On 3/23/2013 1:42 AM, Moose wrote:
> 1. I use until full, then erase or format. If I change cards before one is
> full, say to be ready for a long day of
> shooting that likely won't fit on the card in the camera, the process is the
> same, full clearing of files before re-use.
> That shouldn't cause this problem even if I only erase. Quite different case
> than lots of random erasing and writing.
>
> 2. I was under the impression that the FAT the system/camera sees is virtual,
> as the on-card controller moves stuff
> around for load leveling. That means any fragmentation in the FAT is
> illusory, and actual fragmentation is invisible.
>
> 3. I'm not sure sequential reads are meaningfully quicker than random reads
> or even quicker at all. The random access
> test in HDTune just shows a flat line for flash cards. There's no mechanical
> component, just memory addressing - and, as
> above, that access is always through an internal lookup (different FAT)
> hidden to the outside world.
>
> Am I wrong anywhere here? I know it's really easy to forget the load leveling
> isolation of internal and external file
> structures on these cards. The look exactly like physical HDs to our
> computers.
Yes, you are wrong here. I think you just haven't thought through the
problem from the camera's perspective rather than the flash card's
perspective. Although flash cards do use load leveling firmware to
determine the actual storage location of a device defined "block" it
makes no difference to the FAT software in the camera. The FAT data
structure is still the FAT data structure and works in units of
"clusters". As you say, the actual storage location on the card is
virtual. But the camera's software doesn't know that.
From the camera's standpoint fragmentation occurs in the FAT data
structure as it always has. The camera is still faced with the
complexity of managing a fragmented structure although, as you suggest,
the performance implications are not the same as on a physical disk
drive. NAND flash memory devices (the type we're familiar with) are not
truly random access like RAM but, without the mechanical delays of a
disk, there may be no observable performance hit.
But you are perfectly correct that filling the card and then erasing all
or reformatting avoids any possibility of fragmentation.
>> Dr (Flash) Disk always advises formatting in camera vs erasingimages.
>> He also recommends an occasional full format in the computer since it
>> will verify correct operation of all parts of the card.
>I don't believe that is necessarily true of a load leveling flash
>card. For all we know, the full formatting could be
>writing/reading more than once to some of the same physical locations
>and never writing to others.
>It's the job of the card controller to keep 'wear' even and note
>memory locations that are failing.
When the card is formatted with a full format the entire address space
will be written. Regardless of how the load leveler does its work it
must still find and write to a unique storage location equivalent to the
full capacity of the device. Whether hard disk or flash drive the drive
controller may choose to substitute a spare storage area for one that is
failing but that's of no short term consideration to you. You will
still have written to and verified the operation of the device across
the full range of storage locations addressable by the FAT.
>> Follow that with format in camera to install folders the camera is
expecting to see.
>
> This is another safety rule that I think is no longer necessary.
>
> These same ideas live forever on the web - check the dates of advice!
- and get repeated endlessly by those who don't
> know any better and want to be beyond failsafe.
Agreed, I do know and understand this but remain
Dr. (beyond failsafe) Flash
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|