Archived from groups: alt.comp.periphs.mainboard.asus (
More info?)
Each module has one terminated channel, one unterminated channel. The
modules are arranged (installed) in "opposite directions" (physically,
although they are parallel to each other on the board, one is "flipped"
relative to the other -- one is "face up", if you will, the other is
"face down"). Each module terminates the channel that is unterminated
on the other module. Both modules are identical, and a terminator
module terminates the same "channel" as a live module. Beyond that,
it's not easy to explain without being able to sit down together with a
piece of paper at a table.
Paul wrote:
> In article <JbPsd.17912$4m5.10310@fe1.columbus.rr.com>, Barry Watzman
> <WatzmanNOSPAM@neo.rr.com> wrote:
>
>
>>The P4T533-C uses 16-bit RDRAM modules (184 pins), the P4T533 (no-C) use
>>32-bit modules (232 pins). 16 bit modules are always used in pairs in
>>P4 systems (but singly in Pentium 3 systems), and P4 motherboards using
>>them have 4 sockets (e.g. 1 or 2 pairs). The 32-bit modules are used
>>singly, and motherboards that use them only have 2 sockets.
>>
>>You can tell the module width because the 232-pin sockets have pins in
>>the middle and only one notch (unless someone has cut out the space for
>>a 2nd notch, trying to use the 232-bit module in a 184-pin motherboard.
>> This won't work, but if it's done carefully, it won't damage the
>>module). The 184 pin modules have 2 notches and an "empty" area (no
>>pins) between them, in the middle.
>>
>>In all cases, empty slots must be filled with C-RIMM terminator modules
>>(which are different for 16-bit vs. 32-bit modules).
>>
>>Paul's comments about the architecture of the 32-bit modules and how
>>they work is incorrect, but the explanation is too complex to go into
>>here (and, fortunately, is irrelevant to the end-user).
>>
>>Generally, the modules within a pair must be architecturally identical
>>(including the number of chips per module, see below), but the two pairs
>>can be different. I'm not 100.0% sure that applies to ECC, but it does
>>apply to most everything else. [In other words, I THINK you can have
>>one pair of ECC and one pair of non-ECC, but I'm not certain -- and it
>>may depend on the BIOS, because you may need to turn off the ECC
>>manually to do so].
>>
>>As to other parameters, there are only a few that matter:
>>
>>-Total module size must be matched between the two modules within a pair
>>but may be different between the pairs. For example, you can mix two
>>128's and two 256's to get 768 megs of total memory.
>>
>>-"PC" speed -- You need PC800 for a CPU with a 100/400MHz front side
>>bus, or PC1066 for a 133/533MHz front side bus.
>>
>>-Access time -- For a 100/400 FSB on an Intel 850 chipset, 45 nSec is
>>adequate, but for an 850e, you need 40 nSec (both designated as
>>"PC800"). For PC1066, you need 32 nSec (there are some 35 nSec PC1066
>>modules, avoid them). Faster access times (smaller numbers) are
>>"backwards compatible".
>>
>>-Internal module architecture - this is occasionally, but rarely,
>>relevant. For example, a 128 meg module can be made with eight chips of
>>16 megs each or 4 chips of 32 megs each. This only if you are going for
>>really high memory capacity (there is a limit on the total number of
>>chips, but it's fairly high -- 32, I think). Other things being equal,
>>modules of the same size with fewer chips are better (but, on the other
>>hand, 820 and 840 chipsets (Pentium III) can't recognize the newer,
>>largest chips, and in all cases, this needs to be matched between the
>>modules comprising a pair). This is shown as 128/8 or 128/4, where the
>>number after the slash give you the number of chips inside the module.
>>
>>Hope that helps.
>>
>
> I love a challenge, Barry
>
> Perhaps my mistake on the 32 bit modules, is the fact that the
> termination is on the module ?
>
>
http://rambus.com/products/rdram/documentation/32BitRIMM_w256Mb_d_0164_V0_1.pdf
>
> That datasheet suggests the wiring looks like:
>
> P4T533c P4T533
>
> 16 bit 16 bit 16 bit
> | | | | | | T T
> 184pinModule 184pinModule 232pinModule
> | | | | | | | |
> 184pinModule 184pinModule 232pinModule
> T T T T T T | |
> 16 bit
>
> On the 232 pin module, there are only three 16 bit busses. By connecting
> the modules back to back, and limiting the architecture to a two
> module pair, the need for a fourth 16 bit bus is avoided.
>
> References:
>
> Page 7 shows the pinout of a 232 pin module, and suggested the connection
> diagram on the right above...
>
http://rambus.com/products/rdram/documentation/32BitRIMM_w256Mb_d_0164_V0_1.pdf
>
> Page 5 shows the pinout of a 184 pin module, and suggests to me the
> connectivity on the left of the diagram above.
>
http://rambus.com/downloads/RIMM256_d_0110_V1_1.pdf
>
> This Intel reference schematic verifies the left diagram. There are
> two channels, with two RIMMs each.
> page15 (850) ch1 --> page18 (rimm1) --> page 20 (rimm2 + ext 28ohm term)
> page15 (850) ch2 --> page19 (rimm1) --> page 21 (rimm2 + ext 28ohm term)
>
>
ftp://download.intel.com/design/chipsets/designex/298245_CRB.pdf
>
> I cannot find an Intel ref schematic for the P4T533 type design.
>
> Another thing I'm having trouble with, is understanding the
> pinout of the 184pin module. The P4T533-C has the two modules
> in a channel oriented the same, as opposed to having the modules
> facing one another. And yet, if you look at the copper on the
> PCB, the signals look like they just flow through. That means
> the use of "L" and "R" between the two modules in a channel
> isn't consistent, and looks something like this:
>
> P4T533c P4T533
>
> 16 bit 16 bit 16 bit
> | | | | | |
> A2 L| | | | | | T T <--on module
> 184pinModule 184pinModule 232pinModule
> A83 R| | | | | | | |
> A83 R| | | | | | | |
> 184pinModule 184pinModule 232pinModule
> A2 L| | | | T T | |
> T T T T <-- on mobo | |
> 16 bit
>
> Compare to the motherboard pictures in the manuals:
>
http://www.asus.com/pub/ASUS/mb/sock478/p4t533/e1152_p4t533.pdf (pg.17)
>
http://www.asus.com/pub/ASUS/mb/sock478/p4t533-c/e1151_p4t533-c.pdf (pg.17)
>
> Great fun!
> Paul
>
>
>>Racer X wrote:
>>
>>>Paul,
>>>You are unbelievable!!
>>>Thanks so much for your attention to this matter. It was a great help.
>>>
>>>