The system has 3 slots, and 1 is occupied by a 128MB 9-chip single sided
DIMM, so it has ECC but is apparently unbuffered.
I want to bring it up to 512MB, or 640MB if possible.
The Dell website has 2 RAM options listed for the GX1 -- 1 ECC registered
and 1 non-ECC unbuffered.
This seems strange since I thought mobos take either registered or
unregistered but not both. So if it can take both then wouldn't they have
to match, and if all slots were filled then would it need buffered RAM to
reduce loading of the clocks?
Another thing is the installed module's labelling. It's PC100-222-620. The
middle digits are well documented but the last 3 digits are not.
Thanks for any clarification.
--
Reply in group, but if emailing add another
zero, and remove the last word.
Tom Del Rosso wrote:
> The system has 3 slots, and 1 is occupied by a 128MB 9-chip single sided
> DIMM, so it has ECC but is apparently unbuffered.
>
> I want to bring it up to 512MB, or 640MB if possible.
>
> The Dell website has 2 RAM options listed for the GX1 -- 1 ECC registered
> and 1 non-ECC unbuffered.
>
> This seems strange since I thought mobos take either registered or
> unregistered but not both. So if it can take both then wouldn't they have
> to match, and if all slots were filled then would it need buffered RAM to
> reduce loading of the clocks?
>
> Another thing is the installed module's labelling. It's PC100-222-620. The
> middle digits are well documented but the last 3 digits are not.
>
> Thanks for any clarification.
>
>
The Dell specs I see don't mention registered, only ECC or not. They do
seem to indicate that ECC is required to use 256MB DIMMs, which is odd
unless they are buffered. http://support.dell.com/support/edoc...gx1/Memory.htm
However, Crucial seems to feel that unbuffered is all that is required. http://www.crucial.com/store/listpar...0MHz+processor
I'd go with Crucial, since they do guarantee compatibility.
Some motherboard chipsets are capable of handling both ECC registered
and non-ECC unbuffered. The ECC registered stuff is/was used mostly in
servers, most of which are now decommissioned, so it is even cheaper than dirt
lately.
If you are running one of the dumbed down Windows operating systems, 95, 98, or
ME, 512MB is the maximum supported. If you are running Linux, NT, 2000, or XP,
the board handles 768MB max.
I recommend all one type of memory or the other type, else the motheboard might
get a bit confused... Ben Myers
On Fri, 15 Jun 2007 12:45:27 -0400, "Tom Del Rosso" <td_01@att.net.invalid>
wrote:
>The system has 3 slots, and 1 is occupied by a 128MB 9-chip single sided
>DIMM, so it has ECC but is apparently unbuffered.
>
>I want to bring it up to 512MB, or 640MB if possible.
>
>The Dell website has 2 RAM options listed for the GX1 -- 1 ECC registered
>and 1 non-ECC unbuffered.
>
>This seems strange since I thought mobos take either registered or
>unregistered but not both. So if it can take both then wouldn't they have
>to match, and if all slots were filled then would it need buffered RAM to
>reduce loading of the clocks?
>
>Another thing is the installed module's labelling. It's PC100-222-620. The
>middle digits are well documented but the last 3 digits are not.
>
>Thanks for any clarification.
"Ben Myers" <ben_myers_spam_me_not@charter.net> wrote in message
news:ibi573d7rcc13t091il9kdsl5ucsbj2bb0@4ax.com
> Some motherboard chipsets are capable of handling both ECC registered
> and non-ECC unbuffered. The ECC registered stuff is/was used mostly
> in servers, most of which are now decommissioned, so it is even
> cheaper than dirt lately.
How come ECC often (apparently always?) coincides with registered? The
registered memory reduces the load on the clock lines by regenerating the
clock with a built-in PLL, and uses some kind of register pipeline to keep
it's signals in sync. The application note at
"Pen" <nospam@nospam.com> wrote in message
news:faydnQJP3qCpV-_bnZ2dnUVZ_tSunZ2d@comcast.com
>>
> The Dell specs I see don't mention registered, only ECC or not. They
> do seem to indicate that ECC is required to use 256MB DIMMs, which is
> odd unless they are buffered.
Why is that odd, 'unless they are buffered'? This is sort of academic now,
because you gave me enough to buy one, but I'd like to know why ECC and
registered go together.
This page lists 2 available upgrade modules, both 256MB, but one
ECC/registered, and the other non-ECC/unbuffered. (Not that I would buy
from them.)
The following is copied from the the Technicnical Specifications manual I
still have from an old GX1 (still in service)
Memory
Architecture: 64-bit (non-ECC) or 72-bit (ECC), noninterleaved,
"PC100" 100 MHz
DIMM sockets: three
DIMM capacities: 32-, 64-, 128-, and 256-MBSDRAM
Standard RAM: 32 MB (GX1) or 128 MB (GX1p) minimum
Maximum RAM: 768 MB
BIOS address: F0000h
It's been a long time since I had hands-on this box but I do remember it is
a PII-350, the last of the PIIs befort the PIII 400
Dean
"Tom Del Rosso" <td_01@att.net.invalid> wrote in message
news:Zozci.27$fP7.6@newsfe12.lga...
> The system has 3 slots, and 1 is occupied by a 128MB 9-chip single sided
> DIMM, so it has ECC but is apparently unbuffered.
