Re: hdparm: invalidate: busy buffer
On Sun, Dec 08, 2002 at 06:04:11PM -0500, Neal wrote:
> Could this be a hardware or connecting cable problem? The transfer
> speeds you reported are far slower than what one might expect from a
> UDMA66 hard drive.
> My desktop has a Western Digital UDMA66 drive and a VIA 82C686A
> controller. hdparm reports:
> hdparm -tT /dev/hda2
> Timing buffer-cache reads: 128 MB in 1.15 seconds =111.30 MB/sec
> Timing buffered disk reads: 64 MB in 2.64 seconds = 24.24 MB/sec
> It doesn't matter which partition is interrogated, the average reports
> are all about the same 24MB/sec +/- a few hundred K.
Garrh. See nate's reply; he has suggested that it is due to my
chipset. So I need a new controller. Garrh. On my chipset x!=686A.
I've got an 82C596A and the other one has a big heatsink stuck to it.
The BIOS knows I have an 80-conductor cable, and seems quite happy to
set UDMA66 as far as I can tell. But I have always had to pass
ide0=ata66 to the Linux kernel to stop it coming up as UDMA33. It
would be interesting to measure the transfer speeds in DOS, but I
don't have anything that'll handle that drive in DOS, much less
produce results that can be related to hdparm (ie. no idea exactly
what it's testing).
For buffer-cache reads I get just under 50MHz, from any drive; the
UDMA66, some SCSI-1 (10MHz) drives and even a slow old WDC2850. So I
reckon this is measuring my system bus rather than the drives.