> Kris, I am somewhat confused now. What you say makes perfect sense > except one detail. I thought that CF cards were those that had > controller on board. The SD cards as I understand don't have controller > on board. Therefore it makes certain sense (may be not too much sense, > but still) to write to/format the card in the same controller (the > camera). I am not sure if reading from the card can actually damage its > contents...
SD cards have controllers inside as well. This is "hard data" from an Apacer datasheet of 2005: "The SD Memory Card includes an intelligent controller that manages interfaced protocols and data storage and retrieval as well as Error Correction Code (ECC) algorithms, defect handling and diagnostics, power management and Content Protection for Recordable Media related functions". So using SD cards you are essentially shielded from hardware failures of the storage by that controller. What it should be able to do is relocate data even for bad writes and even relocate file system records - the most write-intensive parts of the memory. Whether a card fails or not is directly dependent on how this controller operates and what failures it is programmed to circumvent. As all those safeguards potentially take processing time on the card, expect high-speed cheap cards to have suckier controllers that don't perform 100% on-the-fly checks. Now as this is a market economy we are talking about, there are bound to be cheaper less complicated (or just older) controllers out there that the cheaper cards use, with luck that should never be a problem for an end user but if that controller is slower and there are errors to be corrected and the camera cuts power to the card too quickly when powered off - anything could happen. I am using SSDs in all my computers and there the problem is way more pronounced because of the frequent random writes. It happens that a drive is put on market with defective firmware and because of the frequent writes the users see the problem already in a couple months. SSDs usually have user-upgradeable firmware that can at least partly solve the problems. SDs don't have user-upgradeable firmware so if you put a substandard card on the market users will probably start experiencing issues in a couple years - when nobody can do a thing about it. > Another question I'd like to ask - how many read/write cycles there has > to be made before a certain location on the card becomes flaky? I mean > what is card's MTBF? You see, I still have that 1GB SD card (SanDisk) > that I bought back in 2006 that still works. My empirical understanding > is that several tens of thousands of read/writes don't have significant > influence on the card performance. Controllers in modern cards are designed so as to balance writes across the card. That coupled with good error detection and correction routines should make the card last forever under normal load. That is assuming "normal" error rate. Could happen that the memory on the cards is produced from crappy materials or shipped with some obvious faults (like the first batch of K-5 sensors) - that makes this discussion a purely theoretical one, we have no knowledge of what quality materials are used for which cards > Boris kris -- PDML Pentax-Discuss Mail List PDML@pdml.net http://pdml.net/mailman/listinfo/pdml_pdml.net to UNSUBSCRIBE from the PDML, please visit the link directly above and follow the directions.
