Kernel 2.4.2-0.1.49 detects 80w IDE cable as 40w with VIA VTC82C686A
chipset, thus setting UDMA 33 instead of UDMA 66. Kernel compiled using
VIA chipset and Use DMA by Default options.
From /proc/ide/via using Kernel 2.4.2-0.1.28:
< Cable Type: 80w 80w
< Cycle Time: 30ns 60ns 30ns 180ns
< Transfer Rate: 66.0MB/s 33.0MB/s 66.0MB/s 11.0MB/s
From /proc/ide/via using Kernel 2.4.2-0.1.49:
> Cable Type: 40w 40w
> Cycle Time: 60ns 60ns 60ns 180ns
> Transfer Rate: 33.0MB/s 33.0MB/s 33.0MB/s 11.0MB/s
I'm sorry to say this, but this is by design.
VIA is the only chipset that even offers the ability to autodetect the 80ribbon
cable, but we have disabled this in our kernel as we found that the chipset
misdetected this (and as a consequence, corrupted / crashed) in cirtain
You should be able to override this on the lilo commandline (or in a lilo
"append" line) by specifying "ideX=ata66", where X is 0 or 1 (0 is primary
cable, 1 is secondary cable). I have to warn you though. Some versions of the
VIA chipset have a bug where they corrupt data on disk or crash, even if you
do have the 80 ribbon cable.
If you do try the "ide0=ata66" option, and it does not work, please reopen this
"VIA is the only chipset that even offers the ability to autodetect the 80ribbon
Somehow I doubt it. I have an Intel 815E chipset and it happily autodetects 40-
and 80-ribbon cables. This chipset was available at least from June 2000.