Re: P5W DH Deluxe - Memory Bus Speed & RAM Timings

Bob Huntley wrote:
Hi,

I've just bought a new PC which came with a Asus P5W DH motherboard and Corsair CM2X1024-8500C52 DDR2 RAM (Dominator RAM with their new premium heat sinks).

The RAM is supposed to be capable of running at a bus speed of 266 MHz, but it was set to 200 on arrival and I can't get it to run stably over 222 MHz. The RAM is running at the correct latency settings for that speed, and is also set to the correct voltage.

I've checked the Crucial web site, and have one clue. The RAM is rated at 266 at a 2T Command Rate - but according to Sisoft Sandra the Asus motherboard is running it with a 1T rate. I can't find anything in the BIOS that alters the Command Rate - does anyone know how to do this?

Thanks in advance - and happy Xmas for tomorrow.

Bob,


Official support for Corsair RAM is this forum. Staff from Corsair
can answer your questions. The main site is corsairmicro.com , for
datasheets and the like.

http://www.houseofhelp.com/forums/forumdisplay.php?forumid=128

One way to set Command Rate, is to use an Athlon64 AM2 socket processor :-)

Intel chipsets generally don't expose the Command Rate setting. Maybe
there is a secret location in the memory controller registers for something
like that, but it is not something you normally see in the BIOS
for an Intel chipset. An Xbitlabs article, suggested the Intel chipset
always runs at Command Rate 2T (article was back in the DDR days), but
from a performance perspective, it just doesn't seem reasonable. But
one thing I have noticed, at least with DDR and my Intel chipset, is
performance did not change one iota, when I went from two matched sticks
to four matched sticks. That tells you that the same command rate was
used by the hardware, for both cases, and since memory bus loading with
four sticks generally requires 2T to work well, it suggests to me that
Xbitlabs may be right. What I cannot figure out, is why Intel doesn't
suffer more, from a performance perspective, if that is indeed what they
are doing.

At extremely high clock speeds, even with one DIMM, there is a need
to use 2T command rate. Perhaps having a look around the Anandtech
forums, for articles where they test the elite DDR2 memory, may give
some info on what to expect in that regard.

AMD review articles show the Command Rate 1T or 2T. 2T might be
handy for four sticks.

http://www.anandtech.com/mb/showdoc.aspx?i=2819&p=4

Intel (+ Intel chipset) review articles don't even have room for a command
rate value. Notice how CPUZ leaves the Command Rate field blank.

http://www.anandtech.com/mb/showdoc.aspx?i=2871&p=7

There is a command rate reported on RD600 from ATI. CPUZ does
have a Command Rate field with RD600.

http://www.anandtech.com/mb/showdoc.aspx?i=2891&p=3

Here is an Intel processor (+ 650i Nvidia chipset) and it
also shows Command Rate. The accompanying text states they
used 2T to get above DDR2-800.

http://www.anandtech.com/mb/showdoc.aspx?i=2894&p=3

Have you verified the settings with the latest version of CPUZ ?
If you've just been looking in the BIOS, you should boot into
Windows (use timings that you know are stable, not the bleeding
edge settings), and verify that you understand the timings right.
Overclocking the FSB, affects the memory also. Just in case
you've actually pushed past the spec value.

At this point, I'd give the Corsair official forums a try,
and ask them exactly how you are supposed to verify that
you are using Command Rate 2T and are doing all the right
thing on your 975X based motherboard.

You can look on PDF page 90 here, for a Command Rate
bitfield in the 975X registers. This is the 975X Northbridge
datasheet. I've tried looking for this before, on some
other Intel chipset, but did not find what I was
looking for.

ftp://download.intel.com/design/chipsets/datashts/31015801.pdf

I just thought of another test you can run. If you place the
two memory sticks on the same channel (i.e. that would be a
single channel mode), does Sisoft Sandra still report a
1T command rate. You'd think, if the chipset was running
1T in dual channel mode with two sticks, it would have to
switch to 2T to run single channel with two loads. See if
Sandra can tell the difference. It the setting does not
change, chances are Sandra is wrong.

HTH,
Paul
.