endurance flash mem chips are.

                              also, high write endurance doesnt necessarily mean high reliability or it wont fail. it doesnt say anywhere in the marketing vernacular for the high endurance products that it wont fail as much or it has a longer lifespan than other flash mem products. they only say it can be written to more for more hours of video recording compared to other products. the 2 or 3 year warranty is pretty telling since other non-high endurance flash mem products also have similar warranty lengths. so like all other types of storage technology, ymmv and it depends on your luck sometimes too. hope this overview helps people decide if these high endurance products are worth the added cost...

                              if u constantly write a lot of data to your micro sd cards, then it is worth the added cost. for light heavy write users, the sandisk high endurance may be more cost effective. for extremely heavy write users, splurging on the samsung pro endurance would then be better.
                              Attached Files

                              Comment


                                #16
                                Re: uSD cards (probably applies to other, similar media, as well)

                                they could just be lying and assuming most people by the time they fail will not have a receipt or just bin them and buy something bigger (as by then the price/capacity will have shifted).

                                Comment


                                  #17
                                  Re: uSD cards (probably applies to other, similar media, as well)

                                  Originally posted by ChaosLegionnaire View Post
                                  just thought i'd post some comments about both the sandisk and samsung high endurance micro sd cards since i recently got both of 'em in 32gb capacity.

                                  so from the above, we can surmise these about current flash memory technology. firstly, even though flash mem manufactured on a smaller process node has lower write-erase cycles, the lower write-erase cycles are offset by the higher capacities permissible by the smaller process node, thus higher capacities give higher write endurance (as in having a higher TBW limit) as well. so it is for this reason, the lower capacity micro sd cards have a shorter warranty length.

                                  what im unable to find out is whether these high endurance cards use flash mem on a larger process node thereby allowing higher write-erase cycles which gives it high write endurance at a hardware level. or if they simply use highly advanced wear-leveling and special writing algorithms to maximise the write-erase cycles on the flash. in that case, u are at the mercy of how well the flash controller and its firmware are written by the manufacturer. i suspect the best solution would be both, so only time will tell just how enduring those high write endurance flash mem chips are.
                                  There are several problems trying to "intelligently" infer the quality/durability of consumer FLASH from the data provided by the manufacturer.

                                  First, there is absolutely no information provided on how the controller does (or DOESN'T!) cope with read and program disturb corruption.

                                  Second, there's no way to tell if you're seeing the "native" size of the device or some subset (e.g., in an overprovisioned case)

                                  Third, there's nothing that explains how the controller handles/reports an irrecoverable error (to the user) -- and differences between those and recoverable errors. How do I know when the device is in need of replacement?

                                  Fourth, you're never really sure whether its an SLC, MLC, TLC, QLC, etc. device hiding under the "media interface" as the interface can be gimmicked to make any sort of native device APPEAR to be any other.

                                  In my case, I use most FLASH as WORM. But, even that doesn't let me know how the device is being accessed.

                                  E.g., if I store music on the device and NEVER play a particular song, is the portion of the medium that hosts that song EVER touched? This SEEMS like it would be the case but, consider: songs are usually listed by artist, album name, genre, etc. -- i.e., things that are obtained from tags stored IN the file (NOT the file name). Does the app cache all of this information? Or, count on being able to re-extract it from the original files whenever it needs to refresh its display/catalog?

                                  If I opt to play a song, is the file read exactly once? What if I pause and resume -- possibly putting the player to sleep in the interim? Does it reread all or part of the file? Or, does it remember the offset into the file (at the nearest page/block boundary) so it can pickup where it left off?

                                  I.e., are folks writing apps using legacy notions of files (and the media on which those files have traditionally been stored)? Or, are they tuning their apps with knowledge of the actual medium being used?

                                  Comment

                                  Working...
                                  X