I have an XP Pro system with the A7N8X Deluxe board with a Barton
2500+ (not overclocked) running at its default speed (166 x 11). I
currently have two sticks of PC3200 256mb RAM (CAS 3) in slots 1 & 2
and a 512mb stick of PC3200 RAM (CAS 2.5) in slot 3. I have been led
to believe a congruent amount of like memory in slots 1+2 and 3 (the
combination of what is 1 and 2 should equal what is in 3) will allow
dual channel performance.

My dilemma is whether to add another 512mb RAM to the system. I am
finding I need to add memory in order to accommodate a variety of
memory-intensive applications which I run simultaneously and have open
for long stretches at a time. The system becomes somewhat sluggish as
the RAM ceiling is met and exceeded. There are some good deals on
PC3200 RAM at the moment and purchasing a couple sticks of 512mb would
allow the size upgrade desired. The tradeoff, I believe, is I would
lose dual channel performance as the congruency size-wise would be
lost (1024 and 512).

So, is the dropoff in performance from losing this dual channel
capability a greater factor than the benefit that would be derived by
adding 512mb of RAM to address the aforementioned conditions?

Thanks for your insight,


--Howie

"Howie" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
>I have an XP Pro system with the A7N8X Deluxe board with a Barton
> 2500+ (not overclocked) running at its default speed (166 x 11). I
> currently have two sticks of PC3200 256mb RAM (CAS 3) in slots 1 & 2
> and a 512mb stick of PC3200 RAM (CAS 2.5) in slot 3. I have been led
> to believe a congruent amount of like memory in slots 1+2 and 3 (the
> combination of what is 1 and 2 should equal what is in 3) will allow
> dual channel performance.
>
> My dilemma is whether to add another 512mb RAM to the system. I am
> finding I need to add memory in order to accommodate a variety of
> memory-intensive applications which I run simultaneously and have open
> for long stretches at a time. The system becomes somewhat sluggish as
> the RAM ceiling is met and exceeded. There are some good deals on
> PC3200 RAM at the moment and purchasing a couple sticks of 512mb would
> allow the size upgrade desired. The tradeoff, I believe, is I would
> lose dual channel performance as the congruency size-wise would be
> lost (1024 and 512).
>
> So, is the dropoff in performance from losing this dual channel
> capability a greater factor than the benefit that would be derived by
> adding 512mb of RAM to address the aforementioned conditions?
>
> Thanks for your insight,
>
>
> --Howie

Really tough to say without trying it. I would assume that dual channel
would be more efficient at 1g than non-dual channel at 1.5g. Just an
assumption though. The extra 512 is not a lot of gain anyway and I wouldn't
even bother to add it unless you have found that the 1g you have is being
fully used for sure and if so, I would add another gig. If the apps you are
using are utilizing the full gig you have, I would say that another 512
would be used fairly quickly. Video editing and some memory-intensive CAD
stuff are the only things I have seen that suck up memory like you are
seeing.

Ed

On Sun, 20 Nov 2005 17:26:54 GMT, "Ed Medlin" <(E-Mail Removed)> wrote:

>
>"Howie" <(E-Mail Removed)> wrote in message
>news:(E-Mail Removed)...
>>I have an XP Pro system with the A7N8X Deluxe board with a Barton
>> 2500+ (not overclocked) running at its default speed (166 x 11). I
>> currently have two sticks of PC3200 256mb RAM (CAS 3) in slots 1 & 2
>> and a 512mb stick of PC3200 RAM (CAS 2.5) in slot 3. I have been led
>> to believe a congruent amount of like memory in slots 1+2 and 3 (the
>> combination of what is 1 and 2 should equal what is in 3) will allow
>> dual channel performance.
>>
>> My dilemma is whether to add another 512mb RAM to the system. I am
>> finding I need to add memory in order to accommodate a variety of
>> memory-intensive applications which I run simultaneously and have open
>> for long stretches at a time. The system becomes somewhat sluggish as
>> the RAM ceiling is met and exceeded. There are some good deals on
>> PC3200 RAM at the moment and purchasing a couple sticks of 512mb would
>> allow the size upgrade desired. The tradeoff, I believe, is I would
>> lose dual channel performance as the congruency size-wise would be
>> lost (1024 and 512).
>>
>> So, is the dropoff in performance from losing this dual channel
>> capability a greater factor than the benefit that would be derived by
>> adding 512mb of RAM to address the aforementioned conditions?
>>
>> Thanks for your insight,
>>
>>
>> --Howie
>
>Really tough to say without trying it. I would assume that dual channel
>would be more efficient at 1g than non-dual channel at 1.5g. Just an
>assumption though. The extra 512 is not a lot of gain anyway and I wouldn't
>even bother to add it unless you have found that the 1g you have is being
>fully used for sure and if so, I would add another gig. If the apps you are
>using are utilizing the full gig you have, I would say that another 512
>would be used fairly quickly. Video editing and some memory-intensive CAD
>stuff are the only things I have seen that suck up memory like you are
>seeing.
>
>Ed
>