>
> I want to bring it up to 512MB, or 640MB if possible.
>
> The Dell website has 2 RAM options listed for the GX1 -- 1 ECC registered
> and 1 non-ECC unbuffered.
>
> This seems strange since I thought mobos take either registered or
> unregistered but not both. So if it can take both then wouldn't they have
> to match, and if all slots were filled then would it need buffered RAM to
> reduce loading of the clocks?
>
> Another thing is the installed module's labelling. It's PC100-222-620.
> The
> middle digits are well documented but the last 3 digits are not.
>
> Thanks for any clarification.
>
>
> --
>
> Reply in group, but if emailing add another
> zero, and remove the last word.
>
>
In article <u%Dci.245$fP7.234@newsfe12.lga>, td_01@att.net.invalid
says...
> "Ben Myers" <ben_myers_spam_me_not@charter.net> wrote in message
> news:ibi573d7rcc13t091il9kdsl5ucsbj2bb0@4ax.com
> > Some motherboard chipsets are capable of handling both ECC registered
> > and non-ECC unbuffered. The ECC registered stuff is/was used mostly
> > in servers, most of which are now decommissioned, so it is even
> > cheaper than dirt lately.
1GB registered ECC sticks are still rather expensive ($125 for DDR1
for my Tyan S2875S) compared to $40 for DDR2 unregistered nonparity
for my laptop.
> How come ECC often (apparently always?) coincides with registered? The
> registered memory reduces the load on the clock lines by regenerating the
> clock with a built-in PLL, and uses some kind of register pipeline to keep
> it's signals in sync. The application note at
>
> http://www.pericom.com/pdf/applications/AN009.pdf
>
> says registered RAM is needed when using many modules, but I don't see what
> that has to do with ECC.
>
Neither are *required* for home desktop machines. Both are required
for servers and most workstations. There is little in the middle.
There is no technical reason, just what the market wants; cheap or
good. ;-)
"krw" <krw@att.bizzzz> wrote in message
news:MPG.20dd0e8fcef004e998a680@news.individual.ne t
>
>> How come ECC often (apparently always?) coincides with registered?
>> The registered memory reduces the load on the clock lines by
>> regenerating the clock with a built-in PLL, and uses some kind of
>> register pipeline to keep it's signals in sync. The application
>> note at
>>
>> http://www.pericom.com/pdf/applications/AN009.pdf
>>
>> says registered RAM is needed when using many modules, but I don't
>> see what that has to do with ECC.
>>
> Neither are *required* for home desktop machines. Both are required
> for servers and most workstations. There is little in the middle.
> There is no technical reason, just what the market wants; cheap or
> good. ;-)
That makes sense. The same reason SCSI drives are made better. Thanks.
--
Reply in group, but if emailing add another
zero, and remove the last word.
Yes! My desktop system uses SCSI drives, and nearly all of my earlier ones
have, too. Also a SCSI scanner and a SCSI read-write-all-types-of-flash-cards
device. A little noisier than PATA/SATA, and much slower to boot up, but I
would not have it any other way... Ben Myers
On Tue, 19 Jun 2007 02:58:35 -0400, "Tom Del Rosso" <td_01@att.net.invalid>
wrote:
>"krw" <krw@att.bizzzz> wrote in message
>news:MPG.20dd0e8fcef004e998a680@news.individual.n et
>>
>>> How come ECC often (apparently always?) coincides with registered?
>>> The registered memory reduces the load on the clock lines by
>>> regenerating the clock with a built-in PLL, and uses some kind of
>>> register pipeline to keep it's signals in sync. The application
>>> note at
>>>
>>> http://www.pericom.com/pdf/applications/AN009.pdf
>>>
>>> says registered RAM is needed when using many modules, but I don't
>>> see what that has to do with ECC.
>>>
>> Neither are *required* for home desktop machines. Both are required
>> for servers and most workstations. There is little in the middle.
>> There is no technical reason, just what the market wants; cheap or
>> good. ;-)
>
>That makes sense. The same reason SCSI drives are made better. Thanks.
In article <F8Ldi.293$ys7.263@newsfe12.lga>, td_01@att.net.invalid
says...
> "krw" <krw@att.bizzzz> wrote in message
> news:MPG.20dd0e8fcef004e998a680@news.individual.ne t
> >
> >> How come ECC often (apparently always?) coincides with registered?
> >> The registered memory reduces the load on the clock lines by
> >> regenerating the clock with a built-in PLL, and uses some kind of
> >> register pipeline to keep it's signals in sync. The application
> >> note at
> >>
> >> http://www.pericom.com/pdf/applications/AN009.pdf
> >>
> >> says registered RAM is needed when using many modules, but I don't
> >> see what that has to do with ECC.
> >>
> > Neither are *required* for home desktop machines. Both are required
> > for servers and most workstations. There is little in the middle.
> > There is no technical reason, just what the market wants; cheap or
> > good. ;-)
>
> That makes sense. The same reason SCSI drives are made better. Thanks.
>
Actually, SCSI drives are not made better. They're often *exactly*
the same drive with a SCSI controller board slapped on, in place of
the IDE board.
Dell Optiplex GX1 RAM 
Linear Mode
