64bit amd 3700++ and they don't want me to use ecc mem -- Peter can't read

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]



On Wed, 2005-12-28 at 23:26 -0500, Peter Arremann wrote:
> Wrong. Several higher end boards for socket 939 chipsets can enable ECC. 

Peter, you have a continual, nasty habit of not reading my _entire_
post.  Revisit ...

1)  Socket-939 Opteron *DOES* support unregistered ECC ...

 "So if you want unregistered ECC, get a Socket-939 Opteron
  If you want registered ECC, get a Socket-940 Opteron
  If you want to use unregistered, non-ECC, get Socket-939 A64
 
  The BIOS/POST/ACPI is going to be so configured, respectively,
  on Socket-939 and 940 mainboards, depending on what CPU is
  inserted."

2)  I'm sure there are _some_ mainboards that support ECC on A64 ...

 "There may be "hacks" to support otherwise on some mainboards,
  but I seriously doubt it is well supported/reliable."

If AMD doesn't officially support it, please take that into
consideration.  In other words, if a mainboard sets up the APIC** and
other registers to enable ECC for an Athlon 64, be wary that AMD does
_not_ test for it.

[ **NOTE:  In my previous post, I _incorrectly_ stated ACPI when I meant
APIC. ]

> The only difference between Athlon64 and Opterons (940, not 939)
> is the registered part. 

That's _not_ a difference in processor type, but _socket_ type.

Again, there are *2* variables ...

1)  Socket-940 = registered, Socket-939 = unregistered *AND*
2)  Opteron = ECC (AMD tested) and Athlon 64 = non-ECC (typical)

Again, there _may_ be Socket-939 mainboards that will setup the APIC to
enable ECC.  But make no mistake, it's the CPU _itself_ that has the ECC
logic -- and if it's doesn't work, then you don't get ECC.

You can typically use (stability is another story) ECC memory in a non-
ECC system -- it just won't be ECC.

> MSI K8N Diamond Plus (manual page 3-11 for enabling ECC)
> ASUS A8N32-SLI Deluxe - QVL and spec sheet list both ECC and non ECC.

Yes, I've noted most of the new nForce4 SLI x16, which is basically a
commodity version of the workstation/server nForce Pro 2200+2050
combination, have this option.  That's because they are AMD tested and
approved for the new Socket-939 Opterons.

Remember, just because it is in the setup/POST, doesn't mean the CPU can
support it.  It wouldn't surprise me if all Athlon 64 processors have
ECC logic.  But whether the Socket-939 versions are tested for it is
another story.

I've only seen ECC touted on the original Socket-940 Athlon 64/FX
processors.
 
> AMD Athlon64 spec sheet lists that both ECC and non-ECC modules are supported. 

Again, be _careful_ on that statement, since there _are_ older
Socket-940 Athlon 64/FX processors.  It all depends on the
_exact_revision_ and its features.

Remember, Socket-939 is _newer_ than Socket-754 and 940.  And that
includes various changes to the core to reduce transistor count and
increase efficiency.

If you would be so kind to refer to which spec sheet/revision of the
Athlon 64, I will verify this for you.  ;->

> Most low end boards do not have a way to enable ECC. Many don't work if you 
> insert ECC modules. Most higher end socket 939 boards allow you to use ECC 

The setup/POST may limit the APIC setup of ECC.  But it's the CPU with
the actual ECC logic.  ;->


-- 
Bryan J. Smith   mailto:b.j.smith@xxxxxxxx
http://thebs413.blogspot.com
------------------------------------------
Some things (or athletes) money can't buy.
For everything else there's "ManningCard."



[Index of Archives]     [CentOS]     [CentOS Announce]     [CentOS Development]     [CentOS ARM Devel]     [CentOS Docs]     [CentOS Virtualization]     [Carrier Grade Linux]     [Linux Media]     [Asterisk]     [DCCP]     [Netdev]     [Xorg]     [Linux USB]
  Powered by Linux