Thanks for the input. Believe it or not, I don't use the types of
applications one would normally identify as the "memory suckers" but
rather the following type of lineup:

Firefox (A dozen or so open tabs)
MS Outlook
IE
iTunes
MS Excel
Google Earth
MS Money
ActiveSync (Pocket PC synching software)
Zone Alarm
MS AntiSpyware
And a few others in the background...

I suspect part of the issue in bumping up against the current 1gb
ceiling is that some of these applications do not return resources
efficiently (at all?). I guess I am trying to avoid the need to
reboot every couple of days when I am using the system intensely for
all of (my) life's activities...

I know that 2gb is probably the best solution, but unfortunately I
don't have a 1gb stick already (read: cost prohibitive to buy 2gb of
RAM and receive little or nothing for my existing lineup of 1gb RAM).
The 1gb "kits" (2 sticks of 512mb) are priced nicely at the moment and
that is what triggered my thinking relative to upping things to 1.5gb
-- but if performance would be jeopardized due to losing the dual
channel feature, then perhaps my dilemma has been resolved.

Regards,


--Howie

You can't run dual channel mode with three sticks of memory, no matter
what the size. Dual channel requires two hence the term Dual. The
memory slots are configured to run in pairs. Thats why most dual
channel boards have colored memory slots. So you can make sure you
install the memory in the proper slot to get dual channel operation.

Now running a pair of 256 sticks in there proper slots for dual channel
and a pair of 512 sticks in there proper slots should give you Dual
Channel support.

You must also remember the sticks of ram must be by the same, make,
size and layout to run in dual channel mode. They do not need to be
Dual Channel kits but these kits also have added support for operation
in dual channel mode.

Custom Computers wrote:
> You can't run dual channel mode with three sticks of memory, no matter
> what the size. Dual channel requires two hence the term Dual. The
> memory slots are configured to run in pairs. Thats why most dual
> channel boards have colored memory slots. So you can make sure you
> install the memory in the proper slot to get dual channel operation.


You can run a perfect dual channel with 3 sticks if you only match them up
so that you get the same amount of memory in the 2 channels, just as OP is
doing right now.
If there is an imbalance in ram amount in the 2 channels, only parts of the
ram will run dual channel.


> Now running a pair of 256 sticks in there proper slots for dual
> channel and a pair of 512 sticks in there proper slots should give
> you Dual Channel support.

Not with nforce 2 that has 3 memory slots, 2 for the 1. channel, 1 for the
other.


> You must also remember the sticks of ram must be by the same, make,
> size and layout to run in dual channel mode. They do not need to be
> Dual Channel kits but these kits also have added support for operation
> in dual channel mode.

Not necessary at all. Currently I mix old PC2100, PC2700 and PC3200 in my
rig (all Kingston valueram). Works like a charm.

"Egil Solberg" <(E-Mail Removed)> wrote in message
news:4380fe2c$0$41143$(E-Mail Removed)...
> Custom Computers wrote:
>> You can't run dual channel mode with three sticks of memory, no matter
>> what the size. Dual channel requires two hence the term Dual. The
>> memory slots are configured to run in pairs. Thats why most dual
>> channel boards have colored memory slots. So you can make sure you
>> install the memory in the proper slot to get dual channel operation.
>
>
> You can run a perfect dual channel with 3 sticks if you only match them up
> so that you get the same amount of memory in the 2 channels, just as OP is
> doing right now.
> If there is an imbalance in ram amount in the 2 channels, only parts of
> the ram will run dual channel.
>
>
>> Now running a pair of 256 sticks in there proper slots for dual
>> channel and a pair of 512 sticks in there proper slots should give
>> you Dual Channel support.
>
> Not with nforce 2 that has 3 memory slots, 2 for the 1. channel, 1 for the
> other.
>
>
>> You must also remember the sticks of ram must be by the same, make,
>> size and layout to run in dual channel mode. They do not need to be
>> Dual Channel kits but these kits also have added support for operation
>> in dual channel mode.
>
> Not necessary at all. Currently I mix old PC2100, PC2700 and PC3200 in my
> rig (all Kingston valueram). Works like a charm.
>

Hmm...it may work, but not very efficiently. If you are mixing up different
speeds / brands / types, clock skew will be killing off most of your
performance. It'd be interesting if you benchmark your memory operation with
the different combinations and then post the results.

Wrong. Three sticks of RAM used in a motherboard precludes it running in
dual channel mode.

--
DaveW

----------------
"Howie" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
>I have an XP Pro system with the A7N8X Deluxe board with a Barton
> 2500+ (not overclocked) running at its default speed (166 x 11). I
> currently have two sticks of PC3200 256mb RAM (CAS 3) in slots 1 & 2
> and a 512mb stick of PC3200 RAM (CAS 2.5) in slot 3. I have been led
> to believe a congruent amount of like memory in slots 1+2 and 3 (the
> combination of what is 1 and 2 should equal what is in 3) will allow
> dual channel performance.
>
> My dilemma is whether to add another 512mb RAM to the system. I am
> finding I need to add memory in order to accommodate a variety of
> memory-intensive applications which I run simultaneously and have open
> for long stretches at a time. The system becomes somewhat sluggish as
> the RAM ceiling is met and exceeded. There are some good deals on
> PC3200 RAM at the moment and purchasing a couple sticks of 512mb would
> allow the size upgrade desired. The tradeoff, I believe, is I would
> lose dual channel performance as the congruency size-wise would be
> lost (1024 and 512).
>
> So, is the dropoff in performance from losing this dual channel
> capability a greater factor than the benefit that would be derived by
> adding 512mb of RAM to address the aforementioned conditions?
>
> Thanks for your insight,
>
>
> --Howie

"DaveW" <(E-Mail Removed)> wrote in message
news_ydnTqRevJBjRzeRVn-(E-Mail Removed)...
> Wrong. Three sticks of RAM used in a motherboard precludes it running in
> dual channel mode.
>
> --
> DaveW
>
You might want to first read the manual for that particular motherboard
(which only has 3 memory slots).

In article <9l7gf.21457$q%.(E-Mail Removed)>, "Greysky"
<(E-Mail Removed)> wrote:

> "Egil Solberg" <(E-Mail Removed)> wrote in message
> news:4380fe2c$0$41143$(E-Mail Removed)...
> > Custom Computers wrote:
> >> You can't run dual channel mode with three sticks of memory, no matter
> >> what the size. Dual channel requires two hence the term Dual. The
> >> memory slots are configured to run in pairs. Thats why most dual
> >> channel boards have colored memory slots. So you can make sure you
> >> install the memory in the proper slot to get dual channel operation.
> >
> >
> > You can run a perfect dual channel with 3 sticks if you only match them up
> > so that you get the same amount of memory in the 2 channels, just as OP is
> > doing right now.
> > If there is an imbalance in ram amount in the 2 channels, only parts of
> > the ram will run dual channel.
> >
> >
> >> Now running a pair of 256 sticks in there proper slots for dual
> >> channel and a pair of 512 sticks in there proper slots should give
> >> you Dual Channel support.
> >
> > Not with nforce 2 that has 3 memory slots, 2 for the 1. channel, 1 for the
> > other.
> >
> >
> >> You must also remember the sticks of ram must be by the same, make,
> >> size and layout to run in dual channel mode. They do not need to be
> >> Dual Channel kits but these kits also have added support for operation
> >> in dual channel mode.
> >
> > Not necessary at all. Currently I mix old PC2100, PC2700 and PC3200 in my
> > rig (all Kingston valueram). Works like a charm.
> >
>
> Hmm...it may work, but not very efficiently. If you are mixing up different
> speeds / brands / types, clock skew will be killing off most of your
> performance. It'd be interesting if you benchmark your memory operation with
> the different combinations and then post the results.

