poll: problems with MTRRs?

[Lennart Sorensen]

On Tue, Mar 17, 2009 at 03:44:07AM -0400, D. Hugh Redelmeier wrote:
> | From: D. Hugh Redelmeier <hugh-pmF8o41NoarQT0dZR+AlfA at public.gmane.org>
> 
> | Somehow Intel didn't notice when they cloned the architecture so the
> | P4 didn't have an IOMMU.  I presume they fixed this for Intel Core.
> | But where?  The memory controller is on support chips, not the CPU
> | chip.
> 
> I'm having trouble finding out about VT-D, Intel's name for this
> feature.  Specifically: what chips support it.

Seems the answer is something like:
Q35, Q45, 5400XS.  Maybe a few more.

Apparently the X38 and X48 are also able to do it although many boards
don't enable it from the BIOS.

> According to this document it is supported in the 82Q45 Graphics and
> Memory Controller Hub (GMCH) and not the other ones described there.
> Furthermore, there is a bug that can only be mitigated by not using
> VT-D.
>   http://www.intel.com/Assets/PDF/specupdate/319971.pdf
> 
> It seems to also be supported by the 82q35 according to
>   http://www.intel.com/design/chipsets/datashts/316966.htm
> This is also the only family member that supports Intel Trusted
> Execution Technology and Intel Active Management Technolgy, whatever
> that is.
> 
> So my 4 year old Athlon 64 systems have an IOMMU but my six month old
> Core2Quad system does not (it has an 82q33 chip).

That would be correct.  My P35 certainly doesn't have it.

Now if your devices are 64bit address capable, then you don't need it.
So AHCI controller is fine, but many PCI cards are not.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists