Re: Poor IDE performance with VIA MVP3

[Ryan Hayle <hackel@walkingfish.com>]

On Mon, 05 Nov, 2001 at 04:12:00PM +0000, vda wrote:
> Are you saying that hdparm -T -t is yielding wildly varying results?
> Looks similar to failing hd symptoms or bug in IDE layer causing retries
> after error/timeout. What's in the logs?

hdparm -T (buffer-cache reads) gives good numbers, 43v69 M/sec just
now.  But that's only mesuring reading from the drive's 2M cache (I
believe?).  It is the hdparm -t test (buffered disk reads) that is the
problem.  As I said, I don't receive any errors or retries
whatsoever.  Everything seems to be working perfectly, just very, very
slow...

> Well, I had problems with drives refusing to do [u]dma.
> On my home machine I found out that compiling kernel with support for VIA 
> chipset allowed udma to work ok (hdparm -T -t = ~20mb/s). Without that 
> support, my hd was stuck in pio, ~6mb/s.

That's the thing--it claims to be in UDMA mode, and again I get no
errors.  Even when I do 'hdparm -d1 -X66 /dev/hda", everything works fine,
without errors, only the speed problems persist.  Oh, and I have compiled
my kernel with VIA IDE support, it makes no difference in the performance.

It's sounding more and more like this isn't a driver/chipset problem, but
something wrong with the HD itself.  Thanks for your insight.

Ryan