Netserver Memory

[gael]

Archived from groups: comp.sys.hp.hardware (More info?)

Hello,
We have a HP Netserver E800 with 512 Mb of RAM.
We want to upgrade it to 2 Gb.
Due to financial problems, we want to know if we could use SDRAM non ECC to
upgrade it or, is it obligatory to use ECC SDRAM ?

Thanks

Gael

 

[Guest]

Archived from groups: comp.sys.hp.hardware (More info?)

ECC is mandatory in nearly all HP servers. The signalling characteristics of
ECC memory are usually different, even though the DIMM socket is physically the
same... Ben Myers

On 23 Nov 2004 08:38:35 GMT, Gael <gael@lemouhaer.net> wrote:

>Hello,
>We have a HP Netserver E800 with 512 Mb of RAM.
>We want to upgrade it to 2 Gb.
>Due to financial problems, we want to know if we could use SDRAM non ECC to
>upgrade it or, is it obligatory to use ECC SDRAM ?
>
>Thanks
>
>Gael

 

[Guest]

Archived from groups: comp.sys.hp.hardware (More info?)

ben_myers_spam_me_not @ charter.net (Ben Myers) wrote:

> ECC is mandatory in nearly all HP servers. The signalling
> characteristics of ECC memory are usually different, even though the
> DIMM socket is physically the same... Ben Myers

That's not quite correct. ECC memory uses the same signalling spec like
non-ECC memory. All chipsets which support ECC also work with non-ECC
memory. If a system works with non-ECC memory is a matter of the BIOS. If it
can disable ECC verification the server will work with non-ECC memory as
well...

On the other side, there is a reason why servers use ECC. The OP should
think twice about if sacrifying some safety for a few bucks is worth it...

Benjamin

 

[Guest]

Archived from groups: comp.sys.hp.hardware (More info?)

There IS a signalling difference between registered and unregistered DIMMs, also
between buffered and unbuffered. ECC DIMMs are most often registered and
buffered.

Barring information to the contrary, the E60 BIOS would require ECC. If it was
my server brand, not HP's, I would require it, and use of anything less goes
against the design criteria for high reliability.

This is a bit better than the old days when early HP EDO ECC DIMMs had an extra
PAL to identify them as HP-branded. If you put in a non-HP DIMM otherwise
meeting the server's memory specs, the BIOS would complain loudly about the
non-HP memory it found, but fortunately the server would then boot up and run
without incident... Ben Myers

On Tue, 23 Nov 2004 16:50:22 +0100, "Benjamin Gawert" <bgawert@gmx.de> wrote:

>ben_myers_spam_me_not @ charter.net (Ben Myers) wrote:
>
>> ECC is mandatory in nearly all HP servers. The signalling
>> characteristics of ECC memory are usually different, even though the
>> DIMM socket is physically the same... Ben Myers
>
>That's not quite correct. ECC memory uses the same signalling spec like
>non-ECC memory. All chipsets which support ECC also work with non-ECC
>memory. If a system works with non-ECC memory is a matter of the BIOS. If it
>can disable ECC verification the server will work with non-ECC memory as
>well...
>
>On the other side, there is a reason why servers use ECC. The OP should
>think twice about if sacrifying some safety for a few bucks is worth it...
>
>Benjamin
>

 

[Guest]

Archived from groups: comp.sys.hp.hardware (More info?)

The converse is definitely not the case. I acquired some new ECC registered
buffered SDRAM DIMMs once in an auction lot, all good memory, tested OK in a
server motherboard. When they were installed in an equally good quality
Intel-branded desktop board, the board would not boot at all.

If the differences between ECC and non-ECC are not due to signals, they are due
to timing. Or the timing of signals, hence my assertion about different
signalling. Different timing of signals or different signals? The difference
is largely one of semantics... Ben Myers

On Tue, 23 Nov 2004 16:50:22 +0100, "Benjamin Gawert" <bgawert@gmx.de> wrote:

>ben_myers_spam_me_not @ charter.net (Ben Myers) wrote:
>
>> ECC is mandatory in nearly all HP servers. The signalling
>> characteristics of ECC memory are usually different, even though the
>> DIMM socket is physically the same... Ben Myers
>
>That's not quite correct. ECC memory uses the same signalling spec like
>non-ECC memory. All chipsets which support ECC also work with non-ECC
>memory. If a system works with non-ECC memory is a matter of the BIOS. If it
>can disable ECC verification the server will work with non-ECC memory as
>well...
>
>On the other side, there is a reason why servers use ECC. The OP should
>think twice about if sacrifying some safety for a few bucks is worth it...
>
>Benjamin
>

 

[Guest]

Archived from groups: comp.sys.hp.hardware (More info?)

ben_myers_spam_me_not @ charter.net (Ben Myers) wrote:

> The converse is definitely not the case. I acquired some new ECC
> registered buffered SDRAM DIMMs once in an auction lot, all good
> memory, tested OK in a server motherboard. When they were installed
> in an equally good quality Intel-branded desktop board, the board
> would not boot at all.
>
> If the differences between ECC and non-ECC are not due to signals,
> they are due to timing. Or the timing of signals, hence my assertion
> about different signalling. Different timing of signals or different
> signals?

Well, the timing is basically the same like on non-ECC DIMMs (assuming both
RAS/CAS are the same, i.e 3-2-2). The problem is the buffering, which
doesn't work with a lot of desktop boards. The buffers add additional signal
delay, and often the desktop boards don't work with buffered modules,
regardless is ECC or non-ECC. On the other side, some chipsets require
buffered modules and won't work with unbuffered memory due to the high loads
on the bus. But even this has nothing to do with ECC or non-ECC.

Then there are systems that only work with modules in a certain
organisation, or with special presence detection circuitry (to bind
customers to the manufacturer). Or computers that refuse to work with ECC
modules because of the systems firmware (i.e. old Macs)...

Benjamin