The point *I* at least was trying to make is that if you take a hundred (S)SDs from the same manufacturer and technology type, and subject them to identical conditions - temperature, and write cycles., they will fail in similar time-scales, and that is *not random*.
Random is when they just fail at totally arbitrary times.
Its not a question of prediction, its a question of statistics. Random events have a different shape of histogram to non random events.
-- Truth welcomes investigation because truth knows investigation will lead to converts. It is deception that uses all the other techniques.
Hi,
On 31.01.2016 11:42, Dave Liquorice wrote: ...
:-)
...
Mine is still working.
It seems a little bit strange to me that SD cards are dying after such a short usage.
Mine (Raspberry B) does exactly the same job: one picture every 30s (each approximately 2.9 MB). Plus: for each picture taken it calls 'convert' to downscale the image and stores the scaled image on the SD as well.
I just took a look at the shell script that runs the above mentioned actions triggered by a cron job. I installed it on 18 October 2014 (sic!) Since then this routine is running day and night.
The (full size) SD card I use is a 32 GB class 10 (200x) Lexar.
I dont think the SD card's brand makes the difference. I own another Raspi running about the same time housing a high frequented JBoss server which accesses a MySQL database running on the same machine. Not at all optimized to do less writing on the SD card. This one uses a micro SD card from SanDisk - no problems.
My experience: SD cards are not so bad for this purpose.
Regards Ronald
That may only seem to be two write operations, but depending on how the software works, it may involve a few dozen writes to the disc meta data, which is the same part of the disc each time.
---druck
Shouldn't the wear levelling algorithms move it around occasionally?
I was talking about this last night to an old friend from college who actually design(a/ed) the memory chips for these things.
There is of course a huge amount of subsurface error correction and 'bad block' switch-out going on, as well as caching with maybe DRAM on the SSDs at least to minimise the need to access the flash memory too often.
It's a steady development cycle with better manufacture, better design, better algorithms and so on all playing a part, to hide the fact that a given memory cell only has a certain number of writes before it dies.
And the pace of development is fearsomely commercial he said. its weeks, not months. So what was true 6 months ago my be less true today.
What this means is that the question "When does a program write to the SD card" does not actually equate to "When does a program write to the SD card memory"
And its hard to say how many low level writes it actually results in.
-- Canada is all right really, though not for the whole weekend. "Saki"
If there is any wear levelling. On fast cards just designed to store large files until full, then erased, there may be very little wear levelling. On cards designed for smaller writes, there may be more.
Unfortunately there is no way to tell which sort of card you have. The class 10's are certainly fast, but the lower classes may be designed for better small writes, or they could just be less fast, using old tech.
The only thing I can recommend is find a brand which gives good performance and lasts a reasonable amount of time, and stick to that. (I use Samsung EVO branded SDXC's).
But don't expect any card to last forever, in fact them expect to fail regularly. So buy several and rotate out your main card to be a backup after a period, say 6 to 24 months depending on how heavily used.
---druck
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.