"Nick Ford" said:
Hi,
Here's my system---
Asus A7N8X rev.2.0 Deluxe with BIOS flashed to 1008
AthlonXP 2500+ overclocked at 11 X 200Mz (POST as XP3200+)
Windows XP Professional with Service Pack 2
MAXTOR 120 GB @7200 rpm
Corsair XMS 1GB (2x512) DDR Dual Channel PC3200 (DDR400)
About every third boot, I'll hear "System Failed Memory Test. Computer Now
Booting from Operating System." On one boot, the POST Memory Test wouldn't
launch at all! It will always boot into the OS fine, but the system will
lock up during such activities as virus checking (Norton 2003), and games
(Rome: Total War).
This is the third time I've replaced the memory. Prior to the Corsair, I
had Dual Channel OCZ PC3200, and prior to that Geil Dual Channel PC3200.
All new stuff (mostly from Egghead). I've tried running the CPU at specs
(and lower). I've set the memory setting back to optimal. But nothing
seems to help. (I've run the Geil memory in another system, with no
problems).
What am I dealing with here? Is it a problem with the board itself? Any
help would be greatly appreciated!
Nick
OK. First off, there are some hacked BIOS here. I use a Trats
modified BIOS for my A7N8X-E and had no trouble with it. This
thread seems to be for A7N8X Deluxe 2.0, but obviously you want
to be really careful that it is in fact the right BIOS.
http://forums.pcper.com/showthread.php?t=349979
You didn't give details about your RAM, but what I noticed on
my board, is using CAS2 memory made all the difference in
the world. CAS3 PC3200 didn't seem to want to work higher than
about 185MHz. There are a couple of reports of CAS3 memory
working in this newsgroup, but I really cannot tell you how common
it is to get CAS3 working at the full DDR400 rate.
Your options are:
1) Run single channel. Put both sticks on one channel. The
other channel remains electrically idle, and the reduction in
electrical noise allows a bit more clock speed to be used.
I didn't find this worked well enough with my CAS3 memory
to make this solution acceptable.
2) Run dual channel but select "83%" for memory speed. That
would allow the FSB to clock at 200MHz, while the memory
clocks at 166MHz. Test with memtest86+ and Prime95, to ensure
that it works error free.
3) Use a mobile processor, adjust multiplier and CPU clock to
get the best from processor and memory at the same time.
This means you are willing to accept a reduced memory bandwidth.
In this case, your FSB and memory remain in sync (100% memory
setting), but you adjust the bus clock to its max stable,
then set the multiplier to get 2200MHz or so for the core.
If a multiplier over 12.5x is required, a "wire trick" may
be needed for the most significant multiplier bit.
4) Use a modified BIOS. There is a setting in the Northbridge
that selects either Command Rate 1T (so called Command Per
Clock "on"). All official Asus BIOS run at the more aggressive
Command Rate 1T. There are modified BIOS, where a permanent
selection of Command Rate 2T is used. This reduces the memory
bandwidth, but allows the FSB to be raised maybe 30MHz. So,
if you were stable at 185MHz before, perhaps you could squeeze
more than 210MHz out of it. This might be a preferred solution
if you are not using a Mobile, like a locked Barton.
5) Buy some CAS2 memory. Be happy. Mine runs 200Mhz 2-2-2-5 dual
channel. Processot 1.65V, Vdimm 2.7V or so.
What is the Command Rate setting ? A good question. It is a
timing trick. When Command Rate is set to 2T, a command being
sent to the memory is present for two clock cycles. A signal
used to strobe the information into the memory, is only activated
on the second cycle. The first cycle gives plenty of time for
the command to settle on the bus - in a sense, it is almost like
the clock rate to the memory is dropped just long enough to
communicate with the memory, then the speed picks up again
for the data phase. When the I/O pads on the Northbridge cannot
charge the command onto the bus in time, the Command Rate 2T
(CPC "off") option, improves the interface timing, at the
expense of wasting a bus cycle. Wasted bus cycles drop the
"efficiency" of the bus, so while the interface is running at
DDR400, the bandwidth you get is reduced by the number of
wasted bus cycles needed.
The other question I'd like answered, is why does buying CAS2
memory work ? The memory is supposed to be a synchronous device,
relying only on the clock for timing. There is no logical reason
for the CAS2 memory to be making a difference, and the implication
is, there is some asynchronous behavior inside the memory chip
(the chip is responding before the clock signal rises ?). I don't
understand enough about the internal operation of the memory,
to be able to say why going from CAS3 to CAS2 makes a difference.
As for CAS2.5, my experiments seem to suggest that in fact the
chipset cannot run at CAS2.5. If you get some CAS2 memory, try
all the different CAS values and measure memory bandwidth with
the bandwidth indicator on the memtest86+ test screen. I found
that 2.5-3-3-7 gave the same results as 3-3-3-8, as set in
the BIOS.
HTH,
Paul
