I'm still bemused that so many sellers are offering refurbished and upgraded SD cards at rock-bottom prices, but are still getting away with blatant untruths - or am I wrong?
I have recently purchased a 128GB one for 6GBP which the Windows app h2testw showed having a shortfall of 7.7GB, which is a lot better than expected, but one wonders how reliable the card might be.
Thoughts anyone?
-- Mark J From RISCOS on a BeagleBoard-xM and Raspberry Pi2B - and Linux on a PandaBoard ES
I don't have any first hand experience with this but the defects on a flash chip are statistical. If they allow for say, 10 defects on the die and 95% of the time the devices meet this standard, there will be 5% that don't and end up on the grey market. I would not expect these devices to be any worse in terms of reliability. They just ended up with more than their fair share of the spot defects.
-- Rick
They'll normally get away with it -- but such SD cards could indeed be a particular problem for the Pi.
SD-cards have a processor that rapidly remaps blocks of physical memory in and out of the address space because partial-block writes can not be done to the physical memory. A physical memory block can only be written to without corruption if it is completely cleared first. The actual partial-block write process is: read from an existing block, merge with the new data, write to a cleared block, and remap that into the address space. If a card has too many physical memory defects, then it'll be firmware-programmed to be a smaller size so that it still has a sufficient margin of spare blocks that can be switched around as the data-writes occur.
It's known that some vendors (especially Chinese vendors) will "upgrade" these SD-cards back to their original size. This will *probably* be OK if the cards are used in a camera (their main use) because the writes are typically slow and are writes to whole blocks. However, these cards might not have enough spare blocks left to meet the timing requirements that the card's original designer intended, so rapid small-file (partial-block) writes might cause corruption as the cards's processor is unable to find the expected number of erased blocks for its normal operation. Thus such a card might work fine in a camera but become corrupted by the Pi, and become corrupted in a way that causes it to lose track of which physical blocks are erased, so even remaking its filesystem might not fix the corruption.
On Wed, 06 May 2015 06:49:43 +0100, Mark J declaimed the following:
There is also the matter of whether they are advertising using decimal GB or binary GB...
128,000,000,000 vs 137,438,953,472 -- a difference of 9+GB
--
Wulfraed Dennis Lee Bieber AF6VN
wlfraed@ix.netcom.com HTTP://wlfraed.home.netcom.com/
It's *always* decimal GB.
-- Rick
Only a question. Does anybody have a list of companies that actually manufacturer SD cards? I know that even if I buy something from say, Samsung, it may be a fake but at least I will cut out the risk of buying a rebranded factory second.
The label is always decimal, but Windows reports binary, hence showing apparently significantly less than advertised.
-- Alan Adams, from Northamptonshire alan@adamshome.org.uk http://www.nckc.org.uk/
Windows reports both. Choose your poison.
-- Rick
I don't know for sure, but I expect most fakes *are* factory seconds.
-- Rick
- if I, as the OP, can come back here (with thanks to rickman and all others for taking me seriously)...
I had misread the h2testw result, and had not realised how the running of the app can be upset by what else the Windows box was trying to do at the same time (The device is not ready. Code 21), so re-ran it (something like six hours to write and verify) after Windows had properly settled down. The result, hopefully not unhelpfully formatted, is below:
The media is likely to be defective.
7.7 GByte OK (16205824 sectors) 113.2 GByte DATA LOST (237576192 sectors) 0 KByte overwritten (0 sectors) 0 KByte slightly changed (< 8 bit/sector, 0 sectors) 113.2 GByte corrupted (237576192 sectors) 0 KByte aliased memory (0 sectors) First error at offset: 0x00000001ee900000 Expected: 0x00000001ee900000 Found: 0x0000000000000000 H2testw version 1.3 Writing speed: 10.4 MByte/s Reading speed: 9.14 MByte/s H2testw v1.4It means very little to me, other than the obvious that the card is defective, and is 7.7GB usueable, not 7.7GB short as I originally misunderstood, and far short of the 128GB Class 10 unbranded upgraded/refurbished as sold - and 125GB or 134GB or thereabouts decimal/binary as Windows reported.
Confusingly, to my mind, the card has 121 notionally 1.1GB h2testw files, plus a single 13.6MB one to round off with. That probably caused me to misread 7.7GB as defective/useable. The .h2w filetype appears unreadable under Windows XP, Ubuntu on the Panda, or RISCOS on this Beagle, but a cursory look at Properties suggests that the 121 are all the same size - presumably because the many errors have been padded over.