The Nforce2 is a very flexible chipset, in fact more flexible
than just about any other chipset (only some recent Intel
chipsets come close, and AFAIK, still don't support the
composite mode that the Nforce2 has.). Like many boards in the
past, if you mix different speeds of memory, it is up to the
BIOS to pick the slowest timing spec found in each SPD, and
use that to set the timing for all sticks. A PC3200 stick is
backward compatible to the lower rates, and can run all the way
down to PC1600 if need be. (Download a memory chip datasheet
from Micron if you don't believe this.)

The Nforce2 only needs to balance the total memory in each
channel, to benefit from dual channel operation. And the
Nforce2 still continues to run, unabated, when faced with
an unbalanced combination of memories.

I have a specially modified copy of memtest86 (a three line
mod in the program), in which I test the memory bandwidth
as a function of address space. If I insert 3x512MB DIMMs,
the lower 1024MB of memory has the dual channel bandwidth,
and the upper 512MB has the single channel bandwidth. Thus,
using an unbalanced configuration, where one channel has
more memory than the other, _still_ results in 2/3rds of
the address space being dual channel. It does the dual
channel thing, for as much memory as is common to both
channels. In the 3x512MB case, one channel has 512MB and
the other 1024MB, and both channels have at least 512MB
in common, and that common portion runs dual channel

channel0 channel1

512MB <--- Top portion runs single channel
512MB 512MB <--- This section runs dual channel

Also, if you consider the bandwidths involved for a moment,
the AthlonXP has a 64 bit bus, operating at FSB400, for a
bandwidth of 3200MB/sec. Two sticks of PC3200 give a total
bandwidth of 6400MB/sec. That means the AthlonXP cannot
unleash the full potential of that configuration. That is
why the difference seen in real applications, between
single channel and dual channel, is as small as it is.
If I really needed the memory, I wouldn't hesitate to run
3x512MB, and ran Linux that way for a while.

HTH,
Paul

On Mon, 21 Nov 2005 00:54:50 GMT, (E-Mail Removed) (Paul) wrote:


>
>The Nforce2 is a very flexible chipset, in fact more flexible
>than just about any other chipset (only some recent Intel
>chipsets come close, and AFAIK, still don't support the
>composite mode that the Nforce2 has.). Like many boards in the
>past, if you mix different speeds of memory, it is up to the
>BIOS to pick the slowest timing spec found in each SPD, and
>use that to set the timing for all sticks. A PC3200 stick is
>backward compatible to the lower rates, and can run all the way
>down to PC1600 if need be. (Download a memory chip datasheet
>from Micron if you don't believe this.)
>
>The Nforce2 only needs to balance the total memory in each
>channel, to benefit from dual channel operation. And the
>Nforce2 still continues to run, unabated, when faced with
>an unbalanced combination of memories.
>
>I have a specially modified copy of memtest86 (a three line
>mod in the program), in which I test the memory bandwidth
>as a function of address space. If I insert 3x512MB DIMMs,
>the lower 1024MB of memory has the dual channel bandwidth,
>and the upper 512MB has the single channel bandwidth. Thus,
>using an unbalanced configuration, where one channel has
>more memory than the other, _still_ results in 2/3rds of
>the address space being dual channel. It does the dual
>channel thing, for as much memory as is common to both
>channels. In the 3x512MB case, one channel has 512MB and
>the other 1024MB, and both channels have at least 512MB
>in common, and that common portion runs dual channel
>
> channel0 channel1
>
> 512MB <--- Top portion runs single channel
> 512MB 512MB <--- This section runs dual channel
>
>Also, if you consider the bandwidths involved for a moment,
>the AthlonXP has a 64 bit bus, operating at FSB400, for a
>bandwidth of 3200MB/sec. Two sticks of PC3200 give a total
>bandwidth of 6400MB/sec. That means the AthlonXP cannot
>unleash the full potential of that configuration. That is
>why the difference seen in real applications, between
>single channel and dual channel, is as small as it is.
>If I really needed the memory, I wouldn't hesitate to run
>3x512MB, and ran Linux that way for a while.
>
>HTH,
> Paul

Paul,

Your input does help. Thanks. I have ordered a pair of Patriot
PC3200 512mb sticks...

Thanks to the others who contributed as well, save for those
contending dual channel performance is not available with three sticks
of RAM. Posting information as fact when done from a base of
ignorance is irresponsible when there are legitimate questions in need
of valid input.

Regards,


--Howie