ICH5 ATA DMA timeouts

[Perry Hutchison]

Interesting.  The ICH5 is part of the processor chipset, and it
could be "anywhere" on the system board, so finding/identifying
it (so as to point a fan at it, or figure out a way to add a
heatsink) may be less than straightforward -- but it's worth
keeping in mind as something to look into if problems arise.

Any idea where one finds cheap (but reliable) PCI SATA cards
these days?  It would have to be PCI -- this box does not have
PCI-E slots.

Florian Ermisch <florian.ermisch at alumni.tu-berlin.de> wrote:

> I had some onboard (S)ATA controllers becoming unreliable in the
> past. Two of them due to chipsets overheating under load which
> were fixable with additional cooling. In the third one we just
> added a cheap SATA card and spread the redundant disks between
> onboard and add-on controller to make a temporary failure a minor
> issue.
>
> Regards, Florian
>
> Am 6. Dezember 2015 22:21:35 MEZ, schrieb Steven Hartland <killing at multiplay.co.uk>:
> > Check cables and devices are in good condition. These things are
> > usually a connectivity issue or device failing
> > 
> > On 06/12/2015 03:06, Perry Hutchison wrote:
> > > Does anyone know the condition of the ICH5 ATA support in FreeBSD
> > > 10?
> > >
> > > In preparing to repurpose an elderly Dell Dimension 4600 from
> > > Windows to FreeBSD, and needing to decide what to do about
> > > drives, I found several mentions in the archives* of ICH5 ATA
> > > DMA timeouts -- mostly affecting the SATA ports, but the
> > > prevalence of SATA reports may just indicate which ports were
> > > getting the most use:  a couple of the reports involved the
> > > PATA ports.
> > >
> > > While there have been commits to the ATA code since then, I didn't
> > > find any definitive statement that the DMA timeouts had been fixed.
> > > Did I miss something, or would I be better off using a separate SATA
> > > or PATA PCI card instead of the ICH5's built-in ports?