Archived from groups: comp.sys.ibm.pc.hardware.storage (
More info?)
"Andrew Rossmann" <andysnewsreply@no_junk.comcast.net> wrote in message news:MPG.1b5399a9652114b2989a12@news.comcast.giganews.com
> In article <ccbqp1016cs@enews1.newsguy.com>, ericgisin@graffiti.net
> says...
> > "Andrew Rossmann" <andysnewsreply@no_junk.comcast.net> wrote in message
> > news:MPG.1b5322ffa94ac1c989a10@news.comcast.giganews.com...
> > >
> > > A CRC is computed for each packet of data. If it doesn't match, an error
> > > is flagged and passed to the software. I believe it's up to the
> > > software/drive firmware to figure out what to do next (usually retry a few
> > > times.) I believe UDMA33 and higher have more aggressive error correction
> > > and retries.
> > >
> > All UDMA modes have CRC, and PIO/DMA have none.
>
> I guess it depends on what level everyone is talking about.
Simple: "Does the hardware involved check for errors during *data transfer*?"
> There has always been a CRC stored on the drive
Nope.
> and it's checked when it reads the data.
That's an ECC.
> If there are too many retry failures, it passes a failure code to
> the controller. Back in the MFM/RLL/ESDI days, that was done on a card
> separate from the drive, so you did sort-of check the integrity of the
> signal in the cabling. With IDE, it's all built-into the drive, so before
> UDMA33, it was assumed the data traveling on the cable always made it
> through with no error. It's also why the 80-wire cable was required for
> UDMA66 to give better shielding.