My comment really didn't have anything to do with the underlying
technology of flash cards. It's just that, when doing a write, you
first have to find an empty spot to store something. That has much
higher software and (potentially) device overhead than does just
locating where something is already stored. FAT was always pretty
inefficient to begin with but FAT32 is more so and the advent of very
large cards makes it much more likely that the necessary allocation
table data is too large to be fully cached.
In about 1986 I had developed a proof of concept for a dramatic
improvement in FAT file system performance and was going to prepare a
patent application when I discovered that a co-worker had independently
developed something very similar but his was better than mine in that
his was accomplishing similar performance gains while using less memory
(a big premium in those days). Anyhow, IBM applied for and was
eventually granted the patent but Microsoft refused to use it in DOS.
To this day I still wonder if Microsoft or any other DOS provider has
thought to fill that glaring performance hole.
Chuck Norcutt
Moose wrote:
> Chuck Norcutt wrote:
>> But one should at least be able to infer that, for a given card, its
>> read performance should be better than its write performance. Something
>> is seriously wrong with the implementation if that's not true.
>>
> Sure, if you say so. I certainly don't know enough about the technology
> to venture any opinion. Rob Gailbraith has a section about card to
> computer transfer speeds, and it is indeed true that read speeds are
> faster than write speeds. What I wrote, however, was not intended to
> make any comparison of read to write speeds, only of read vs. read and
> write vs. write of two different brands. Sorry if I wasn't clear enough.
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|