On Wed, 17 May 2006 15:22:59 GMT,
(E-Mail Removed) (Bob) wrote:
>On Tue, 16 May 2006 16:52:56 -0400, kony <(E-Mail Removed)> wrote:
>
>>The more foolproof way to do it is to use a compactflash
>>card and CF-IDE adapter. That will work with just about any
>>motherboard because compactflash was made to work in ATA
>>mode. Many if not most CF cards may use PIO mode instead of
>>UDMA though, which is slower and uses more CPU time. I
>>believe Sandisk's higher-end cards are UDMA capable
>>(probably others now too but you'd need to see that as a
>>feature of the card to be sure), yet an UDMA supportive
>>CF-IDE adapter must also be used.
>
>Will this mode preserve the fast characteristics of flash memory that
>you realize on USB?
It still has low latency, "seek" time. PIO mode will be
slower than USB2 if the system's USB2, and the flash disk's
supportive USB2 bridge chip, are good. Some are better than
others, and none are particularly good at USB2 with many
small files which is ironic since that is exactly where
flash memory's low latency should really shine.
CF using UDMA should be better than USB2, but I don't have
any of the newest/fastest CF cards on a UDMA capable adapter
so it is only theoretical.
>
>>It depends on what you want to boot too. Flash drives may
>>have a removable media descriptor that changes how windows
>>handles them. Certain IDE drivers may not work with them
>>(for example, when I had a thumbdrive booting Win2k, it
>>wouldn't boot (finish booting) if the nVidia nForce2 IDE
>>driver was installed instead of the MS IDE driver. DOS is
>>dumber though, no such issues there.
>
>I am just thinking about my next machine which is a ways off. I would
>like to use non-volatile memory for the two main disks in my setup -
>the boot disk and the backup disk. I will periodically clone these to
>hard drives for archive purposes.
One of the problems with flash for the OS is what OS we
often use, Windows. Windows incessantly writes to the drive
even if the swap file is turned off or delegated to a
ramdisk. Other applications have an annoying tendency to do
so too until one manually disables certain features (if it
is possible to do so). For example TV tuner cards, the
moment you load up the application it may then be writing to
the HDD every X # of seconds to support a replay feature.
Other apps may write less often but still quite a bit, like
editing applications with undo features.
I am wondering why you would use the flash for the main
drives then "periodically clone these to hard drives for
archive purposes". It seems at least as useful to do it the
other way around, use the hard drives for the OS and clone
the important data to the more reliable but more limited
write cycle storage, the flash memory.
>
>I believe flash memory is coming down in price such that you could use
>a 32GB flash drive. I keep my drives very clean - no more than 15GB on
>the two mentioned above. Therefore when 32GB flash memory comes down
>to $100 per stick, I might want to give it a try. By then I am sure I
>can find an appropriate mainboard. However if CF-IDE is as fast or
>faster than USB, or is better overall, then that would be the way to
>go.
If considering speed then you will have to see benchmarks of
the particular product you are considering, to make a good
choice. Performance can vary considerably with the flash
drive or CF cards themselves, though a CF-IDE adapter (one
that supports UDMA) is a fixed entity, only a pin-adapter
that will be no faster or slower whether you picked up a
generic on ebay or something more expensive from a US
company... the latter would probably cost about 4X as much.