As one who started out thinking a C60 was magic, a MicroSD still seems a total hoot! What on earth will be the norm in another 30 years...
Thanks very much for the informed opinion and observations from all. I will now return to what I do understand, if I can remember what that is...
-- Mark J From RISCOS on a BeagleBoard-xM and Raspberry Pi2B - and Linux on a PandaBoard ES
Local file storage will be probably be illegal by then ;)
fruit
It looks like it is an 8GB card that has been programmed to report
128GB :(
Well, that's the upgrade. Presumably repainting the outside to 128GB was the refurbishment!
-- Mark J From RISCOS on a BeagleBoard-xM and Raspberry Pi2B - and Linux on a PandaBoard ES
Was it a Chinese Ebay vendor? :)
You can get good stuff from Chinese Ebay vendors, but I recommend only choosing those that have clocked up several thousand sales and have >99% positive feedback.
On Thu, 07 May 2015 06:53:14 +0100, Mark J declaimed the following:
And seems to have a faster write speed than read speed -- which is not normal either.
I'd also be perplexed at that "first error at...", "expected..." set... What is it doing, writing 64-bit addresses AS the data?
--
Wulfraed Dennis Lee Bieber AF6VN
wlfraed@ix.netcom.com HTTP://wlfraed.home.netcom.com/
It was an eBay seller based at Quinton, Birmingham, UK, with a +ve feedback score of 420. Only one -ve, from someone suitably disgruntled over a card that simply didn't work. I haven't left -ve feedback since eBay results in quid pro quo retailation, and my score is somewhat larger and 100% +ve, so the bad guy has won again. I'm not particularly proud...
-- Mark J From RISCOS on a BeagleBoard-xM and Raspberry Pi2B - and Linux on a PandaBoard ES
My understanding is that it writes blocks then reads them and compares. Since they are all, save the last, 1.1GB, reading them is an insuperable problem to me. The 13.6MB one is entirely zeros, however; but what continues to perplex me is how a card with 7.7GB OK can have
121x1.1GB + 13.6MB of data written on it. That's over 133GB, even if its controller has been "upgraded" to report 125/134GB of capacity! Referring back to Dave F's explanation some way back - all the h2testw playing, and formatting by SDFormatter, was on a Windows XP, not on a RPi. That formatter is supposed to be the one to use for SD, but gives no Full choice, so data is presumed to remain after formatting. Re-enter thoughts of physical memory defects, firmware programming size, and spare blocks; but no matter - the card is a fake!Thanks again all :-)
-- Mark J From RISCOS on a BeagleBoard-xM and Raspberry Pi2B - and Linux on a PandaBoard ES
I believe the retaliatory feedback on ebay has been abolished. It used to be that sellers would hold your feedback hostage so that if you gave them negative feedback they would return the favor. I'm surprised ebay let that stand for so long. But some years back they changed it so that sellers don't leave negative feedback anymore. I have had a number of issues with sellers including a couple where I have had to dispute the transaction through my credit card. None have given me negative feedback.
BTW, I don't think the "score" is so important as the percentage. The problem with low volume sellers is that the percentage can be adversely impacted by just a small number of negative ratings... or conversely look too good because the people they ticked off just haven't rated them. With higher volume the averages will even out and their percentage should be >99%... even *well* above 99%. At 99% they have ticked off 1 in 100 which I don't consider especially good odds when buying.
-- Rick
On 07/05/2015 18:17, Mark J wrote: []
Did the vendor not offer to replace the card, or to refund you?
-- Cheers, David Web: http://www.satsignal.eu
They did indeed, but not the return postage, which would have had to be Recorded Delivery. I have been doubly bitten in the past, when returned goods were stated not to arrive, and that includes 80BGP of faked 1st Class stamps so bad as to be laughable, but eBay was adamant that goods had to be returned for a refund, so they disappeared; and attempts via PayPal resulted in the payer having already closed his account. For 6GBP this card is simply not worth the hastle, but I'm sure mea culpa for not reading its description closely enough. The bottom line was 8GB to 128GB, so I have not a leg to stand on but, as with 450 other buyers, I probably felt it was worth a chance, although I wouldn't have bought it for 8GB of faulty mercandise. I was daft in any case - 128GB is an enormous amount of data to lose, so (to myself) don't be silly. Don't do it again.
An interesting and informative thread nonetheless :-)
-- Mark J From RISCOS on a BeagleBoard-xM and Raspberry Pi2B - and Linux on a PandaBoard ES
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.