AMD/939: four double-sided DIMMs cause problems

[John Doe]

Reading through the archive, I've noticed this problem come up and
(correct me if I'm wrong) but seen it dismissed. Looking at some
mainboard manuals, most common mainboards do not support four
double-sided DIMM memory modules.

Apparently using four double-sided DIMM memory modules will at best
constrain memory bandwidth to 333 MHz on most mainboards.

Specifically, I'm talking about AMD socket 939 boards with four
memory slots.

 

[Paul]

Reading through the archive, I've noticed this problem come up and
(correct me if I'm wrong) but seen it dismissed. Looking at some
mainboard manuals, most common mainboards do not support four
double-sided DIMM memory modules.

Apparently using four double-sided DIMM memory modules will at best
constrain memory bandwidth to 333 MHz on most mainboards.

Specifically, I'm talking about AMD socket 939 boards with four
memory slots.

On an Intel chipset motherboard, where the Northbridge
chip drives the memory, I've never had any problems whether
using two or four DDR DIMMs. That means it is possible to
have four DIMM slots without a problem. So four DIMMs is not
universally a bad thing.

On the AMD S939, the memory controller is inside the processor.
Putting too much circuitry inside one chip, may involve compromises.
The drivers for the memory bus may not be quite the same as
would be used on a separate Northbridge chipset.

I believe there are people using four sticks on S939 boards,
and whether the memory runs at DDR400 Command Rate 2T or
DDR333 Command Rate 1T, I expect the computer will still
get the job done. There are some BIOS on the AMD boards,
that try to enforce the DDR333 setting, but you'll have
to check the archives for evidence of that shortcoming.
The way it is supposed to work, is the BIOS selects whatever
it feels like, when the Auto setting is used. But a good
BIOS will allow the user to experiment with Manual settings.

The Command Rate 2T setting, is a compensation for loading.
Using Command Rate 2T allows more time for the memory address
bus to settle down, before the address is strobed into the
memory. When four DIMMs are used, that loads the memory bus,
and slows down the signals. When the signals are slowed,
then the setup time to the memory chips is violated. By
selecting the Command Rate 2T option, that allows a bit
more time to meet setup at the memory chips.

Combining the cranking down of the memory speed, and/or the
use of the Command Rate 2T memory setting, should allow
your four DIMMs to be used, if you wish.

Many people bitch and moan about running at DDR333, but the
computer will still work, with only a small performance
loss. And unless you've got the stopwatch out, you may not
even notice the difference.

Paul

 

[John Doe]

On an Intel chipset motherboard, where the Northbridge chip drives
the memory, I've never had any problems whether using two or four
DDR DIMMs. That means it is possible to have four DIMM slots
without a problem. So four DIMMs is not universally a bad thing.

Recognizing there are different universes is one reason I included
"AMD/939" in the subject line.

On the AMD S939, the memory controller is inside the processor.
Putting too much circuitry inside one chip, may involve
compromises. The drivers for the memory bus may not be quite the
same as would be used on a separate Northbridge chipset.

I believe there are people using four sticks on S939 boards, and
whether the memory runs at DDR400 Command Rate 2T or DDR333
Command Rate 1T,

Or maybe it has to run at DDR333 Command Rate 2T.

I expect the computer will still get the job done. There are some
BIOS on the AMD boards, that try to enforce the DDR333 setting,
but you'll have to check the archives for evidence of that
shortcoming.

Or you can just look in mainboard manuals like I've been doing.

It's not something you can correct in the BIOS, except as stated in
the owner's manual.

There might be exceptions, but it appears to be widespread.

Many people bitch and moan about running at DDR333,

And you don't even get paid for reading it.
Oh well.

but the computer will still work, with only a small performance
loss. And unless you've got the stopwatch out, you may not even
notice the difference.

I noticed an improvement going from 266 to 400. I'm sure they had a
good reason for designing and manufacturing PC3200 memory.

I think it's probably better to tell the person to look in his (or
her) owner's manual than it is to suggest his problem doesn't exist.